WO2009040777A2

WO2009040777A2 - Leader robot election in collaborative robot groups

Info

Publication number: WO2009040777A2
Application number: PCT/IB2008/053963
Authority: WO
Inventors: Harsh Dhand; Nagaraju Bussa
Original assignee: Koninklijke Philips Electronics, N.V.; U.S. Philips Corporation
Priority date: 2007-09-27
Filing date: 2008-09-29
Publication date: 2009-04-02
Also published as: WO2009040777A3; TW201015455A

Abstract

A system and method for electing a leader in a robot group includes assigning (14) tasks to a group of robots to collect information. A leader is elected (16) by designating (30) each robot as a leader and comparing (32) a criteria related to the collected information between robots of the group to determine a non-leader robot at each comparison until a sole leader is determined by being a remaining robot designated as a leader.

Description

LEADER ROBOT ELECTION IN COLLABORATIVE ROBOT GROUPS

This disclosure relates to robots, and more particularly to a system and method for electing a leader from a group of robots in a collaborative relationship. Collaborative robots are employed today in a wide variety of applications.

These applications may include infrastructure security, new area exploration, knowledge based discovery, and many other applications. Individual intelligent robots contribute by solving a part of the problem. While small groups of collaborating robots are referred to as teams, large groups are referred to as swarms. Another classification of collaborative robots is based on whether the robots are homogenous or heterogeneous in nature. Swarms are typically homogenous in nature because of the nature of tasks that the swarms usually perform, while a team of robots could be either homogenous or heterogeneous.

One important aspect in collaborative robotics is the presence or absence of a centralized controller. The robots could be acting on their own, or the robots could be governed by a central controller. Dynamic task assignments for the individual robots in a swarm using strategies like random-choice decisions, extreme communications with all robots communicating, card-dealers algorithm, etc. have been advocated, but the result is deterministic in nature and will lead to optimal results only in the case of homogenous robots. U.S. Patent Number 5,825,981 to Matsude has proposed a human intervention based task assignment for individual robots in a team for carrying out actions like multi-functional manufacturing, cleaning, sowing or interacting with humans. In this approach, the human always needs to be aware of the situation and appropriately control the individual robots, leading to a situation where the swarm behavior is not automated. Horace et al. (International Publication No. WO 2004/018158) and Neal E.

Solomon (US Publication No. 2006/0167917) have both proposed self-organizing dynamic multi-robot systems. Their proposed methodology involves each robotic agent learning together and making a distributed decision model. In such a model, robots will again perform actions on their own (not having a centralized controller for assigning appropriate tasks to appropriate robots at that instant) and can lead to a conflicting task assignment/end goal. The robots revisit the task assignment problem once a conflicting action takes place and try to rectify the task's allocation, leading to a suboptimal solution in terms of achieving the goal.

Therefore, it would be advantageous to provide a leader/controller election and team co-ordination method for efficient collaboration in dynamic action scenarios.

In accordance with the present principles, a novel information based dynamic leader election, which enables collaborative robots to deal with dynamic scenarios in an optimal manner, is provided. A correct leader election is followed by leader-initiated co-ordination, which further improves a solution for collaborative scenarios. In accordance with one embodiment, the solution is adaptive, since it depends on the information at a particular time. The controller may even be elected dynamically and regularly updated to enable the best possible solution, without any human intervention.

A system and method for electing a leader in a robot group includes assigning tasks to a group of robots to collect information. A leader is elected by designating each robot as a leader and comparing a criteria related to the collected information between robots of the group to determine a non-leader robot at each comparison until a sole leader is determined by being a remaining robot designated as a leader.

In other systems and methods for electing a leader in a robot group, a temporary leader robot from a group of robots is randomly assigned. Tasks are assigned to the group of robots by the temporary leader to collect information. A new leader is elected by labeling each robot as the new leader, and comparing a criterion related to the collected information between robots of the group to determine a non-leader robot at each comparison until the new leader is determined by being a sole remaining robot labeled as the new leader.

