A Team Formation with Unknown Size in Multi-agent Environment
Amar Nath, Rajdeep Niyogi · 2016
Usually, in the Multi-Agent System (MAS), a number of agents required to finish a job are known a priori. But, there are many situations in which it is not known prior that how many agents will exactly be required to finish the job, i.e., the rescue operation. The Multi-Agent Planning problems (MAP) contains collective (several agents are performing different actions at the same time such that the combined effect achieves a common goal) and joint actions (a kind of collective action where several agents are performing the same action at the same time such that the combined effect achieves a common goal) makes the planning a bit challenging. The execution of such actions demand team formation prior to their execution. The plan execution, involving team formation of the size that is not known a priori is very hard and challenging. The key contributions of the paper include: (1) identification of the domain where the number of agents required to finish a job is not known a prior, (2) proposing a distributed consensus algorithm based on asynchronous communication (messages passing) to form a team and (3) The design specification and validation of the proposed algorithm with Specification and Descriptive Language (SDL) and Message Sequence Charts (MSCs).