Towards Distributed and Dynamic Task Reallocation

Distributed and dynamic task reallocation in multi-robot organization is motivated by the fact that most critical information for organizational performance must be obtained during problem solving and static techniques for task allocation cannot take the dynamic information into account. This paper presents Distributed Organizational Task Network (DOTN) as a representation for the problem, and proposes two local search techniques: trading tasks and trading responsibilities among robots, for searching the best DOTN based on global feedback. Experiments of an implemented algorithm called SOLO show a number of advantages of this approach, and illustrate some promising future research directions.