A system of robots includes a plurality of robots. Each robot includes a processing device and memory. The memory includes a leadership metric where the leadership metric is based upon a feature of a given robot. A communication mechanism of each robot is configured to permit each robot to communicate their leadership metric to each other robot. The processing device is configured to compare the leadership metric of the given robot to other robots to which the given robot has communicated, until a determination that the given robot has a leadership metric which makes the given robot less suitable for leadership. Otherwise, the given robot is determined a leader if no remaining robots have a more suitable leadership metric. These and other objects, features and advantages of the present disclosure will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in connection with the accompanying drawings.

This disclosure will present in detail the following description of preferred embodiments with reference to the following figures wherein:

FIG. 1 is a block/flow diagram showing an illustrative system/method for selecting a leader in a group of robots;

FIG. 2 is a block/flow diagram showing an illustrative system/method for selecting the leader in greater detail; FIG. 3 is a block diagram showing a team of robots communicating to compare information size before a leader is elected in accordance with one illustrative embodiment;

FIG. 4 is a block diagram showing the team of robots of FIG. 3 after a leader is elected and showing the leader coordinating the other robots in accordance with one illustrative embodiment; and

FIG. 5 is a block diagram showing a plurality of robots where one robot is shown in an exploded view to show some of its features in accordance with one embodiment.

Robots today are employed in a large domain of activities. Collaborative robots are becoming popular, with individual robots having the intelligence and the communication capabilities to work in a team. When a collaborative action needs to be done the coordination between the individual intelligent robots plays a role. To enable the optimum utilization of the individual strengths towards a common goal, a centralized monitor is needed. In particularly useful embodiments, a novel way of achieving collaborative behavior is by employing an information-based dynamic leader (e.g., a centralized monitor) election in a team or swarm. This helps in achieving the best team behavior and enables adaptation to different scenarios, since the leader also utilizes information from the other swarm members. The present solution is better than both conventional swarm behavior and pre-defined central or control robots. The present embodiments will improve the performance of many collaborative systems.

It should be understood that the present invention will be described in terms of robots. A robot as described herein includes a device or machine capable of basic decision making capabilities, but may also include highly advanced robots with artificial intelligence. In preferred embodiments, robots may include one or more processing chips, and may include memory for storage of data, software program(s) and possibly an operating system. Some robots may include telecommunications capabilities such that the robots can communicate with each other. Communication between robots may be carried out using wired or wireless technology, light signals, sound or any other suitable communication method. In some embodiments, the robots may include artificial intelligence for making decisions or to adaptively learn from dynamic scenarios. Such intelligence can be employed in adding additional features and improving the accuracy and efficiency of performing assigned tasks. The robots may include transportation features to permit movement of the robots, or the robots can be stationary depending on the application. The robots may include their own power source or derive power from external sources.

While applicable to groups (teams, swarms, etc.) of robots, the teachings of the present invention are much broader and are applicable to any components that are capable of electing a hierarchy from among members of the group. The robots contemplated herein can be arranged or grouped in any number of scenarios and may include homogenous and heterogeneous configurations.

It should be understood that the elements shown in the FIGS, may be implemented in various forms of hardware, software or combinations thereof. Preferably, these elements are implemented in a combination of hardware and software, where the software runs on one or more appropriately programmed general-purpose digital computers or computer-like devices (including robots) having a processor and memory and input/output interfaces. These elements can provide functions which may be combined in a single element or multiple elements.

