Research into multi-agent swarming leverages multiple low-cost agents to perform tasks resulting in a more resilience, scalable, and efficient systems compared to systems using single agents. To suppo..
Research into multi-agent swarming leverages multiple low-cost agents to perform tasks resulting in a more resilience, scalable, and efficient systems compared to systems using single agents. To support research on multi-agent swarming, there is a need for a low-cost system that reliably localizes multiple agents moving on a plane in three degrees of freedom (x, y, yaw). Existing approaches utilize an overhead camera with fiducial markers to perform localization but suffer performance degradation when agents undergo maneuvers featuring fast rotations and translations. The induced motion blur causes intermittent detection loss. This work proposes a localization system that maintains agent localization histories and predicts poses during intermittent detection loss. The predicted search region constrains the search area for redetection, which should enable recovery of tracks using modified detection algorithms. In a bounded region, less exclusive search parameters will be utilized to detect lost agents, with decreased prevalence for false positives than would be seen than if applied to a larger region of the arena. The intended outcome is an improved localization robustness for swarming research, with an emphasis on maintaining identity and pose fidelity during fast maneuvers.
