Surrogate Modeling of Optimal Control Based Collision Avoidance System for Multirotor Unmanned Aerial Vehicles

collision avoidance, polynomial chaos expansion,
multi-rotor vehicle, successive convexification

A dynamically changing operating environment, along with constraints imposed through applicable safety requirements, pose significant challenges to autonomous multi-rotor manned and unmanned aerial vehicle operations in urban areas. Safety-critical onboard collision avoidance capability requires fast decision making accounting for uncertainties arising in complex environments. Successive convexification approach is applied to generate collision avoidance trajectories assuming both static and moving obstacles. The uncertainties arising in estimated state of moving obstacles are accounted for by construction of Polynomial Chaos Expansion based surrogate model. The obtained surrogate model can be evaluated in real-time to update the collision avoidance trajectory in case of change of tracked obstacle's state. The designed trajectories are subsequently tracked using a closed-loop Model Predictive Control scheme assuming a quadcopter configuration.