Parameterized-Agents Collaboratively Tracking (PACT)

The need for mostly-autonomous robotic systems to perform robust tracking and identification of dismounts is especially acute in urban asymmetric scenarios where the possibility of losing track of a potentially armed dismount makes the risks to an organic asset too large. The constraints of line-of-sight sensors coupled with urban canyons and tree canopies makes loss of track inevitable, meaning that more than one tracking agent, cooperating in a tightly-coupled fashion, must be used to get acceptable tracking performance. Using an unmanned aerial vehicle (UAV) and unmanned ground vehicle (UGV) allows the composite system to observe the dismount using very distinctive viewing geometries, which are robust to different types of occlusion. We propose a system called Parameterized-Agents Collaboratively Tracking (PACT), which incorporates a UAV with a UGV in a cooperative tightly-coupled system, enabling improved tracking performance by combining the capabilities of the two platforms. Leveraging our expertise in image stabilization, target detection, particle filter tracking, and sensor fusion, PACT will combine observations from multiple vehicles into a global scene description and choose appropriate actions for vehicles in the system to maximize track continuity and focus agents on targets of particular interest.