Agent-Task Assignation Based on Target Characteristics for a Swarm of Specialized Agents

摘要

A systemic agent-task specialized assignation approach for collaborative unmanned systems is introduced in this paper. It is formulated by matching specialized robotic agents with the recognition of specific characteristics on targets in an environment. The recognition stage leverages visual features detection on objects present in the surrounding of the robotic agents via embedded sensors. A given target's zone of influence is generally defined as a circular area surrounding any detected target object, with the latter corresponding to a specific task to be dealt with by a specialized robot. Within the zone of influence around a target a switching process for the leadership of the swarm is initiated. Appropriateness of assignation, smoothness and safety of the transition form the major factors of performance considered. The framework is formulated to deal with static and moving targets and their corresponding zone of influence, which leads to the consideration of overlapping task zones. The proposed system involves robots control, leader switching and agent-task assignation processes in relation with the detected target characteristics. Simulation experiments are conducted to validate the proposed system, and demonstrate that an effective coordination of the specialized agents around the corresponding targets is achieved.