Referring now to the drawings in which like numerals represent the same or similar elements and initially to FIG. 1, a block/flow diagram 10 shows one illustrative system/method for electing a lead robot in a group of robots. The leader election is autonomous by the group. Advantageously, no external intelligence is needed to initiate and perform the leader election. A synchronization mechanism is preferably employed such as a global timer so that all of the robots are synchronized. Many synchronization techniques may be employed to provide the global timing. For example, a timing signal may be received by an external source, or the robots may have their own synchronization device or mechanism which is synchronized intermittently with the other robots. In block 12, a leader is initially elected randomly, to execute a task of collecting initial information in a given time (e.g., T₁ - Time for information collection) in block 14. The initial election may include a lottery selection by one of the robots or a program may be employed to randomly select a temporary leader. After the information collection phase in block 14, leader election takes place in block 16. The leader election in a group of individual intelligent robots performing tasks collaboratively is preferably performed on-the-fly. The leader election in block 16 is shown in greater detail in FIG. 2. Referring to FIG. 2, a time T_c is allocated for the robots to communicate with each other. In block 30, the status of all robots is set to a variable 'Y' (denoting that "Yes, I am leader"). Each robot then compares its information with the others in block 32. If a robot has less information than any other robot, then it assumes 'N' (denoting not-a-leader) in block 34. This ensures that only one robot is left with the 'Y' assumption in block 36. To deal with the case of equal information (e.g., a tie), one robot retains the 'Y' value while the other assumes 'N'. This can be performed in block 38 by applying secondary criteria or randomly selecting one of the remaining robots. Secondary criteria may be dynamic or arbitrary. It may be based on a lower serial number or ID number, amount of data, power level, or any other criteria. The amount of information used to make the comparison in a leader election may include, e.g., a number of data points collected, a number of memory locations with stored data, an amount of stored data, an activity time (time the robot was actively involved in performing a task), stored energy remaining, distance traveled, current speed, or any other useful criteria or information that dynamically determines fitness to be elected a leader. Referring again to FIG. 1, in block 18, now the elected leader collects information about the capabilities and parameters of the group of robots. The leader may collect a plurality of different types of information, e.g., distributed information about a particular problem to be solved or task and/or any other parameters, such as position information, speed, power levels, computational capacity, environmental conditions, etc. The robots can continue updating their information to the leader. In block 20, the elected leader assigns the tasks that need to be performed to the other robots or need to be performed in accordance with a particular task. The task assignments are preferably optimized to provide efficient execution of the tasks at hand. This may include selecting which robots perform which tasks based on the capacity of the robots to perform the given tasks. For example, in a data collection task, a robot having a highest amount of unused memory and a fastest data collection rate may be preferred. Such a robot or robots would be selected by the leader to perform the task over other available robots. Once the tasks are completed by the individual robots, as determined in block 22, the process of leader election can take place again in block 16.

In accordance with one illustrative embodiment, the present principles may be described in terms of a particular problem to be solved. In this example, a specific collaborative robot scenario of a maze navigation will be described. The objective in collaborative maze navigation is to traverse the maze from one end to another by a team of robots. One important step in collaboration is the exchange of appropriate maze map information between robots to improve solution time.

In swarm behavior with no leader, the robots would all try to find a path and hence act as individual robots. The individual robot solution is sub-optimal in terms of the time taken for navigating through the maze. Using embodiments of the present invention, an appropriate robot can be selected as the leader at every appropriate time instant, and can be used to guide all other robots to navigate in an efficient manner reducing the collection of redundant information and more rapidly determining the solution. The present embodiments build on peer-to-peer communication among robots, which is highly scalable to any number of robots, and may be employed for a swarm of robots. In addition, the leader election assigns only one robot to the leader role, which avoids timely negotiations and conflicts that can occur in the prior art solutions. The bottleneck in many applications is this negotiation and communication which the present invention avoids. The assignment of tasks to the robots in accordance with the present principles is preferably based on an optimal allocation from information about robot capabilities.

Referring to FIG. 3 , robotic entities 102 are shown which exchange information needed for leader election. In this example, a team 100 of four robotic entities 102 exchange information to elect or choose a leader robot where robot 1 emerges as the leader (FIG. 4).

Each of robots 1-4 includes a label 104, which is set to 'Y' indicating that each robot initially believes itself to be the leader. Each robot includes information 106 which will be employed for comparison between robots to elect the leader. Information 106 in this embodiment is stored in a vector structure {V}_n of size n. According to this example, robot 1 includes {V}s , robot 2 includes {V}3, robot 3 includes (V)₄ and robot 4 includes {V}₂. The robots communicate over communication channels or media 108 to compare information size (n) to each of the other robots.

