A Reinforcement Learning Algorithm in Cooperative Multi-Robot Domains

Fernández, Fernando; Borrajo, Daniel; Parker, Lynne E.

doi:10.1007/s10846-005-5137-x

Cited by 49 publications

(35 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…H 10 and H 20 were accepted with P-values equal to 1.66·10 −6 and 0.01, respectively. This means that the average time and average reward to complete emptying the contents of the bag measured over the last ten learning episodes are not equal while comparing Q(λ) with CQ(λ) exhibiting lower average time to extract objects at a higher reward.…”

Section: Resultsmentioning

confidence: 99%

A Human-Robot Collaborative Reinforcement Learning Algorithm

Kartoun

Stern

Edan

2010

J Intell Robot Syst

View full text Add to dashboard Cite

This paper presents a new reinforcement learning algorithm that enables collaborative learning between a robot and a human. The algorithm which is based on the Q(λ) approach expedites the learning process by taking advantage of human intelligence and expertise. The algorithm denoted as CQ(λ) provides the robot with self awareness to adaptively switch its collaboration level from autonomous (self performing, the robot decides which actions to take, according to its learning function) to semi-autonomous (a human advisor guides the robot and the robot combines this knowledge into its learning function). This awareness is represented by a self test of its learning performance. The approach of variable autonomy is demonstrated and evaluated using a fixed-arm robot for finding the optimal shaking policy to empty the contents of a plastic bag. A comparison between the CQ(λ) and

show abstract

Section: Resultsmentioning

confidence: 99%

A Human-Robot Collaborative Reinforcement Learning Algorithm

Kartoun

Stern

Edan

2010

J Intell Robot Syst

View full text Add to dashboard Cite

show abstract

“…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. However, it still has two main problems.…”

Section: Vqqlmentioning

confidence: 98%

“…The VQQL algorithm (vector quantization for Q-learning) 22,23 is based on the unsupervised discretization of the state space. We use vector quantization methods, and more specificall , the generalized Lloyd algorithm (GLA), 24 also called k-means.…”

Section: Vqqlmentioning

confidence: 99%

“…That algorithm requires a model of the environment. However, we have transformed it into a model-free version, following the ideas introduced for the ENNC-QL algorithm 9,29 and the fitte Q iteration algorithm. 30 In this case, instead of having a model of the environment as input, it needs a set of experience tuples obtained from the interaction of the agent with the environment.…”

Section: A Rl Algorithm Based On a Supervised Learning Function Appromentioning

confidence: 99%

See 1 more Smart Citation

Two steps reinforcement learning

Fernández¹,

Borrajo²

2008

Int. J. Intell. Syst.

Self Cite

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

“…So MRSs have received considerable attention during the last decade [1] [2]. Currently, there has been a great deal of research on multi-agent reinforcement learning (MARL) in MRSs [3].…”

Section: Introductionmentioning

confidence: 99%

Multi-robot Cooperation Based on Hierarchical Reinforcement Learning

Cheng

Shen

Liu

et al. 2007

Computational Science – ICCS 2007

View full text Add to dashboard Cite

Abstract. Multi-agent reinforcement learning for multi-robot systems is a challenging issue in both robotics and artificial intelligence. But multi-agent reinforcement learning is bedeviled by the curse of dimensionality. In this paper, a novel hierarchical reinforcement learning approach named MOMQ is presented for multi-robot cooperation. The performance of MOMQ is demonstrated in three-robot trash collection task.

show abstract

A Reinforcement Learning Algorithm in Cooperative Multi-Robot Domains

Cited by 49 publications

References 15 publications

A Human-Robot Collaborative Reinforcement Learning Algorithm

A Human-Robot Collaborative Reinforcement Learning Algorithm

Two steps reinforcement learning

Multi-robot Cooperation Based on Hierarchical Reinforcement Learning

Contact Info

Product

Resources

About