CA2120520A1 - A radio frequency local area network - Google Patents

Abstract

An apparatus and a method for routing data in a radio data communication system having one or more host computers (10), one or more intermediate base statios (20-52), and one or more RF terminals (100) organizes the intermediate base stations into an optimal spanning-tree network to control the routing of data to and from the RF terminals and the host computer efficiently and dynamically. Communication between the host computer and the RF terminals is achieved by using the network of intermediate base stations to transmit the data.

Description

~WC9 93/07691 2 l 2 ~ ~ h a PCI'/US92/08610 TIq!l.15: ~ RADIO FREQ~J~ICY LOS:~AI. ARl :A ~:~ORR

~A~RGRO~D CSF ~B I~ ~ION
In a typical radio data communication system having on~ or more host computers ~nd multiLple RF
termin 15, communication between a host computer and an RF terminal is provided by one or more base stations. Depending upon ~he application and the operating conditions, a large number of thes~ b~se stations may be required to adequately serve the systeDl. For example, a radio data c:ommunication system in~;talled in a large factory may re~uire dozens , - :~; 10 of bc,se stations in order to co~er the entire factory î loor .
In earlier RF data commlmlca~ ion systems, the bas~ 6ta~ions were typically co~ulected dir2ctly to a : host computer through mlllti-dropped conne~tions to an ;5 Ethexnet: communic:ation line. To c:ommunicate between an ~ terminal and a host computer, in ~;uch z system, the RF terminal sends d~ta to a base station and the base ~station passes the data direc ly to the host computer. Communicating with a host c~mputer through 2 0 ~ a base station in this manner is commonly known as holpping. These earlier RF data c~mmunicatic~n systems ~; ~used a ~ingle-hop method of communication.
In ord~3r to c:over a larger area with an RF data comm~nic:atiorl ~3ystem and to take advantage of the deregulation of the spread-spectrum radio fre~encies, later-developed RF data ccDunication æy~tems are : organiæed into layers OI base sta*ions~ As in earlier :~ :

WO93/~7691PCT/US92/08610 2 ~ S 2 RF data communications systems, a typical system includes multiple base stations which communicate directly with the RF terminals and the host c~mputer.
In addition, the system al~o includes intermediate 5stations that communicate with the RF terminalsj the multiple base stations, and other inter~ediate stations. In such a system, communication from an RF
terminal to a host computer may be achieved, for example, by ha~ing the RF terminal send data to an ::~: 10intermediat~ station, the intermediate station send the data to a base station, and the base station send the data directly to the host computer. Communicating with a~host:computer through more than one station is commonly known as a multiple-hop communication system.
15Difficulties often arise in maintaining the integrity of such multiple-hop RF data communication : systems.: The system must be able to handle both wireless~ and :hard-wired station connections, ;;efficient :dynamic routing of data information, RF
20terminal mobility, and interference from many dif f erent sources .

~: : ; : ::

~093/07691 2 ~ 2 ~ PCT/US92/08610 _UNNARY OF_TH~ ~NVBN~O~
The present invention solves many of the problems inherent in a multiple-hop data communication system.
The present invention comprises an RF Local-Area Network capable of afficient and dynamic handling of : data by routing communications between the ~F
Terminals and the host computer through a network of ~ int~rmediate base~stations.
:~ ~: In one embodiment of the present invention, the RF data:communication system contains one or more host - computers~ and multipIe gateways, br~dges, and RF
terminals.: Gateways are used to pass messages to and from;a host computer and the RF Ne work. A host port is used~to provide a link between the gateway and the host computer. In addition, gateways ma~ include bridging~functions and may pass information from one RF terminal to another. Bridges are intermediate relay~nodes~which repeat data messages. Bridges can : repeat~data~; to :and: from bridges, gateways and RF
: ::terminals~and:are~ ùsed to extend the range of the gateways~
The~RF~ erminals are attached logically ~o the :host~computer~ and::use:a network formed by a gateway and~ hé~bridges:to communicate with the host computer.
2~5~ :To~ set~up~the network, an optimal configuration for conducting~;Detwork communication spa~ning~tree is : created:to:control the flow of data communication. To : aid:~ understanding by providing a more visual description, ~:this configuration is referred to hereafter~as:a "spanning tree" or "optimal spanning tree".~
Specifically, root of the ~p~nning tree are the : gateways;~ the branches are the~ bridges; and non : bridging stations, such :as RF terminals~, are the 35~ ~ leaves~of the tree. Data are~sent along the:branches :

:: :

WO93/07691 P~T/US92/0~610 h~5~v 4 of the newly cre~tsd opti~al spanning tree~ Nodes in the network use a backward learning technique to route packets along the correct branches~
~: One object of the present invPntion is to route:~ 5 data efficiently, dynamically, and without looping.
Another object of the present invention is to make the routing of the data transparent to the RF terminals.
The RF terminals, transmitting data intended for the ; ~ host computer, are unaffected by the means ultimately used by~the RF Network to deliver their data.
It~is a further object of the present invention : for the network to be capable of handling RF ~erminal mobility:and~lost nodes with minimal impact on the entire~RF~data communication system.

,~ :
: ~ :

: ~::
~: :::

W~g3/~7691 ~ 2~ 2~ P~r/US92/08610 ~R~F~DBg~RIPTTO~ OF T~ DRA!IrI~G
The FIG. 1 is a func ional bloc:k diagram of arl RF
data co~munication syst~m incorporating the RF local-area ne~work of the present invention.
.
. .

:~

~ ~ s :

:

WO 93~076g1 PCI`/USg2/08610 '~ i2 0 5 2 ~ - 6 -DBTAIL~D DE~CRIPTION OF T~ I~VgNTION
The FIG. 1 is a functional block diagram of an RF
data communication ~ystem. In one embodiment of ~he present invention, the RF data communication system S has a host computer 10, a network controller 14 and bridges 22 and 24 attached to a data communication link 16. A}so attached to the data communication li~k 16 is a gateway 20 which acts as the root node for the spanning tree of the ~F data network of the present invention. A bridge 42 is attached to the gateway 20 through a hard-wired communication link and bridges 40 : : and 4:4 are logically attached to gateway 20 by two inde`pendent:RF links. Additio~al bridges 46, 48, 50 and 52 are~also connected to the RF ~etwork and are ~:: 15 shown in the FIG. 1. Note that, al~hough shown separate :from the host computer 10, the gateway 20 the ;spsnning~ tree root node) may be part of host : : comput~r~;l0.~ ;
The~IG. 1 further shows RF terminals 100 and 102 2:0~ attached~to:bridge 22 via ~F links and RF terminal 104 attached:~to~ bridge 24 via an RF link. Also, RF
;terminals~106, 108, 110, 112, 114, 116, 118, and 120 can~ e~seen logically attached to th~ RF~ Metwork : : through~their respective RF links. The RF terminals 25~ in FIG.~ l are::representative of non-bridging stations.
In aIte~nate ~ odiments of the present invention, the Network;could contain any:type of device capable of supporting~the~functions needed to c~mmuni~ate in the ; ; RF~ Network such as hard-wired :terminals, re~ote printers,~stationary bar code scanners, or the liXe.
The RF:data:communication~system, as sh~wn in FIG. 1, représénts~the configuration~of the fiystem at a discrete moment in time after~;:the~ initialization of : the cystem.~The RF links,:as shown, are dynamic and subject to change. For example, changes in the :
-::

~;:: : :