Referring to FIG. 4, a leader (robot 1) is elected based on the current information that the robots have. The robot having the maximum information (n=5, in the example) is elected as the leader (robot 1 in the example). As each robot communicates, the label 104 is changed to 'N' in accordance with a determination that the amount of information for that robot is less than that of the other robot being communicated with.

The elected leader will ensure efficiency in terms of resources being used. The leader himself has maximum information and hence is able to lead in the tasks. Swarm intelligence involves multiple robots acting towards a common objective. There might be replication of certain tasks or tasks not being addressed in such a scenario. The leader chooses an optimal task distribution. This may include determining that all tasks have an assigned robot or robots. With complete information of all the peer robots, the leader is in a position to perform task allocation. Criteria may be set to optimize this distribution. For example, a number of robots may be allocated for a given task in a way computed to complete the task before a given time has elapsed. In another example, power consumption for the group of robots can be optimized by selecting robots with the lowest power requirement to perform certain tasks to reduce power. Other optimizations are also contemplated. Such optimizations may employ task or work flow optimization programs.

While the amount of information is employed in the above example as the criteria for electing a leader, other parameters may also be employed to make the leader election decision, e.g., a robot having maximum speed, robot having maximum lifting potential, maximum slope navigation capability, maximum stored energy, a combination of special features or desirable features, etc., based on the specific scenario of the collaborative action. In one example, a robot having a combination of features such as a desirable position, speed and payload at the time of the election may be employed as the criteria for leader election.

The information gathered and used for leader election may include a plurality of different types or combinations thereof. The information may include, e.g., maps (to solve a navigation problem), information about weights (e.g., a weight transport problem), terrain geography (outdoor/exploratory environment), etc. The present invention has application in all collaborative robotic scenarios and is particularly useful in dynamic scenarios where future decisions are based on collected information. The leader is preferably selected based on data collected on the fly. The present approach ensures efficiency and optimality in the collaborative robot problem, solves the leader election problem and also ensures that the team of robots performing a task is best suited to perform that task.

Referring to FIG. 5, a system 200 of robots, which collaborate to perform one or more tasks is illustratively shown in accordance with the present principles. The system 200 includes a plurality of robots 202, which may be a team or swarm of robots. As before, the robots may be heterogeneous or homogeneous. Each robot 202 includes a processing device 216 and memory 210. The memory 210 stores collected information 212 and other information including, e.g., a leadership metric 214. The leadership metric 214 may be based upon a feature of a given robot. For example, the features may be the amount of information collected during an initial or other collection phase period. The leadership metric 214 may also include a power or memory capacity, an amount of information collected by a robot, a speed or lifting capacity, a position or location, an operating condition, a measured environmental condition or any other metric. Each robot 202 includes a communication mechanism or device 208 configured to permit each robot to communicate their leadership metric to each other robot. Communication can be carried out through wireless or wired (docking) communications, light signals, optical processing, sound, etc. The processing device 216 may include logic circuitry, a microprocessor or other processing device(s). The processing device 216 is configured to compare the leadership metric 214 of a given robot to other robots to which the given robot has communicated. A determination is made as to whether the given robot has a leadership metric 214 which makes the given robot less suitable for leadership. If a robot is less suitable for leadership, the robot is designated as a non-leader (N). This exchange continues until only one robot remains designated as the leader. A robot is determined a leader if no remaining robots have a more suitable leadership metric. The memory of each robot may include a leader designation label 206, which is defaulted initially to indicate that a robot is a leader. During the exchange phase, the designation label 206 is switched to "not-a-leader" when a robot with a more suitable leadership metric is encountered.

The system 200 includes a synchronization mechanism 204 which is configured to synchronize the plurality of robots. Synchronization mechanism 204 may be a local device, distributed over the robots or receive synchronization signals from an external source. The synchronization mechanism 204 monitors elapsed time, and a new leader is preferably elected after an elapsed time period. This ensures a new leader or at least a reevaluation of the leadership to provide a best possible leader under given conditions. The leader is configured to assign tasks to other robots in the system. Other systems 218 on robots 202 may include sensors, power sources or supplies, guidance systems, physical storage space, mechanical devices, such as arms, wheels tracks, cutters, etc.

