Cooperative multi-robot observation of multiple moving targets

Parker, Lynne E.; Emmons, B.A.

doi:10.1109/robot.1997.619270

Cited by 190 publications

(253 citation statements)

References 8 publications

Supporting

Mentioning

249

Contrasting

Order By: Relevance

“…26 The vector quantization technique takes advantage of the statistical characteristics of the domain to reduce its size, considering only relevant areas from the whole domain. The algorithm has been applied to learn the ball interception skill of a goalie, 22 and in cooperative multi-robot observation of multiple moving targets (CMOMMT), 27 to obtain collaborative behaviors in a multirobot observation task. 23,28 These experiments show that the nearest prototype approach used to discretize the state space produces better results than uniform discretizations.…”

Section: Vqqlmentioning

confidence: 99%

Two steps reinforcement learning

Fernández¹,

Borrajo²

2008

Int. J. Intell. Syst.

View full text Add to dashboard Cite

When applying reinforcement learning in domains with very large or continuous state spaces, the experience obtained by the learning agent in the interaction with the environment must be generalized. The generalization methods are usually based on the approximation of the value functions used to compute the action policy and tackled in two different ways. On the one hand by using an approximation of the value functions based on a supervized learning method. On the other hand, by discretizing the environment to use a tabular representation of the value functions. In this work, we propose an algorithm that uses both approaches to use the benefit of both mechanisms, allowing a higher performance. The approach is based on two learning phases. In the firs one, a learner is used as a supervized function approximator, but using a machine learning technique which also outputs a state space discretization of the environment, such as nearest prototype classifier or decision trees do. In the second learning phase, the space discretization computed in the firs phase is used to obtain a tabular representation of the value function computed in the previous phase, allowing a tuning of such value function approximation. Experiments in different domains show that executing both learning phases improves the results obtained executing only the firs one. The results take into account the resources used and the performance of the learned behavior. C

show abstract

Section: Vqqlmentioning

confidence: 99%

Two steps reinforcement learning

Fernández¹,

Borrajo²

2008

Int. J. Intell. Syst.

View full text Add to dashboard Cite

show abstract

“…In this paper we instead extend our previous findings aimed to investigate surveillance tasks by multi-robot systems where individual agents use sensors with limited capabilities. We started this research thread with two papers [5][6] aimed to extend the CMOMMT (Cooperative Multi-robot Observation of Multiple Moving Targets) problem initially posed by Parker [11]. One of the main limitations of these algorithms is the requirement that robots operate in open areas.…”

mentioning

confidence: 99%

Multi-robot surveillance: An improved algorithm for the GRAPH-CLEAR problem

Kolling

Carpin

2008

2008 IEEE International Conference on Robotics and Automation

View full text Add to dashboard Cite

Abstract-The main contribution of this paper is an improved algorithm for the GRAPH-CLEAR problem, a novel NP-complete graph theoretic problem we recently introduced as a tool to model multi-robot surveillance tasks. The proposed algorithm combines two previously developed solving techniques and produces strategies that require less robots to be executed. We provide a theoretical framework useful to identify the conditions for the existence of an optimal solution under special circumstances, and a set of mathematical tools characterizing the problem being studied. Finally we also identify a set of open questions deserving more investigations. I. INTRODUCTIONThe use of multi-robot systems for the surveillance of vast regions is one of the well established areas in multi-robot research. Up to now, however, there have been still very few on-field deployments of these systems for real world applications. Besides the obvious matter of cost, another reason for their moderate use is the fact that many basic questions about the efficient coordination of these systems are still unanswered. A big fraction of former theoretical research developed models where robots were equipped with sensors abstractions pretty far from realistic applications, e.g. sensors with infinite range and the alike. In this paper we instead extend our previous findings aimed to investigate surveillance tasks by multi-robot systems where individual agents use sensors with limited capabilities. We started this research thread with two papers [5][6] aimed to extend the CMOMMT (Cooperative Multi-robot Observation of Multiple Moving Targets) problem initially posed by Parker [11]. One of the main limitations of these algorithms is the requirement that robots operate in open areas. Our following efforts have therefore been devoted to scenarios where robots operate in cluttered environments [7]. In particular, we modeled the problem of discovering multiple intruders in a complex environment using a novel graph theoretic problem, dubbed GRAPH-CLEAR. Informally speaking, the problem asks what is the minimum number of robots needed to detect all possible intruders in a given complex environment that can be modeled as a graph. In [8] we proved that the associated decision problem is NP-complete. As clarified later on, a way to circumvent the intractability of the problem on graphs, is to perform certain guard operations that turn graphs into trees. In [7] and [8] we have provided two algorithms that produce search strategies for trees, i.e. course of actions for a robot team that ensures each intruder will be discovered. Both algorithms are known to be suboptimal. In this paper

show abstract

“…Independent of swarming, the problem of multi-robot task allocation(MRTA) has been investigated using different techniques such as physical modeling [18], distributed planning [17] and market-based techniques [6,10]. Task allocation in a multi-robot distributed system using a contract-net based protocol for the COMETS-UAV system has been described in [12].…”

Section: Related Workmentioning

confidence: 99%

Distributed Task Selection in Multi-agent Based Swarms Using Heuristic Strategies

View full text Add to dashboard Cite

Abstract. Swarm-based systems have emerged as an attractive paradigm for implementing distributed autonomous systems for various applications in commercial, military and business domains. One of the major operations in a swarm-based system is to ensure that the individual swarm units process the tasks in the environment in an efficient manner. This can be achieved using a suitable task selection mechanism that allocates the desired number of swarm units to each task while reducing inter-task latencies and communication overhead, and, ensuring adequate commitment of resources to tasks. In this paper, we describe a multi-agent based distributed task selection mechanism for swarm-based systems. We show that the distributed task selection problem is NPcomplete and propose polynomial-time heuristic-based algorithms. Our simulation results show that heuristics in which each swarm unit considers both the effects of other swarm units on tasks and its own relative position to other swarm units achieve better task processing efficiency and improved distribution of swarm units over tasks.

show abstract

Cooperative multi-robot observation of multiple moving targets

Cited by 190 publications

References 8 publications

Two steps reinforcement learning

Two steps reinforcement learning

Multi-robot surveillance: An improved algorithm for the GRAPH-CLEAR problem

Distributed Task Selection in Multi-agent Based Swarms Using Heuristic Strategies

Contact Info

Product

Resources

About