Objective:

To present a methodology for motion planning in autonomous systems with multiple agents.

Methodology:

The physical behavior of a team of autonomous navigation systems is parameterized and defined. Then, a control policies algorithm is described and implemented, which interprets these descriptions, which are converted into LTL formulas, and a model is generated which allows making automatic abstractions. From generic solution configurations, the case of multiple robots with a single task in an environment with fixed obstacles is derived. The methodology is validated in different scenarios, and the results are analyzed.

Results:

The proposed methodology for motion planning in autonomous systems with multiple agents combines two state-of-the-art techniques, thus allowing to mitigate the combinatorial explosion of states in traditional approaches.

Conclusions:

The proposed methodology solves the automaton synthesis problem for high-level control with task changes during the execution. Under certain criteria, the problem of combinatorial explosion of states associated with these systems is mitigated. The solution is optimal with regard to the number of transactions performed by the team members.

Funding:

Universidad Tecnológica de Pereira