Real-time safe trajectory generation for quadrotor flight in cluttered environments

We address the problem of real-time generation of smooth and collision-free trajectories for autonomous flight of a quadrotor through 3-D cluttered environments. Our approach starts by generating a sequence of variable sized 3-D free space grids as the virtual flight corridor using an OctoMap-based environment representation and search-based planning method. The key contribution is a quadratic programming-based formulation for generating multi-segment polynomial trajectories that are entirely fit within the corridor, and thus collision-free. Our formulation also allows incorporating higher-order dynamical constraints to ensure that the trajectory is feasible for the platform. A novel non-iterative constraint relaxation method is also proposed for implicit optimization of segment duration. The proposed algorithm runs real-time onboard our quadrotor experimental testbed. Both simulation and online experimental results are presented for performance verification.