Naslov
Real-time planning for cooperative maze exploration

Autori
Zovko, Mila ; Crnković, Bojan ; Ivić, Stefan

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
2nd IMA Conference On Mathematics Of Robotics Abstract Book / - , 2021, 17-17

Skup
2nd IMA Conference On Mathematics Of Robotics

Mjesto i datum
Manchester, Ujedinjeno Kraljevstvo, 08.09.2021. - 10.09.2021

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
maze, exploration, coopperative, multi-agent system

Sažetak
As robotics evolves, more and more work is being done to have the robot replace man in dangerous working conditions. Examples of such dangerous working conditions are searching for victims in burning buildings, passing through a minefield, or patrolling the streets with an increased crime rate. All of the above problems can be viewed as a problem in which one or more agents are exploring an unknown maze. Many single-agent maze exploration algorithms have been implemented, but extending them to multi-agent systems is not always simple and feasible. We propose a solution for the problem where a cooperative multi-agent system of automated vehicles is used to explore an unknown maze with stationary targets. The algorithm uses a potential field to explore the maze with a built-in cooperative behavior of agents which includes collision avoidance, coverage coordination, and optimal path planning. The proposed algorithm was created on the basis of HEDAC (Heat Equation Driven Area Coverage) , and which has already been successfully applied for motion control for autonomous heterogeneous multi-agent area search in uncertain conditions, and for motion control for multi-agent non-uniform spraying. We made a comparison with the algorithm of similar objectives and applications. The proposed algorithm results show obvious improvement and we can demonstrate that the new algorithm can be applied to all maze types and not only to special cases. In performed tests, the proposed algorithm proved to be robust, adaptive, scalable, and computationally inexpensive which enables real-time planning.

Izvorni jezik
Engleski

Znanstvena područja
Matematika, Interdisciplinarne tehničke znanosti

POVEZANOST RADA