Richard West

Zhuoqun Cheng, Richard West, Craig Einstein

Timing guarantees are crucial to cyber-physical applications that must bound the end-to-end delay between sensing, processing and actuation. For example, in a flight controller for a multirotor drone, the data from a gyro or inertial sensor must be gathered and processed to determine the attitude of the aircraft. Sensor data fusion is followed by control decisions that adjust the flight of a drone by altering motor speeds. If the processing pipeline between sensor input and actuation is not bounded, the drone will lose control and possibly fail to maintain flight. Motivated by the implementation of a multithreaded drone flight controller on the Quest RTOS, we develop a composable pipe model based on the system's task, scheduling and communication abstractions. This pipe model is used to analyze two semantics of end-to-end time: reaction time and freshness time. We also argue that end-to-end timing properties should be factored in at the early stage of application design. Thus, we provide a mathematical framework to derive feasible task periods that satisfy both a given set of end-to-end timing constraints and the schedulability requirement. We demonstrate the applicability of our design approach by using it to port the Cleanflight flight controller firmware to Quest on the Intel Aero board. Experiments show that Cleanflight ported to Quest is able to achieve end-to-end latencies within the predicted time bounds derived by analysis.

William Koch, Renato Mancuso, Richard West et al.

Autopilot systems are typically composed of an "inner loop" providing stability and control, while an "outer loop" is responsible for mission-level objectives, e.g. way-point navigation. Autopilot systems for UAVs are predominately implemented using Proportional, Integral Derivative (PID) control systems, which have demonstrated exceptional performance in stable environments. However more sophisticated control is required to operate in unpredictable, and harsh environments. Intelligent flight control systems is an active area of research addressing limitations of PID control most recently through the use of reinforcement learning (RL) which has had success in other applications such as robotics. However previous work has focused primarily on using RL at the mission-level controller. In this work, we investigate the performance and accuracy of the inner control loop providing attitude control when using intelligent flight control systems trained with the state-of-the-art RL algorithms, Deep Deterministic Gradient Policy (DDGP), Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO). To investigate these unknowns we first developed an open-source high-fidelity simulation environment to train a flight controller attitude control of a quadrotor through RL. We then use our environment to compare their performance to that of a PID controller to identify if using RL is appropriate in high-precision, time-critical flight control.