- ~w~ 93/076912 ~ 2 ~ ~ ?~ a PCT/US92/08610 s~ruature of th~ RF data communication system can be caused by movement of the RF t~rminals and by interfer~nce that affects the RF ~ommunication links.
In the preferred embodiment, the host computer 10 5is an~IBM 3090, the network controller 14 is a model ; : RC3250 of the Norand Corporation, the data ; ~ ~com~unication link 16 is an Ethernet link, the :~ nodes 20, 22, 24, 40, 42, 44, 46, 48, 50 and 52 are ntel}igent base transceiver uni~s of the type RB4000 10of the Norand Corporation, and the RF terminals 100, 102, 104, 106, 108, 110, 112, 114, 116, 118 and 120 are of type:RT1100 of the Norand Corporation~
The~optimal spanning tree, which provides the data~pathways throughout the communication system, is 15stored and~maintained by the network as a whole. Each node~in the:~network stores and modifies information : which specif~ies how local communication tr~ffic should : flow~ Optimal spanning trees assure efficient, adaptive~;(dynamic)::routing o~ infoxmation without 20 ~ ooping.~
::To~initialize the RF data communication system, ; the:~gateway z0 and the other nodes are organized into an;optimal~spanning tree rooted at the gateway 20. To : form~ the~::optimal spanning tree, in the preferred 25~embodiment~ the gateway 20 is assigned a status of ATTAC~ED and all other bridges are assigned the status :: U~ATTACHED~ he gateway 20 is considered attached to the~:spanning tree because~ it is the root node.
Initially,~all other bridges are unattached and lack : 30a parent in ths ~panning tree. At ~his point, the : : ::attache~gateway nod~ 20 periodically broadcasts a speciic~type of pollinq:packet referred to hereafter as "HELLO packetsl'. m~: NELLO packets can be broadcast~:using known methods of communicating via ; 35radio frequency (RF) link:or ~ia a direct wire link.

::
:~
:

WOg3/076gl PCr/US92/08610 ,~;., In the preferred embodiment of the present invention, the RF li~k is comprised of spr~ad spectrum 2 ~ 2 transmissions using a polling protocol. Although a polling protocol is preferred, a carrier-sense S multiple-access (CSMA), busy-tone, or any other protocol might also manage the communication traffic on the RF link.
HELL0 packets contain 1) the address of the sender, 2) the hopping distance that the sender is from the root, 3) a source addreæs, 4) a count of nodes in the subtree which flow through that bridge, : and 5) a list of system parameters. Each node in the network is~assigned a unique network service address and a node-type identifier to distinguish between ~:: 15 different nodes and different node types. The : : distan~e of~ a node from the root node is measured in hops times the bandwidth of each hop. The gateway root is considered to be zero hops away from itself.
~:; : The unattach~d bridges are in a LISTEN state.
;~ 20 ~Durin~the~LISTEN state, a br~dge will listen to the HELLO~messages that are broadcast. By listening to ~. , , the HELLO;messages~, bridges can learn which nodes are attached~to the sp~nning treeO The unattached bridges analyze~ the contents of the HELL0 messages to 25~ ~determine~ whether to requ~t attachment to the broadcasting node. In the preferred embodiment, a bridgè atte~pts tG attach to the node that is logically~closest to the root node. In khe preferred emboaim~nt~, the logical distance i~ based upon the ; . 30 numbex o~ hops needed to reach the root node and the bandwidth:of those hops. The distance the attached node is~away from the root node is found in the second : field of the HELL0 message that is broadcast.
, In~another embodiment of the present invention, the bridges consider the nu~ber of nodes attached to ~ .5~ h ~ .~3 2`û
~r `~O 93/07691 PCT/US92/08610 _ g _ the attached node as well as the logical distanc~ of the at~ached node from the root node. If an attached node is ~verloaded with other attached nodes, the unattached bridge may request attachment to a less : 5 loaded node.
After attaching to an attached node, th- newly : attached bridge~ (the child) must determine it5 distance from the root node. ~o arrive at the distance of the child from the root node, the child adds the broadcast distance o~ its paxent from the root node to the distance of the child from its parent. :In the:preferred embodiment, the distance of a child from its:parent is based on the ~andwidth o~
the data~ communication link. For example, if the child:attache~s to its parent via a hard-wired link (data :rate~26,000 baud), then the distance of that communication li~ might e~ual, for example, one hop.
However,~if the child attaches to its parent via an RF
link ~data rate:960U baud~:, then the distance of that 20~ :co~munication li~k might correspondingly be equal 3 : hops~. The n a r of the hop corresponds directly to the co., unication speed of the link. This may not only~take~:into consideration baud rate, but also æuch factors;as~channel interference.
25~ Initia~lly, only the root gateway node 20 is broad~asting: HELLO: messages and only nodes 40, 42 and 44~:~are ~within range of the ~ELL0 messages broadcast :`by :the gateway. ~herefore, after the liste~ing~period has expired, nodes 40, 42 and 44 3~ request attachment to the gateway node 20. The unattached~;nodes 40, 42, and 44 send A ~ CH.request packets~and the:at ached gateway node 20 acknowledges the ATTACH~.request packets with l~cal ATTACHOconfirm packets.~ The newly attached~bridges are assiyned the ~:: 35 ~ sta~us ATTACHED and begin broadcasting th~ir own HELL0 , ~

:
: ~:

WO93/076gl PCT/US92~08610 ~
. ~ .

2 ~2 C ~ackets, looking for o~her unattached bridges. Again, ~ ~ ~the remaining unattached nodes attempt to attach to the attached nodes that are logically closest to th~
root node. For example, node 48 i~ within range of HELLO messages from both nodes 40 and 42. However, node 40 is three hops, via an RF link, away fr~m ~he : gateway root node 20 and node 42 is only one hop, via a hard-wired link, away from the gateway root node 20.
Therefore, node 48 attaches to node 42, the closest node:to the gateway root node 20.
:The ~ending o~ HELLO messages, ATTACH.request packets and AT~ACH.confirm packets aontinues until the entire sp~anning:tree is established. In addition, attached bridges may also respond to HELLO messages.
If a HELLO message indicates that a much closer route ; ~ to the root node is available, the attached bridge ; : sends :a DETACH packet to its old parent and an ATTACH~.request packet to the closer node. To avoid instabi-lity:in the system and to avoid overloading any given node, an~attached bridge would only respond to a HELLO message if the hop count in a HELLO packet is greater~ than a certain threshold value, ;CHANGE_~RES~OLD~. In the preferred e~bodiment, the value ;of the CHANGE_THRESHOLD equals 3. In this :25 ;~anner,~an~optimal spanning tree is formed that is capable of~ transmitting data without looping.
Nodes,~ other than the gateway root node, after acknowledging ~an ATTA~H.re~ est packet from a :;previously ~ unattarhad node, will send the ~TTACH.request packet up the branche of the spanning tree to the gateway root node. As the A~TACH.request packet is~being sent to the gateway root node, other nodes:~attached on the ~s~ame branch record the destinat:ion of the newly attached node in their routing:entry table.~ When the ATTACH.request packet ~ :
:

