We introduce a multi-model variant of the EMT-based control algorithm. The new algorithm, MM-EMT, is capable of balancing several control tasks expressed using separate dynamic models with a common action space. Such multiple models are common in both single-agent environments, when the agent has multiple tasks to achieve, and in team activities, when agent actions affect both the local agent's task as well as the overall team's coordination. To demonstrate the behaviour that MM-EMT engenders, several experimental setups were devised. Simulation results support the effectiveness of the approach, which in the multiagent scenario is expressed in the MM-EMT algorithm's ability to balance local and team-coordinated motion requirements.
Zinovi Rabinovich, Nir Pochter, Jeffrey S. Rosensc