Sensor Network based Conflict Resolution in Autonomous Multiagent Systems

The design of intelligent and sensor-based autonomous agents learning by themselves to perform complex realworld tasks is a still-open challenge for artificial and computational intelligence. In this paper a concept of a framework for an autonomous robotic agent is presented. The structure of an intelligent robotic agent consists of two independent subsystems: the action and motion planning system and the action and motion reactive control system with integrated conflict resolution methods. The action planning system uses an aggregated world model storing knowledge about all static and dynamic objects in the surrounding environment. The action controller solves space conflicts in a reactive manner making use of information from local sensors and a distributed sensor network. Each dynamic object registered from sensor network field is inserted into the world model as a new obstacle in 2,5D form. Based on the updated world model a conflictfree robot motion is calculated in a one step motion planning cycle.