This paper presents a control architecture for the motion planning problem of vertical takeoff and landing unmanned autonomous vehicles with obstacle avoidance and consideration of additional constraints on the state of the vehicle. It uses the Google Tango Tablet for indoor navigation and as provider of a dense point cloud representing a huge number of obstacles in the vicinity. The motion planning algorithm is based on a constrained optimal control problem with receding horizon that is approximately solved by using only a finite set of control input trajectories. The design of the necessary functions of the motion planner allows for a very efficient, parallel implementation on the graphics card of the tablet which permits a huge number of obstacles to be considered without loosing existing information about the environment. This is made possible by using the large number of computing cores in the graphics card. Runtime tests show the efficiency of the architecture and that real-time execution is possible.
展开▼
