Autonomous UAV Swarm: Behavior Generation and Simulation

We present a software framework to produce and simulate autonomous behaviors of UAV swarm tasked with search and reconnaissance missions. With proper control and motion planning, multiple-vehicle swarm allows for a more efficient and cost-effective search. The main challenge, however, is on the generation of the desired autonomous behaviors of the UAV swarm, with minimum centralized control and human intervention. The proposed framework allows UAV agents to adopt different behavior models and use a rule-based architecture to select suitable behavior that is best for given situations. In particular, the UAV's search and tracking behavior are treated by two different POMDP decision models to reflect different reward mechanisms. The proposed framework is evaluated in a SITL environment which boasts both realistic low-level flight dynamic and sensor inputs. A two-UAV teamed search and tracking scenario is presented to test the performance of the proposed model against an evasive ground target.