Simultaneous planning and execution for autonomous quadrotors
Zhou Liu · 2022
The application of quadrotors in inspection, agriculture, transportation, and other fields has increased recently. In these applications, the quadrotor should autonomously generate safe trajectories and execute planned trajectories accurately in the presence of uncertainties. Therefore, this thesis aims to develop an enabling technology to improve tracking accuracy, agility, and safety of autonomous quadrotors encountering disturbances. In this thesis, a Simultaneous Planning and Execution (SPAE) system strategy is proposed to achieve this objective. First, a control scheme based on the SPAE strategy is proposed to improve tracking accuracy for the quadrotor. Due to limited control inputs, a constraint model is induced to reduce three independent variables, which allows the proposed strategy to handle the underactuation of the quadrotor system. The proposed control law consists of an online calculation of desired trajectory and disturbance estimation to compensate for the state errors. Among this, the calculated trajectory is an explicit function of time that is derived by the highest order derivative of the polynomial function constrained by current states, desired states, and the given finite time interval. Meanwhile, an observer based on acceleration and angular acceleration is developed to estimate the unknown disturbance. The simulations and outdoor experiments demonstrate that the proposed SPAE strategy can accurately carry out the task of point and trajectory tracking within a given finite time in the presence of disturbance. Then, the proposed SPAE strategy is extended for the quadrotor to achieve agile flight, including flip maneuvers and high-speed tracking. In the flipping maneuver, feasible trajectories of flipping angle and thrust are planned to achieve the minimum height loss. Meanwhile, a nonlinear filter with outlier detection and online adjustment of sonar variance is developed to mitigate the accuracy degradation of sonar such that the vertical position and velocity can be obtained. Finally, a circle tracking task and a flipping task were carried out on the F450 quadrotor to verify the effectiveness of the proposed planning and execution strategy. The tracking error results of both attitude and position show that the proposed strategy can handle disturbances and accurately track the planned trajectory in both the position and attitude tasks. In addition, this thesis addresses another case of agile flight, an aggressive flight through a narrow gap. This task requires the quadrotor to maintain alignment with a 6-degree of freedom (DOF) pose of the gap as it transverses the gap. There are six geometric constraints to be satisfied at the gap plane, but the quadrotor only has four rotors as inputs. Therefore, a constraint model under SPAE is proposed to project the narrow gap rotation angles and disturbance into acceleration constraint. After determining boundary constraints, a minimum time trajectory solver based on a piece-wise cubic function is proposed to realize aggressive flight of the quadrotor. The proposed solver can generate an optimal time trajectory satisfying the nonzero boundary condition at the gap and coupled dynamics feasibility. Meanwhile, the computation time of this solver on the random test is the level of microseconds on the embedded system, so the proposed solver is suitable for quadrotors with limited onboard computational capability. Simulation and outdoor experiment results demonstrate the effectiveness of the proposed SPAE strategy and trajectory solver for a quadrotor flying narrow gap. The safety problem of the quadrotor suffering one rotor loss and disturbance also is investigated in the proposed SPAE system. First, Euler angle solutions at equilibrium state are analyzed to guarantee the feasibility of control inputs with only three rotors. Then an emergency control strategy is proposed to allocate priorities of control variables according to the location of mass center offset and fault rotor. Meanwhile, this control strategy sacrifices the yaw angle control to handle further underactuation for the damaged system with three rotors. For validation, the proposed system strategy is implemented on the quadrotor to realize take-off and keep-attitude tasks with three rotors. In addition to rotor losses, unsafe flight caused by actuator saturation is considered in the SPAE strategy. Specifically, the dynamic capability of the quadrotor is real-time evaluated by estimating the disturbance information online and combining the maximum thrust. The time interval of the trajectory is then adjusted online based on the updated capability. As a result, the saturation of the actuator caused by the underestimation of disturbance can be avoided under a gentle regenerated trajectory. The effectiveness of the proposed time adjustment mechanism is verified experimentally with a 200g load disturbance. We conclude this thesis by prospecting future research works.