Abstract—As the community strives towards autonomous multirobot systems, there is a need for these systems to autonomously form coalitions to complete assigned missions. Numerous coalition formation algorithms have been proposed in the software agent literature. Algorithms exist that form agent coalitions in both super additive and non-super additive environments. The algorithmic techniques vary from negotiation-based protocols in multi-agent system (MAS) environments to those based on computation in distributed problem solving (DPS) environments. Coalition formation behaviors have also been discussed in relation to game theory. Despite the plethora of MAS coalition formation literature, to the best of our knowledge none of the proposed algorithms have been demonstrated with an actual multi-robot system. There exists a discrepancy between the multi-agent algorithms and their applicability to the multi-robot domain. This paper aims to bridge that discrepancy by unearthing the issues that arise while attempting to tailor these algorithms to the multi-robot domain. A well-known multi-agent coalition formation algorithm has been studied in order to identify the necessary modifications to facilitate its application to the multi-robot domain. This paper reports multi-robot coalition formation results based upon simulation and actual robot experiments. A multi-agent coalition formation algorithm has been demonstrated on an actual robot system. Index Terms—Coalition formation, coalition imbalance, task allocation.

Robot foraging, a frequently used test application for collective robotics, consists in a group of robots retrieving a set of opportunely defined objects to a target location. A commonly observed experimental result is that the retrieving efficiency of the group of robots, measured for example as the number of units retrieved by a robot in a given time interval, tends to decrease with increasing group sizes. In this paper we describe a biology inspired method for tuning the number of foraging robots in order to improve the group efficiency. As a result of our experiments, in which robots use only locally available information and do not communicate with each other, we observe self-organised task allocation. This task allocation is effective in exploiting mechanical differences among the robots inducing specialisation in the robots activities.

In this paper, we present a paradigm for coordinating multiple robots in the execution of cooperative tasks. The basic idea in the paper is to assign to each robot in the team, a role that determines its actions during the cooperation. The robots dynamically assume and exchange roles in a synchronized manner in order to perform the task successfully, adapting to unexpected events in the environment. We model this mechanism using a hybrid systems framework and apply it in different cooperative tasks: cooperative manipulation and cooperative search and transportation. Simulations and real experiments demonstrating the effectiveness of the proposed paradigm are presented. 1

Largely because of significant qualitative differences between the respective systems under study, the multi-agent and multi-robot research communities have each developed their own methods for perception, reasoning, and action in individual agents/robots. In particular, the multi-robot community has historically studied both implicit and explicit coordination techniques. Implicit coordination techniques employ dynamics of interaction among the robots and the environment in order to achieve the desired collective performance, often in the form of designed emergent behavior. These methods, while often very elegant and efficient, have so far defied general analysis, but show great promise in particular for large-scale teams of simple individuals, and are being studied actively in robotics as well as in several other fields (including mathematics, artificial life, etc.). Explicit coordination techniques, in comparison, deal with comparatively more sophisticated agents/robots, and employ intentional communication and collaboration methods much like those employed in multi-agent systems. Therefore, at the level of explicit coordination among multiple individuals, the differences between techniques used in multi-agent systems (MAS) and those used in multi-robot systems (MRS) are in fact very few. This is not to say that MAS and MRS are equivalent in any fundamental way, but rather that although robotics researchers employ sophisticated specialized techniques of various sorts (e.g., controltheoretic

Introduction Machine learning (ML) [24] is a means of automatically generating solutions that perform better than those that are hand-coded by human programmers. Such improvement is possible in problem domains where optimal solutions are di#cult to identify, i.e., when there are no models available that can accurately relate a system's dynamics to its performance. One such domain is the control of multi-robot systems. Mobile robots are notoriously di#cult to control in a robust, reliable, and repeatable fashion. The challenges stem from uncertainty inherent in physically embodied systems, including in sensors, e#ectors, and interactions between the system components and the environment. The behavior-based (BB) control paradigm [22, 1] provides a means of structuring robot controllers into collections of task-achieving modules or behaviors, such as exploration and obstacle avoidance. The modules operate in parallel and interact within the system and also through their e#ects on the en

