Reactive navigation of a mobile robot using a hierarchical set of learning agents

Within the context of learning sequences of basic tasks to build a complex behavior, a method is proposed which uses a hierarchical set of incrementally learning agents. Each one has to respect a particular perceptive constraint. To do so, an agent must choose either to execute basic tasks or to call another agent in order to use its decision-making competency, according to its perception. The learning procedure of each agent is achieved by a reinforcement learning inspired algorithm based on a heuristic which does not need internal parameters. A validation of the method is given, using a simulated Khepera robot. A hierarchical set of 4 agents is created. Each one is dedicated to the exploitation of particular perceptive data. They use 5 basic tasks in order to achieve a goal-reaching behavior which is formulated by a high level strategy composed of logical rules using perceptive primitives.