Learning and relearning of target decision strategies in continuous coordinated cleaning tasks with shallow coordination1

Yoneda, Keisuke; Sugiyama, Ayumi; Kato, C.; Sugawara, Toshiharu

doi:10.3233/web-150326

Cited by 8 publications

(2 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The route is then divided into fragments and allocated to individual agents to minimize the working time. Yoneda et al [21] proposed a distributed method in which agents autonomously decide their search/exploration strategies in a multi-robot sweeping problem using reinforcement learning. Sampaio et al [17] proposed the gravity-based model in which the locals that were not visited for a long time have the stronger gravity, and thus, agents tend to visit such locations for uniform patrolling.…”

Section: Related Workmentioning

confidence: 99%

“…The first approach is for agents to share the working area and clean it in a coordinated manner. For example, agents could patrol the area by using different cleaning algorithms or different visitation cycles to uniformly cover the entire area [5], [13], [21]. Another strategy for this approach is for the agents to move around the area in formation (e.g., Refs.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Coordinated Area Partitioning Method by Autonomous Agents for Continuous Cooperative Tasks

Sea

Kato

Sugawara

2017

Journal of Information Processing

Self Cite

View full text Add to dashboard Cite

Abstract:We describe a method for decentralized task/area partitioning for coordination in cleaning/sweeping domains with learning to identify the easy-to-dirty areas. Ongoing advances in computer science and robotics have led to applications for covering large areas that require coordinated tasks by multiple control programs including robots. Our study aims at coordination and cooperation by multiple agents, and we discuss it using an example of the cleaning tasks to be performed by multiple agents with potentially different performances and capabilities. We then developed a method for partitioning the target area on the basis of their performances in order to improve the overall efficiency through their balanced collective efforts. Agents, i.e., software for controlling devices and robots, autonomously decide in a cooperative manner how the task/area is partitioned by taking into account the characteristics of the environment and the differences in agents' software capability and hardware performance. During this partitioning process, agents also learn the locations of obstacles and the probabilities of dirt accumulation that express what areas are easy to be dirty. Experimental evaluation showed that even if the agents use different algorithms or have the batteries with different capacities resulting in different performances, and even if the environment is not uniform such as different locations of easy-to-dirty areas and obstacles, the proposed method can adaptively partition the task/area among the agents with the learning of the probabilities of dirt accumulations. Thus, agents with the proposed method can keep the area clean effectively and evenly.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%