~W~93~07691 ~ 2 ~ PC~/US92/~8610 reaches the gateway root node, the gateway root node returns an end-to-end A~TACH. conf irm packet .
After the spanning tree is initialized, th~ RF
terminals listen for periodically broadcasted Hello packets to dstermine which attached nodes are in range. After recei~ring HELL0 messages from attached nodes t an RF terminal responding to an appropriate poll sends an ATTACH. request packet to a~ttach to the node logical~y closest to the rootO For example, RF
terminal 110 is physically closer to node 44.
~Iowever, node 44 is three hops, ~ria an RF link, away ~Erom the gateway root node 20 and node 42 is only one hop, ~ria a hard-wired link, away from the gateway root node 20. Therefore, RF terminal 110, after hearing HELL0 messages from both nod~s 42 and 44, attaches to node 42, the cIosest node tQ the gateway root node 20.
~: Similarly, ~ terminal 114 hears ~IELL0 messages fromnod~s ~8 and 50. Nodes 48 and 50 are both four hops away from~ ~e gateway roc~t node 20. ~lowever, nod~ 48 has ~wo RF terminals 110 and 112 already attached to it while node 50 has only one RF terminal 116 attached to it~ :m erefore, ~F terminal 114 will ~ttach to node 50, the least busy node of equal distance to the :: gateway root node 20.
: 25 The attached node acknowledges the ATTACH.request : and sends the ~TTACH.re~ue~t pack t to the gateway root node. Then, the gateway root node returns an :
: end-~o-end AT~ACH.confirm packet4 In ~his manner, the end-to-end ~TTACH.raquest Xunctions as a discovery packet enabling the gateway root node, and all other nodes ~long the same branch, to learn the addre~s of . ~ tha RF terminal quickly. This proc~ss is called backward learning. Nodes learn the addresses of terminals by monitoring the traf~ic from terminals to the root. If a packet ~rrives from a terminal that is WO93/07691 PCT/US92/08610 ~.

~Z~20 not con~ained in the routing table of the node, an entry is made in the routing table. The entry includes the terminal address and the address of the node that sent the packet. In addition, an entry timer is set for that terminal. The entry timer is used to determine when RF terminals are actively using the attached node. Nodes maintain entries only for terminals that are actively using the node for communication. If the entry timer expires due to lack of communication, the RF terminal entry is purged from the routing table.
;~ ~; The RF links among the RF terminals, the bridges, and the gateway are oft n lost. Therefore, a connection-oriented data-link service is used to maintain the logical node-to-node links. In the absence of network traffic, periodic messages are sent and received:~to ensure th~ stability of the RF link.
As a result, the loss of a link is quickly detected and the RF Network can attempt to establish a new RF
link before~data transmission ~rom the host computer to an RF termi~aI is advers~ly affected.
; Communication between terminals and the host : compu~er; is~accomplished by using the resulting RF
Network.~To;com~unicate with the host computer, an RF
~ terminal sends a data packet in response to a poll from :the ~bridge closest to the host computer.
Typically,~:the RF terminal is attached~.to the bridge : ¢loses: to;~the host computer. However, RF termina}s : :zre ~ constantly listening for HELL0 ~nd polling i: ~ 30 messages from other bridges and may attach to, and then co~municate with, a bridgP in the table of bridges t~at i~s~closer to the particular RF terminal.
, ~
nder certain operating conditions, duplicate data packets can bs transmitted in the RF Network.

-~WO93/07591 ~ PCT/US92/08610 For example, it is possi~le for an RF terminal to transmit a data packet to its attached node, for the node to transmit the acknowledgement frame, and ~or the RF terminal not to receive the acknowledgement.
Under such circumstances, the RF terminal will retransmit the data. If the duplicate data packet is updated into ~he database of the host computer, the database would become corrupt. Therefore, the RF
Network of the present invention detects duplicate data packets. ~o ensure data integrity, each set of data transmissions receives a sequence number. The sequence numbers are continuously increme~ted, and duplicate sequence numbers are not accepted.
When a~ bridge receives a data packet ~rom a terminal directed to the host computer, the bridge ~ forwards the data packet to the parent node on the :: branch~ The parent node then forwards the data packet :~ to it ! parent node. The forwarding of the data packet continues until ~he gateway root node rereives the data packet and sends it to the hGSt co~puter.
; Similarlyl when a packet arrives at a node ~rom the host compu~er directed to an RF terminal, the node checks lts routing entry table and forwards the data ~: packet to its child nodP which is along the branch destined:~for the RF terminal. It i~ not necessary for th~ nodes along the branch co~taining the RF terminal to ~now the ultimate location of the RF terminal. The ~orwarding of the data packet continues until the data pa~ket reaches the final node on the branc~, which then forwards the data packet directly to the terminal itsel~.
Communication is a~so po ~ible between RF
; termi~als. To communicate with ano~her RF terminal, the RF ~erminai sends a data packet to its attached bridge. When the bridge recei~es the data pack~t from W093/~769t PCT/USg~/0~610 2 a terminal direated to the host computer, the bridge checks to see if the destination address o~ the RF
terminal is located within its ruuting table. If it is, the bridge simply sends the message to ~he S intended RF terminal. If not, the bridge forwards the data packet to its parent node. The forwarding of the data packet up the branch continue~ until a common parent between the RF terminals is found. Then, the common parent (often the gateway node it~elf) sends the data packet to the intended RF terminal via the branches of the RF Network.
During the normal operation of the RF N~twork, RF
terminals can become los~ or unattached to their attached node.;;If an RF terminal becomes unattached, for whatever reason, its routing entry i5 purged and the RF terminal listens for HELL0 or polling messages from any attached nodes in range. After recei~ing HELLO~or~pol~ling message~ from attached nodes, the RF
terminal~sends an ATTACN.request packet to the attached node`closest to the root. That attached node acknowledges~ the~ A~TACH~re ~ est and sends the ; A~ ACH~.request ~packet onto the gateway root node.
Then, ~the gatewa~ root node returns an end-to-end ATTA~H.confirm packet~
25~ Bridges~an also become Iost or unattached during normal~ operations of the RF Ne~work. I~ a bridge becomes~ lost or unattached, all routing entries contalning the bridge are purged. The ~ridge~ then broadcasts~;a HELLO.request with a global bridge ;30 destina~ion address. Attached nodes will broadcast HELLO packets~ immediately if they receive an ~TTA~H.xequest packet with a global d~stination address.~ miS helps the lost node re-attach. Then, the bridge~ enters the LISTEN state to learn which attached nodes are within range. The unattached : ~:: : :: ::
: -: ~ : : :

::
:~
, 093/07691 ~ PCT/US92/08610 bridge analyzes the contents of broadcast HELI~
messages to determine whether to request attachment to the broadcasting node~ Again, khe bridge attempts to attach to the node that is logically closest to the root node. After attaching to the closest node, the bridge begins broadcasting HELLO messages-to solicit ~TTACH.requests from other nodes or RF terminals.
The spread-spectrum system provides a hierarchical radio frequency network of on-line 10 ~ terminals for data entry and message transfer in a mobile environment. The network is characterized by sporadic data traffic over multiple-hop data paths consi~s~ing ;of RS485 or e~hernet wired links and single-channel direct seguenced spread spectrum links.
The network architecture is complicated by moving, hidden, and sleeping nodes. The spread spectrum system consists of the following types of devices:
Terminal controller -- A gateway which passes messaqes from a host port to the RF n~.twork; and which 20~ ~passes messages from the network to the host port.
The host port~(directly or indirectly) provides a link between~the~controller and a "host" computer to which the~ erminàls are logically~attached.
Base~station -- An;intermediate relay node which 25~;~ is ùsed to;èxtend the range of the controller node.
8ase~ station-to-controller or ~ase station-to-base station l~inks can b~e wired or wireless RF.
Terminal -- Norand RF~ hand-held terminals, print~rs~ etc. In addition, a controller device has a terminal compohent.
The~de~ices are logically or~anized as nodes in an~optimal) spanning tree,~with the controller at the root,~internal nodes in base stations or controllers ` on branches of the tree, and terminal nodes as (possibly mobile) leaves~on the;~ree. Like a sink :

WO ~3/07691 PCl`/US92/08610 : : 2 ~ 2 ~ 5 ~ tree, nodes closer to the root of the spanning tree are said to be "downstream" from nodes which are further away.~ Conversely, all nodes are "upstream"
from the root. Packets are only sent along branches of the~spanning tree. Nodes in the network use a "BACKWARD LEARNING" technique to route packets along , the b~anches of the spanning tree.
Devices~ in the spanning tree are logically categorized~as~one of the following three node types:
l)~ Root~(or root bridge) - A controller device ;"',~ which ~functions as the root bridge of the network~spanning tree. In the preferred embodimènt, the spanning tree has a single root node.~ Initially~,~ all controllers are root candidates from which a root node is selected.
This~selection may be ~based on the hopping 'distancé~ to~the host, preset priority, r~ndom select~ion,~;etc.~ ~
2)'~,Bridge~ An~internal node in the æpanning , 2~0~ ,trée~which~ is used to "bridge" terminal nodes togethe~r into~an interconnected network. The s~also consider~ed a bridg and t e term~'~",bridge"~ may be used to refer to all non-terminal;~nodes or ali non-te ~ inal nodes except 25~ the~root, depending~on~the context herein. A
E~ node~;consists~ of~a network interface functiàn~and a routing,function.