Summary. Task allocation is an issue that every multi-robot system must address. Recent task allocation solutions propose an auction based approach wherein robots bid for tasks based on cost functions for performing a task. This paper presents RACHNA, a novel architecture for multi-robot task allocation based on a modified algorithm for the winner determination problem in multi-unit combinatorial auctions. A more generic utility based framework is proposed to accommodate different types of tasks and task environments. Preliminary experiments yield promising results demonstrating the system’s superiority over simple task allocation techniques. 1

A problem that has recently attracted the attention of the research community is the autonomous formation of robot teams to perform complex multirobot tasks. The corresponding problem for software agents is also known in the multi-agent community as the coalition formation problem. Numerous algorithms for software agent coalition formation have been provided that allow for efficient cooperation in both competitive and cooperative environments. However, despite the plethora of relevant literature on the software agent coalition formation problem, and the existence of similar problems in theoretical computer science, the multirobot coalition formation problem has not been sufficiently grounded for different tasks and task environments. In this paper, comparisons are drawn to highlight the differences between software agents and robotics, and parallel problems from theoretical computer science are identified. This paper further explores robot coalition formation in different practical robotic environments. A heuristic-based coalition formation algorithm from our previous work was extended to operate in precedence ordered cooperative environments. In order to explore coalition formation in competitive environments, the paper also studies the RACHNA system, a market based coalition formation system. Finally, the paper investigates the notion of task preemption for complex multi-robot tasks in random allocation environments.

Abstract. Task allocation is a fundamental problem that any multirobot system must address. Numerous multi-robot task allocation schemes have been proposed over the past decade. A vast majority of these schemes address the problem of assigning a single robot to each task. However as the complexity of multi-robot tasks increases, often situations arise where multiple robot teams need to be assigned to a set of tasks. This problem, also known as the coalition formation problem has received relatively little attention in the multi-robot community. This paper provides a generic, task independent framework for solutions to this problem for a variety task environments. In particular, the paper introduces RACHNA, a novel auction based coalition formation system for dynamic task environments. This is an extension to our previous work which proposed a static multi-robot coalition formation algorithm based on a popular heuristic from the Distributed Artificial Intelligence literature. 1

Existing multi-robot task allocation algorithms (MRTA) generally do not consider the e#ects of interaction, such as interference, but instead assume that tasks are independent. That assumption, however, is often violated in groups of cooperative mobile robots, where interaction e#ects can have a critical impact on performance. Modeling the e#ects of interaction, or group dynamics, is problematic due to their complexity and volatility, which also makes it di#cult to hand-code optimal solutions to MRTA problems. We present a distributed MRTA algorithm that is sensitive to interaction dynamics. The algorithm uses distributed reinforcement learning to make interference-sensitive estimates of task utilities and relies on stigmergy to let optimal allocations emerge. The algorithm is inspired by vacancy chains, a resource distribution process common in human and animal societies. We present a formal model of task allocation through vacancy chains that defines optimal allocations in MRTA problems in terms of interference-sensitive measures of individual robot contributions. We validate our model by demonstrating, in simulation, how the predicted allocations are produced by our algorithm. As the robots continuously update their individual utility estimates, the vacancy chain algorithm has the additional property of adapting automatically to changes in the environment, e.g., robot breakdowns or changes in task values. We study the problem of cooperative transportation and demonstrate that our algorithm improves on the performance of hand-coded solutions. Finally, as the vacancy chain algorithm uses no communication or unique roles and is more robust and scalable than alternative algorithms with significant communication overheads.

Abstract. A major challenge in the field of Multi-Agent Systems is to enable autonomous agents to allocate tasks efficiently. This paper extends previous work on an approach to the collective iterative allocation problem where a group of agents endeavours to find the best allocations possible through refinements of these allocations over time. For each iteration, each agent proposes an allocation based on its model of the problem domain, then one of the proposed allocations is selected and executed which enables us to assess if subsequent allocations should be refined. We offer an efficient algorithm capturing this process, and then report on theoretical and empirical results that analyse the role of three conditions in the performance of the algorithm: accuracy of agents’ estimations of the performance of a task, the degree of optimism, and the type of group decision policy that determines which allocation is selected after each proposal phase.