An Algorithm for Multi-robot Planning: SGInfiniteVI

Abstract

In this paper we introduce an efficient path planning algorithm called SGInfiniteVI so the resulting path solution facilitates for robots the accomplishment of their tasks. The algorithm is based on the dynamic programming and deals with the problem of distributed equilibrium computation and selection in general-sum stochastic games with infinite horizon. Our approach proposes three Nash selection functions (in case of multiplicity of Nash equilibria) and obtains its benefits by pointing out experimentally which function is the most suitable to favor better interaction between robots. Experiments were conducted on the classical box-pushing problem, with several mobile robots moving objects in dynamic environments.