A',~te inal~ node can~ibe~view,ed as the ~software 30~ ent ~ ~that~terminates a ~ranch in the spanning tree~
i~ A~controller devicé~contains~a terminal~ node(s) and~a-bridge~node. ~The~bri~ e~;node is the root node if;the~controller~is~functioning as the root ~bridge.
3~5, A~base~stàtion contains;~a~ bridge node. A terminal WO93~07691 2 ~ PCTlUS92/08610 devic~ contains a terminal node and must have a n2twork interface function. A "bridging entity"
:~ refers to a bridge node or to the network interface function in a terminal.
The basic requirements of the system are the following. :
a) Wired or wireless node connections.
b) Network layer transparency.
c) Dynamic/automatic network routing , ~ :
con~iguratio~.
d) Terminal mobility. Terminals should be able to move:about the RF network without losing an end-to-: end connection.
e)~:: Ability to accommodate sleeping terminals.
f) Ability to locate terminals quickly.
g)~ Built-in redundancy. Lost nodes should have ~inimal: impact on the network.
h) :P~ysical link independenae. The bridging algorithm;is consistent across heterogenèous physical 20~ ~ links.~
:The~software for the spread-spectrum system is f~unctionally~layered~as follows.

Medium Acc-ss_Controll MAC) 25~ The;~;MAC~ layer :is :responsible for providing reliablê~transmission between any two nodes in the network ~ii.e:. terminal-~o-bridge). The MAC has~ a ; :channel~acoess control component and a link control component.~ & e link control component facilitates and re ~;lates~point-to-point fr~me transfers in the absence~of~ collision: dete~ction.~ The M~C channel ; : access~ control component;~;regulates access to the :: network. :Note that herein~, the MAC layer is also ; referred to~as~ the Data Link~layer.

::
, ~ :

:~;~: :

WO93/07691 PCT~US92~08610 ~.~
~5 ~3rid~ina La~rer : The bridging layer, which is also referred to herein as the network layer, has several functions as follows.
~: S1~ The bridging Iayer uses a "HELLO protocol"
to organize nodes in the network into an optimal spanning tree rooted at the root bridge. The spanning tree is used to' prevent loops in the topology.
Int~rior branches of the spanning tree are relatively : lO ~stable (i.e.:controller and relay stations do not move often). Terminals, which are leaves on the spanning : three:,~ may become unattached, and must be reattached, : frequently.~
2.:~ The bridging layer routes packets from :15terminals to the host, from the host to terminals, and from terminals to terminals along branches of the spanning~tree.
: 3~: ~ The:bridging layer pro~ides a service for storing packets for SLEEPING t,erminals. Packets which : 2~0~ ~ cannot be~delivered immediately can be savea by the bridging;~e~tity in a parent node for one or more HELL0 ~"x~ The~bridging: layer propagates lost node information~throughout the spanning tree.
25~5~ The~ bridging layer maintains the spanning tree~links~
6~ ~:The~ bridging layer distri~utes network :i~terface::addresses. ~ ~

: ' 30~; : Loaioal_Link Cont~ol_Layer A logical link control layer, also known herein as the~ Transport layer: herein, is responsible for providing~reliable transmission between any two nodes in the~network (i.e., terminal-to-base station)~ The 35data-link :: layer pro~ides a connection-oriented ~, -~093/07691 ~ ~P~ PCT/US92/08610 reliable service and a connectionless unreliable service. The reliable service detects and di~cards duplicate packets and retransmits lost pacXets. Th~
unreliable services provides a datagram facility for upper layer protoGols which provide a reliable end to-end data path. The data-link layer provides ISO layer 2 services for terminal-to-host application ~essions which run on top of an end-to-end terminal-to-host transport protocol. However, the data-~ink layer : lO provides transport ~ISO layer 4) services for sessions- ~ contained wi hin the SST netwoxk.

Hi~her Lay~rs : For terminal~to-terminal sessi~ns contained within the SST network, th~ data-link layer provides transport layer services and no additional network or transport layer is required. In this case, the M~C, bridging, and data-link layers discussed abov~ ean be : viewed as a data link layer, a nGtwork layer, and a transport layer~ respectiYely. For terminal-to-host-application sessions, higher ISO lay~rs e~ist on ~op of:the~SST data-link layer and must be implemented in the terminal and host computer, as required. This : documen~ does not define ~or restrict) those layers.
:: 25 ~ ~is ~o~ument does discuss a ~ast-connect V~TP-like ; transport ::protocol which is used for transient internal~terminal-to-terminal sesæions.
Specifically, a n~two~k layex has several functions~, as follows.
3~ 13 The network layer uses a "hello protocol" to organize:nodes in the ne~work into an optimal spannin~
tree rooted:at the controller. (A spanning tree is ~ required to prevent loops in the topology~) Interior : : branches of the spanning tree are relati~ely stable ;~ 35 (iOe., the controller and base 8tation5 do not move , :
: :

WO93/07691 PCT/US92/08610 ~

often). Terminals, which are leaves on the spanning tree, become unattached, and must be reattached frequently.
2) The ne~work layer routes messages from terminals to t~e host, from the host to terminals, and from terminals to terminals along branches of the spanning tree.
3~ The network layer provides a service for storing messagés for SLEEPING terminals. Messages s 10 which cannot be delivered immediately can be saved by the network entity in a parent node for on~ or more hello times.
41 The network layer propagates lost node information throughout the spanning tree.
~- 15 5) The network layer main ains th~ spanningtree links in the absence of regular data traffic.
: A transport layer is responsible for establishing ;: and maintaining: a reliable end-t~-end data path between~transport access points in any two nodes in 20~ the network. ~The transport layer provides unreliable, reliable ~and~ a~transaction-oriented services~ The transport~ layer should ~e immune to implementation :changes~in~the~network layer.
The~ responsibilities of the transport layer 25~ include:~the~:following.
Establishing and maintaining TCP-like connections~ for reliab}e root-to-terminal data transmission. ;
: 2~ Maintaining VMTP-like transaction records :: :
~ :30 for reliable transient message passing between any two `
nodes.:
3):~ Detecting and discarding duplicate packets.

