Flight range and time, as well as autonomous flights in complex environments, are two challenges preventing quadcopters from being more widely used in the industry. This thesis reduces the effect of these two problems via several motion planning methods.
This thesis is composed of three parts. In the first part, two methods are proposed to improve the flight time (endurance) and distance (range) of quadcopters. The first method does so by finding the optimal flight speed using extremum seeking control. The second method extends the first method by finding the optimal flight sideslip as well as speed, and adds a step size adapter to the extremum seeking controller to improve its convergence speed. Both methods do not require the power consumption modeling of the quadcopters and can thus adapt to changing payloads and disturbances.
In the next part, we focus on the problem of fast collision avoidance flight in cluttered environments. First, we propose a computationally efficient memoryless planner for fast outdoor flights, using a depth camera for sensing obstacles and using the visual inertial odometry (VIO) for state estimation. Since the VIO may function poorly in areas with very few visual features or when the flight is overly aggressive. We then take the state estimation quality of the VIO into account during the trajectory planning. This perception-aware planner is able to guide the vehicle to areas with more VIO features and avoid overly aggressive trajectories. It improves the VIO's accuracy and reduces its failure rate, while only slightly reducing the flight speed.
Finally, an inertial navigation motion planning strategy is introduced. This state estimation method only requires the inertial measurement unit (IMU) and can be used as a backup when other methods fail. By breaking a long trajectory into multiple short "hopping trajectories'' and introducing zero velocity updates when the vehicle is stationary on the ground, the state estimation variance can be drastically reduced and can be used in closed-loop control of quadcopters.