In interpreting the appended claims, it should be understood that: a) the word "comprising" does not exclude the presence of other elements or acts than those listed in a given claim; b) the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements; c) any reference signs in the claims do not limit their scope; d) several "means" may be represented by the same item or hardware or software implemented structure or function; and e) no specific sequence of acts is intended to be required unless specifically indicated.

Having described preferred embodiments for systems and methods for leader robot election in collaborative robot groups (which are intended to be illustrative and not limiting), it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings. It is therefore to be understood that changes may be made in the particular embodiments of the disclosure disclosed which are within the scope and spirit of the embodiments disclosed herein as outlined by the appended claims. Having thus described the details and particularity required by the patent laws, what is claimed and desired protected by Letters Patent is set forth in the appended claims.

Claims

CLAIMS;

1. A method for electing a leader in a robot group, comprising: assigning (14) tasks to a group of robots to collect information; and electing (16) a leader by: designating (30) each robot as a leader; and comparing (32) a criterion related to the collected information between robots of the group to determine a non-leader robot at each comparison until a sole leader is determined by being a remaining robot designated as a leader.

2. The method as recited in claim 1, wherein the criteria includes an amount of collected information.

3. The method as recited in claim 1, further comprising collecting (18) information regarding capabilities and parameters of the group by the sole leader.

4. The method as recited in claim 3, further comprising assigning (20) tasks to the group of robots based on collected information regarding capabilities and parameters.

5. The method as recited in claim 1, further comprising, after a time period, repeating (22) the steps of designating and comparing until a task is completed.

6. The method as recited in claim 1, further comprising, when the comparing results in a tie, employing (38) a secondary criteria to break the tie to determine the sole leader.

7. The method as recited in claim 1, wherein the group includes heterogeneous robots.

8. A method for electing a leader in a robot group, comprising: randomly assigning (12) a temporary leader robot from a group of robots; assigning tasks (14) to the group of robots by the temporary leader to collect information; and electing (16) a new leader by: labeling (30) each robot as the new leader; and comparing (32) a criterion related to the collected information between robots of the group to determine a non-leader robot at each comparison until the new leader is determined by being a sole remaining robot labeled as the new leader.

9. The method as recited in claim 8, wherein the criteria includes an amount of collected information.

10. The method as recited in claim 8, further comprising collecting (18) information regarding capabilities and parameters of the group by the new leader.

11. The method as recited in claim 10, further comprising assigning (20) tasks to the group of robots based on the collecting information regarding capabilities and parameters.

12. The method as recited in claim 8, further comprising, after a time period, repeating (22) the steps of designating, and comparing until a task is completed.

13. The method as recited in claim 8, further comprising, when the comparing results in a tie, employing (38) a secondary criteria to break the tie to determine the sole leader.

14. The method as recited in claim 8, wherein the group includes heterogeneous robots.

15. A system of robots (200), comprising: a plurality of robots (202), each robot including: a processing device (216) and memory (210); the memory including a leadership metric (214) where the leadership metric is based upon a feature of a given robot; and a communication mechanism (208) configured to permit each robot to communicate their leadership metric to each other robot; the processing device configured to compare the leadership metric of the given robot to other robots to which the given robot has communicated until a determination that the given robot has a leadership metric which makes the given robot less suitable for leadership, otherwise the given robot is determined a leader if no remaining robots have a more suitable leadership metric.

16. The system as recited in claim 15, wherein the feature includes an amount of information collected (212) by a robot.

17. The system as recited in claim 15, wherein the memory includes a leader designation label (206) defaulted to indicate that a robot is a leader, the designation label being switched to not-a-leader when a robot with a more suitable leadership metric is encountered.

18. The system as recited in claim 15, further comprising a synchronization mechanism (204) configured to synchronize the plurality of robots.

19. The system as recited in claim 18, wherein the synchronization mechanism (204) monitors elapsed time and a new leader is elected after an elapsed time period.

20. The system as recited in claim 15, wherein the leader is configured to assign tasks to other robots in the system.

1 ?