4)~ Retransmitting lost packets.
Layers:l through 4 are self-contained within the Norand~ RF;-:network, and are independent of the host W093tO7691 ~ 2 0 PCT/US92/08610 computer and of texminal applications. The session layer ( and any higher layers) are dependent on speci~ic applications. Therefore, the session protocol (and higher protocols~ must be implemented as S required. Note that a single transport access point i~ sufficient to handle single sessions with multiple ~: nodes. Multiple aoncurrent sessions between any two nodes could be handled with a session identifier in a ~::: session header.
:: 10 Network address requirements are as follows. DLC
framed contain a hop destination and source address in :
the DLC header. network packets contaîn an end-to-end destination and~ a s~urce address in the network header. Transport messages do not contain an address ~: 15 field; instead, a transport connection is defined bynetwork layer~ source and destination address pairs.
~ultiple txansport conne~tions require multiple network address:pairs.
: me transport header contains a TRANSPORT ~CCESS
:~ POINT identifier. DLC and network addresses are : consistent~and have the same for~at. Each node has a : unique LONG~ ~ RESS which is programmed into the node at~ the~ factoxy. ~The long address is used only to vbtain~:~a~SHOR~ ADDRESS from~the root node.
25: ~ ~The~:network entity in each node obtains a S~O~T
ADDRESS~from the root node, whiah identi~ies the no~e uniquely.~The~network e~tity passes the short address to the~DLC~entity. Short addresses are used to minimize~packet~:sizes.
~ Short addresses consist of the following. There :: : is: an:address length bit (short or long).
a~spanning tree identified.
a node-type identifier. Node ~ypes are well know~
~: ~
~ 35 a unique multi cast or~broadcast node identifier.
:.

:~ ~

~093/076g1 PCT/US92/08610 The node-ident~fier parts of root addresses are a~ 3 well known and are constant. A default spanning tree identifier is well known ~y all nodes. A non-default spanning tree identifier can be entered into the root nodQ (i.e., by a network adminis~rator) and advertised to all other nodes in "hello" packets~ The list of non-default spanning trees to which other nodes can attach must be entered into each node.
A node-type identifier of all 1's is used to ~; 10 specify aIl node types. A node identifier of all l's is used to specify all nodes of the specified type.
A DLC~identifier of all 0's is used to specify a DLC
en~ity which~does not yet have an address. The all-O's address is used in~ DLC frames that are used to send and receive network ADDRESS packets. (The network~entity~ in each node filters ADDRESS packets based on the network address.) ` Short-address allocation is accomplished as follows.~ Short node identifiers o~ root nodes are 20~ ~ well kno ~ . ~All other nodes~ must obtain a short node identifier from the root. To obtain a short address, a~node~send an ADDRESS request packet to the root node. The~source~addresse~ (i.e., DLC and network) in the request packet are LONG ADDRESSES. The root 25~ ~ maintains~an~address queue of used and unused SHORT
; aDDRESSES. ~If possible, the root selects an available short~address, associates the short address with the long~address~of the requesting node, and returns the shor~ ~address~ to the~ requestin~ node in an ADDRESS
~, 30 acknowledge packet. SNote that the destination address in the acknowledge packet~is a long address.) A~ node must obtain~;a;; (new) short address initially~and whenever an ADDR~SS-TIMEOUT inactivity period expires without having the nQde receive- a ; 35 pacXet from the network entity in the root.

:

:

~WO93/076912 ~ 2 ~ PCT/USg2/08610 .

The network entity in the root maintains : addresses in the addre~s queue in least recently used order. Whenever a packet is received, the source address is moved to the end of the queue. The address Sat the head~ of the queue is available for use by a requesting node if it has never been used or if it has ~ been inactive for a MAX-ADDRESS-LIFE time period.
:: MAX-ADDRESS-LIFE must be larger than ADDRESS-TIMEOUT to ensure that an address is not in use by any lOnode when it~becomes available for another node. If the root:receives an ADDRESS request from a source for which an;entry exists in the address queue, the root simply~updates:~the:queue and returns the old address.
: The network layer organizes nodes into an optimal lSspanning tree with::the controllsr at the root of the .tree. ~(Note that the spanning three identifier allows two logiaal~trees to exist in theisam~ coverage area.~
;Spanning tree organization is facilitated with a HELL0 protocol~which all~ws nodes to:deitermine the shortest 20~path~to~the~ root be~ore attaching to the spanni~g tree.~: All messages~are routed along branches of the spanning ~ tree . ~
Nodes~in~;the network are generally categorized as ATTAC!~ED or~;:llN~lqACHED. Initially, only the root node 25~is attached~.~ A single controller may be designated as the ;root,~ or ~multiple:: root: candi~ates~ (i.e.
controllers)~ may~negotiate~to determine which node is : ;the~:root.: ~:~Attached bridge nodes and root candidates transmit ~nHELLO~ packets at calculated intervals. The 30HELLO:p~ckets include:
a)~ the~ source; address, which includes the spanning~:tree;ID)~
:b)~a broadcast destination address.
c)~: ~a~ Nseed" ~alue from which the time sche~ule : 35of future hello messages can be calculated.

: :

WO93/07691 PCT/US92/08610 ~
~ 5~ - 24 -d) a hello slot displacement time specifylng an actual variation that will occur in the scheduled arrival o~ the very next hello message (the scheduled arrival being calculated from ~e ~seed").
e~ the distancs (i.e., path cost) of the transmitter from the host. ~he incremental portion of the distance be~ween a node and its parent is primarily a function of the type of physical link (i.e., ethernet, RS485, RF, or the like). If a 0 signal-strength indicator is available, connections are biased toward the link with the best signal strength. The distance component is intended to bias path selection toward (i.P., wired) high-~peed connections. Setting a minimum signal str~ngth threshold helps prevent sporadic changes in the : networX. In addition, connections can be biased to bal~nce the load (iOe., the number of children) on a : parent node~
f~ a p~nding message list. Pending ~essage li~ts consist o~ 0 or more destination address/message-length pairs. Pendin~ messages for terminals are stored in the terminal'~ parent node.
: ~ g) a detached-node list. Detached-node list~
contain the addresses of nodes which have detached 25: f~om the:spanning tree. The root maintains two lists.
A private list consists of all detached node : addresses, and an advertised list con~ists of the ~:; addresses of all detached node~ which have p~nding : tra~sport messages. The ~ddresses in the hello packet ~, 30 are equivalent to the ~d~erti~ed list~
An internal node learns which entrie~ ~hould be : in its li~t from hello messages transmitted by its parent node. The root node bui}ds its detachednode lists from information received in DET~C~ packets.

WO93/07691~ PCT/US92/08610 Entries are included in hello message for DETACH-MSG-LIFE hello times.
Attached notes broadcast "SHORT ~.ELLO" messages immediately if they receive an "HELLO.request" packet 5with a global destination address; otherwise, attached nodes will only broadcast hello messages at calculated time intervals in "hello slots." Short hello messagPs do not contain a pending-messag~ or detached-node : list. Short hello messages are sent independently of :~ 10regular hello messages and do not affect regular hello : timing.
:Unattached nodes (nodes without a parent in the spanning:;tree) are, initially, in an "UNATTACHED
LISTEN" state.~During the listen s ate, a node learns ~: 15which attached base station/controller is closest to :the root;~node:by listening to hello messages. After the listening period expires an unattached node sends : an ATTACH.request packet to the attached node closest to :the~ root.~ The attached node immediately :~: acknowledges~ the ATTACH.req~est, and send the ATTACH.~request packet onto the root (controller) node.
The root~node~returns the re ~ est as an end-to-end : A~TACH.conf;irm~:packet. If the newly-attached node is a;~base~:station,~the node calculates its link distance 25and adds:the:distance to the distance of its parent : beforè:~beginning;to transmit hello messages~
; The`end-to-end ~TTA~H.request ~unctions as a disco~very pa~ket, and enables the root node to learn `the:address of the sou~ce node quickly. The end-to-30~end ATTACH. request, when sent from a node to the root, does not always tra~el the entire distance.
:: When a downstream node receives an ATT~CH.request pàcket ~nd;~already has a correct:routing entry for the :associated:~node, the downstrea~ node intercepts the 35request and~returns the ATTACH.confirm to the source : ~ :

~::: ~ :

WO 93/07691 P~/US92/08610 ,, ~, node. (Note that any data piggy-backed on the 'Z~ A~rTACH. request packet must still be forwarded ~o the host.) This situation occurs whenever a "new" path has more than one node in common with the "old" path.
The LISTEN state ends after MIN HELLO hello time slots if hello messages have been received from at le~st one node. If no hello messages have been received the listening node waits and retries later.
An attached node may respond to a hello message ~ ~ ~ 10 from a node other than its parent (i.e., with an : ~ ATTACH. request) if the difference in the hop count specified in the hello packet exceeds a CHANGE-THRES~OLD level.
:: ` : : `: : : : :
Unattached ~nodes may broadcast a GLOBAL
ATTACH.request ~with a multi-cast base station destination address to solicit short hello messages from attached base~stations. The net effect is that the LISTEN~state may (optionally) be shortened. (Note that only~attached base station or the controller may 2~0 respond~to~ATTACH. requests.) Normally, this facility is~ reserved~for base stations with children and terminals~with ~ransactions in progress.
ATTACH.~requests contain a (possibly empty) CHILD
LISTt;to~ènable~internal~nodes to update th~ir routing 25 ~ tables. ATTACH.requests~also contain a "count" field which indica~tes that a terminal may be SLEEPING. The network~entity in the parent of a SLEEPING terminal con temporarily store messages for later d~livery. If the count field is non-zero, the network entity in a parent node will store pending messages unt~l 1) the message~is ~elivered, or 2) "count" hello times have expired.
Transport layer data can ~e piggy-backed on an attached reguest packet~from a terminal. (i.e., an ~: :
~ :
~ ~ :
:
: : :

~WO g3/07691 2 ~ 2 Q PCT/VS~2/08610 attach request/confirm can be implemented with a bit flag in the network header of a data packet~) Network layer routing~
All messages are routed along branches of the spanning tree. Base stations "learn" the address of terminals by~monitoring tra~fic from terminals (i.e., to the root). When a base station receives (i.e., an ATTACH.reguest) packet, destined for the root, the base station creates or updates an entry in its routing table for the terminal. The entry includes ; the terminal~address, and the address of the base station which sent the packet (i.e., the hop address).
When a base~station receives an upstream packet (i.e., from the root, destined for a terminal) the packet is simply~forwarded to the base station which is in the routing entry for the destination. Upstream messages (i.e.,;~to a~te ~ inal~) are discarded whenever a routing entry do~s~not exist. Downstream mes ages (i.e., from 20 ~ a~terminal to the root) are simply forwarded to the next~downstrea~ node (i.e., the parent in the branch ,of the sp~nning tree.
TEKMINAL--TO-TERMINAL~ ~OMNUNICATIONS is accomplished,~by routing all terminal-to-terminal 25~ tra~ffic~through the nearest~com~on~ancestor. In the ; worst~case~ the~root is the nearest common ancestor.
A ~ADDRESS~SERVER" facilitates terminal-to-terminal com~unicatio,ns (see below~
DELETING INVALID ROUTING T~BLE EN~RIES is accomplished in several ways: connection oriented - tranæport layer ensures that packets will arrive from nodes~ atta~hed to the branch of th spanniny tree within~ the~timeout period, unless a node is disconnected.)~

:

::: : ~ : : :
:

W093f07691 PCT/US92/08610 ~,~
..

2 ¦ 2 3 ~ 2 ~ ~, Wh~never th~ DLC entity in ~ pare~t fails RETRY MAX times to send a message to a child node, the node is logically disconnected from the spanning tre~, with one exception. If the child is a SLEEPING
terminal, the message is retained by the network entity in the parent for "count" hello times. The parent immediately attempts to deliveir the message after it sends its next hello packet. If, after : "count" hello~times, the message cannot be delivered, then the child is logically detached from the spanning : tree. : Detached :node information is propagated downstream to the root nc~e, each nod~ in the path of the DETACH~::packet must adjust its routing tables appropriately according to the following rules: a) if the lost node~ is a child terminal node, the routing : entry for~the terminal is deleted and a DETACH pa~ket is~generàted, b) if the node sp~cified in DETACH
: packet is~a~terminal and the node which delivered the packet~is the:~next hop in the path to the t~rminal, ; 2~0~ then:the ~routing table entry for the terminal is deleted:~and~the DE~ACH packet is forwarded, c) if the lost;:~ node~is~ child base station node then all routing~entries~ which~specify that base s ation as the next hop~are~deleted and a DETAC~ packet is generated 25~ for~:each~lost terminal.
IN~;~;GENERAL,~:WHEMEVER A~NODE DISCO~ERS THAT A
TER~INAL I:S~DETACH D, IT PURGES;ITS ROUTING ~ TRY FOR
THE;~TERMINAL.~;:WHENEVER A NODE DISCOVERS THAT A BASE
STATIQN:IS~:DETACH D, IT PURGES ALL:RO~TING ENTRIES
;: 30 CONTAINING ~THE IBASE STATION~ ONLY ENTRIES FOR
: UPSTREAM NODES ARE DELETED.
When~DETACH~:packets:reach~the root node, they are :: : added to a:~"detached list." Nodes remain in the root node's detached list until a)~ the node reattaches to ; 35 the spanning~tree, or b) the list entry times out.

; ~

~093/076912 ~ 2 ~ S ~ ~ PCT/US92/08610 The detached list is included in hello messages and is propagated throuqhout the spanning tree.
For example, if a te~minal detaches and reattaches to a different branch in the ~panning tre~, all downstream nodes in the new branch (quickly) n learn" the new path to the terminal. Nodes which were a~so in the old path change their routing tables and no longer foxward packets along the old pathO At least one node, the root, must be in both the old and lOn~w path. A new path is established as soon as an en~-to-en~ attach request packet from the terminal reaches a node which was also in the old path.
~ 4) A node (quickly) learn~ that it is detached : whenever~it receives a hello message, from any node, lSwith it~ address in the associated detached list. The detached node can, optionally, send a global TTACH.request, and then ~nters the UNATTACHED LISTEN
state and reattaches as d~scribed above. After :~ :
reattaching,~the node must remain in a HO~D-DOWN state ~: : 20until:its:address is aged out of all detaGhed lists.
During the~HOLD-DOWN state the node ig~ores detached lists.

5) : A node becomes disconnected and enters the UN~TT~CHED LIS~EN state whenever HELLO-RETRY-MAX hello ~ messages ;a~e~miss~ed from it~ parent node.

6) ~A ~node enters the ATTACHED LISTEN state : whene~er a ingle hello message, from its parent, i5 :," ~:
~ missed. SL~EPING terminals remain awake during the : : :
TACHED LISTEN state. The state ends when the ; 30terminal receives a data or hello m~ssage from its .
parent. The terminal becomes UNATT~CHE~ when a~ its address appears in the det~hed list of a hello ~ message from an ode o~her ~han its parent, or b) : ~HELLQ-RETRY-NAX h~llo messages are missed. The total ~093t07691 PCT/US92/08610 ~

-- ~0 --number of hello slots spend in the ~ISTEN state is co~stant.
If a node in the ATTACHED LISTEN ~tate discovers a path to the root which is CHANGE-T~RESHOLD shorter, it can a~tach to the shorter path~ Periodically, SLEEPING terminals must enter the ATTACHED L ~ state :~ to discovery any changes (i.eO, shorter paths) in the network topology.

~ Hello synchronization.
A11 attached non-terminal nodes broadcast periodic "hello" messages in discrete "hello slots" at ; calculated;intervals. Base station nodes learn which hell:o 8~10ts :are busy and refrain from transmitting lS during busy hello slots.
; A terminal refrains from transmitting during the .
hello slot :of its parent node and refrains from transmitting during message slots xeserved in a hello message.~
20~ The~;hello~message contains a "seed" field used in :a well-known~:randomization algorithm to determine the next he1:1O~:s1Ot for the transmitting node and the next geed~. The address o f the transmitting node is used as a~:factor~ in:~the algorithm to ~arantee randomization.
25 ~ Nodès can~execute~the a1gorithm i times to determine : the~ time~;~ (and~ seed) if the~ i-the hello message from the~transmitter. ~ :
After attached, a base station chooses a random initia1~seed~and a non-busy hello s~lot and broadcasts 3 0 a hello message in that slot . The base station : chooses~ ~succeeding hello lots by executing the randomization~:algorithm. If an execution of the algoritlun~ ~chooses a busy slot, the next free slot is used and~ a ~hello "displacement" f i~ld indicates the 3S offset fro~a calculated slot. Cumulative delays are :: . : ~ . :

~ W093/07~91 PCT/US92/08~10 ` ' 2~2~ ~2~

not allowed (i.P., contention delays during ~he i hello transmission do not effect th~ tim~ of ~he i+1 hello transmission).
HELLO-TI~E and HELLO-SLO~-TIME values are set by the root node and flooded throughout the network in hello messages. The HELL0-SLOT-TïME value must be large enough to minimize hello contention.
A node initially synchronizes on a hello message ~rom its parent. A SLEEPING node can power-down with an active timer interrupt to ~aks it just before the n xt expected hello message. The network entity in ba~e station nodes can tore mes~ages for SLEEPIN~
nod~s and transmit them immediately following the he lo messages. This implementation enables SLEEP~NG
terminals to receive unsolicited messages. (Note that : the network layer always tries to deliver mes~ages : immediately, before storing them.) Retries ~or pending messages are tran mitted in a round-robin order when messages are pending ~or more than one destination~
: Note that a child node that misses i h~llo messages, can calculate the time of the i+~ hello mes~age.
~ ~ .
;25~ : Transport~:~1ayer theory and implementation notss.
: The~ ; transport layer provides reliable, unrel~iable,~and transactionoriented services. Two types~:o~ transport:connection~ are de~ined: 1~ a TCP-e~transport connection ~ay be explîcitly requested ~,~ : 30 for long-lived connectionæ or 2) a ~MTP-like connee ion-reeord may ~e implieitly ~et up for transient;eonnections. In addition, a eonnectionless serviee is provided for nodes whieh s~ppDrt an endoto-end transpvre eonnection with the host eomputer.

WO93/076glPCT/US9~/08610 ~
~, r ~ aThe interfaces to the next upper (i.e., - 2;~ 3~application) layer include:
CONNECT (access Point, node name~
LISTEN (access_point) SUNITDATA (access point, node name, bu~fer, length) SEND (handle, buffer, length) RECEIVE (handle, buffer, length) : CLOSE (handle) 10The~"handle" designates the connection type, and is the connection identifier:for TCP-like connections.
SEND messages require a response from the network node:~(root:~or; terminal) to which the message is :: directed.~
lSUNITDATA messages do not re~uire a response.
UNIT~ATA is:;used to~send messages to a host which is capab1e~:of supporting end-to-end host-to-terminal transport connections.
Because~the~network layer provides an unreliable 2~0ser~iGe~ the~;transport layer is required to detect dupli ate~packets and retransmit lost packets.
Detecting~ ~ licates is fa ilitated by numbering transport packets with una~biguous sequence numbers.

5~ Transport connections.
TCP-like:~transport connections are used for message~transmission over long-lived connections~ The connections~may:be terminal-to-root or terminal-to-terminàl~(i.;e.,~:base 5tation5 are not involved in the transport connection).
TCP-1;ike transport Gonnections are established ;:using~a~3-way handshake. Each end selects its initial sequence~ nunber and ~acknowledges the other end's : initial :sequence number during the handshake. The 35 :node whiah~;initiates the connection must wait a MAX-: ~ :
:; : :
:, `

~~93/07691 ~ PCT/US92/08610 .

PACKET-LIFE time, before requesting a connection, to guarantee that initial sequence numbers are unambiguous. S~uence numbers are incremen~ed modulo MAX-SEQ, where NAX-SEQ is large enough to insure that duplicate sequence numbers do not exist in the network. Packet types for establishing and breaking connections are defined as in TCP.
A TC~-like connection is full-duplex and a sliding window is used to allow multiple outstanding transport packets. An ARQ bit in the transport header is used to require an immediate acknowledgment from the opposite end.
VMTP-like connections are u~ed for transient messages (i.e. terminal-to-terminal m~il messages).
VMTP-like connection records are built automatically.
A VMTP~ e :connection recsrd is built (or updated) whenever a ~ P-like transport mescage is received.
The advantage is~ ~hat an explicit connection re~uest is not reguired. The disadvantage is that longer and :20 more carefully se~ected se~uence number~ are required.
A V~P-like connection is half-duplex. (~ full-dup~ex connQction at a~ higher layer can be built with two independent half-duplex VMTP-like connections.) Acknowledgments mu t be handled by highe~ layers.
:
25:: Transport connections are defined by the network end-to-end~destination and sourc~ addresses.
A NAX rP LIFE timeout is assoaiated with transport connections. Transport connection records are purged after a MAX TP LIFE time ~xpires without activity on the connection. Th~ transport entity in ~~ a ter~inal can ensure that its transport connec~ion :~ will not be lost by transmitting an empty time-fill transport packet whenever TP_TIMEOUT time expires withou~ activity.

WO93/07691 PCr/US92/08610 ~à ~ U 3 4 The transport entity in a node stores ~essages for possible retransmission. Note that retransmissions may not always follow the same path (primarily) due to moving terminals and the resulting changes in ~the spanning tree. For examplej the network entity in a parent node may disconneGt a child after the DL~ entity reports a message deli~ery failure. The child will soon disco~er that it is detached~and will reattach to the spanning tree. Now when the transport entity (i.e. in the root) re-sends the message,~it will follow the new path.

Transpo~t message timing and sleeping terminals.
The transport entity in a terminal calculates a separate timeout for SEND and TRANSACTION operations.
Initially,~both timeouts are a $unction of the distance of ~he terminal from the root node.
A TCP-like algorithm is used to estimate the expected propagation delay for each message type.
2;0 Messages,~which require a response, are retransmitted if twice the expected propagation time expires be~ore a respQnss is received. SLEEPING terminals can power down for ~a~ large percentage of the expected propagation~delay before waking up to receive the 25~ response ~message. Note that missed messages may be stored~by~the~network layer for "count~' hello times.

;Medium Access;Control (N~C) ~heory and implementation notes.
~Access to the ne~work communications channel is ~egulated in~several ways: executing the full CSMA
algorit~ ~(see MAC layer~ above). The sender retransmits unacknowledged~messages until a R~TRY MAX
~ ~ , ~ count îs exhausted.
:
:

: :: `

,~WO93/07691 ~.. l ? ~ ` .; PCT/US92/08610 The retry time of the DLC must be relatively short so that lost nodes can be detected quickly.
When the DLC layer reports a failure to deliver a message to the network layer, the network layer can l) save messages for SLEEPING terminals for later attempts, or 2) DETACH the node from ~he spanning tree. Note that most lost nodes are due to moving : terminal The node identifier part of the DLC address is : ~ lO initially all O's for all nodes except the root node.
The all~O's address is used by a node to send and : received data-link frames until a unique node identifier~is`~passed to the DLC entity in the node.
~The unique:node identifier is obtained by the network entity.) Address resolution.
Well-known` names too are bound to network addresses in~several ways:
: ~ ~- The;n~twork address and TRANSPORT ACC~SS
' : ID of~a name~:server, contained in the root, is : well-known b~ all nodes.
A node can register a well-known name with`;the name server contained in the root node.
5~ A node can request the network access address~of another applîcation from the name server~by using the~ well-known name of the ; application.~

` : : : ` :
Possible extensions.
Base~station-to-base station traffic could also be routed~ through th~ controller if the backward : learning~algorithm included b~se station nodes. ~Each ~ : base station~: would simply have to rememb~r which : ~:

WO ~3tO7S91 PCT/US92/08~10 ~
. , ~ 20 - 36 -direction on its branch of the spanning tree to send data directed toward another base station.) The possibility of multiple controllers is kept open by including a spanning-tree identifier in address field~. Each controller d~fines a unique spanning tree. A node can be in more than one spanning tree, with separate network state variables defined for each.

~: 10 Thus, the preferred embodiment of the pre~ent invention describes an apparatus and a method of : ef~iciently routing data through a network of : intermediate base stations in a radio data ; communication system.
In alternate embodiments of the present ~: ~ invention,~the RF Networks contain multiple gateways.
By including a system identifier in the address field : of the nodes, it is possible to determine which nodes are~connected to which networks.
: 20 As :is ::evident from the description that is provided ~above, the implementation of the present invention ~ can vdry~ greatly depending upon the desired goal of the:~user~ However, the scope of the present invention:~is intended to cover a~l variations and ~ subst~itut~ions:which are and which may become apparent ; from~the ~ lustrative embodiment of the present invention that is provided above, and the scope of the ; invention should be extended to the claimed inve~tion and its equivalents.

:

: -:

Claims (10)

WHAT IS CLAIMED IS:

1. A multi-hop data communication network having RF capability comprising:
a host computer node;
a plurality of terminal nodes; and a plurality of bridging nodes which dynamically create and revise communication pathways between any two nodes in the network, each of the bridging nodes independently storing and maintaining local information that specifies how communication traffic should flow through that bridging node, and the plurality of bridging nodes, together, providing a complete specification for the communication pathways in the multi-hop communication network.

2. A multi-hop data communication system comprising:
a root node;
a plurality of terminal nodes;
a plurality of bridging nodes which dynamically arrange communication pathways between the root node and the other nodes in the network; and said nodes using a backward learning technique to independently create and maintain locally stored information to specify how communication traffic should flow through that bridging node.

3. A data communication system providing communication pathways between comprising:
a plurality of terminal nodes; and a plurality of bridging nodes dynamically interconnecting the terminal nodes to provide communication pathways between the terminal nodes, each of the bridging nodes independently storing and updating local information that specifies the communication pathways through that bridging node.

4. A multi-hop data communication system providing communication pathways comprising:
a first communication node;
a second communication node from which it is desired to initiate communication with the first communication node;
a plurality of intermediate communication nodes; and the communication nodes dynamically arranging communication pathways to provide minimum number of hops between the first and second nodes without overloading any one intermediate node.

5. A multi-hop data communication system providing communication pathways comprising:
a first and a second communication node between which communication is desired;
a plurality of intermediate communication nodes;
said communication nodes using backward learning techniques dynamically interconnecting the first and second communication nodes to provide a communication pathway with the lowest number of hops possible without overloading any one intermediate communication node; and said communication nodes maintaining current information specifying how communication traffic should flow therethrough.

6. In a multi-hop data communication system having a computer node and a plurality of communication nodes, a method for providing communication among the nodes comprising the steps of:
(a) establishing a communication link between the computer node and one of the plurality of communication nodes;
(b) indicating from the communication node which has been linked to those of the remaining, unlinked communication nodes that a communication link has been established;
(c) establishing a communication link between the unlinked communication nodes which receive the indication and the linked communication node;
(d) repeating steps (b) and (c) for each communication node which has established a communication link until all of the plurality of communication nodes have been linked; and (e) requesting, by any of the communication nodes which become unlinked, a communication link and branching to step (b).

7. In a multi-hop data communication system having a first and second communication node and a plurality of intermediate communication nodes, a method for providing communication pathways among the communication nodes comprising the steps of:
(a) establishing for the first communication node a communication link with at least one of the plurality of intermediate communication nodes;
(b) indicating by each of the linked intermediate communication nodes that a communication link has been established and providing the hopping distance of that link;
(c) analyzing, by the intermediate communication nodes which receive the indication and by the second communication node if it receives the indication, the indication to determine whether to establish a communication link with the intermediate communication node providing the indication, and, if the analysis so indicates, establishing the communication link;
(d) branching to step (b) if the second communication node has not been linked; and (e) requesting, by any of the communication nodes which become unlinked, a communication link and branching to step (b).

8. In a multi-hop data communication system having a root node and a plurality of communication nodes a method for providing and maintaining communication pathways among the nodes comprising the steps of:
(a) indicating by the root node that communication links may be established;
(b) analyzing, by the communication nodes which receive the indication, the indication to determine whether to establish a communication link with the root node providing the indication, and, if the analysis so indicates, establishing the communication link;
(c) indicating by each of the linked communication nodes the hopping distance of the line to the root node;
(d) analyzing, by the communication nodes which receive the indication, the indication to determine whether to establish a communication link with the linked communication node providing the indication, and, if the analysis so indicates, establishing the communication link;

(e) branching to step (c) until all intermediate communication nodes have been linked; and (f) requesting, by any of the intermediate nodes which become unlinked, a communication link and branching to step (c).

9. In a multi-hop data communication system having a root node, a plurality of communication nodes, and a plurality of terminal nodes a method for providing communication pathways among the communication nodes comprising the steps of:
(a) indicating by the root node a readiness to establish communication links;
(b) establishing communication links between the root node and the communication nodes receiving the indication;
(c) indicating by each of the linked communication nodes the number of communication links which have been established and providing the hopping distance of the link to the root node;
(d) analyzing, by the communication nodes which receive the indication, the indication to determine whether to establish a communication link with the linked communication node providing the indication, and, if the analysis so indicates, establishing the communication link;
(e) branching to step (c) until all communication nodes have been linked;
(f) indicating by the linked communication nodes the hopping distance of the link to the root node;
(g) analyzing, by the plurality of terminal nodes, the indications which are received to determine which linked communication nodes to establish, and, each terminal node establishing at least one communication link;
(h) requesting, by any of the plurality of terminal nodes which become unlinked, a communication link and branching to step (f); and (i) requesting, by any communication node which becomes unlinked, a communication link and branching to step (c).

10. In a multi-hop data communication system having a plurality of communication nodes, a method for providing communication pathways among the communication nodes comprising the steps of:
(a) selecting one of the communication nodes to be a root node;
(b) indicating by the root node that communication links may be established;
(c) establishing communication links between the root node and the communication nodes receiving the indication;
(d) indicating by each of the linked communication nodes the hopping distance of the link to the first communication node;
(e) analyzing, by the linked communication nodes which receive each indication, the indications to determine whether to replace the current communication link with a link to the communication node providing the indication, and, if the analysis so indicates, replacing the communication link;
(f) analyzing, by the unlinked communication nodes which receive each indication, the indications to determine whether to establish a communication link with the communication node providing the indication, and, if the analysis so indicates, establishing the communication link;

(g) branching to step (c) until all communication nodes have been linked.