In this work we present a planning and control method for a quadrotor in an autonomous drone race. Our method combines the advantages of both model-based optimal control and model-free deep reinforcement learning. We consider a single drone racing on a track marked by a series of gates, through which it must maneuver in minimum time. Firstly we solve the discretized Hamilton-Jacobi-Bellman (HJB) equation to produce a closed-loop policy for a simplified, reduced order model of the drone. Next, we train a deep network policy in a supervised fashion to mimic the HJB policy. Finally, we further train this network using policy gradient reinforcement learning on the full drone dynamics model with a low-level feedback controller in the loop. This gives a deep network policy for controlling the drone to pass through a single gate. In a race course, this policy is applied successively to each new oncoming gate to guide the drone through the course. The resulting policy completes a highfidelity AirSim drone race with 12 gates in 34.89s (on average), outracing a model-based HJB policy by 33.20s, a supervised learning policy by 1.24s, and a trajectory planning policy by 12.99s, while a model-free RL policy was never able to complete the race.
A largely untapped potential for aerial robots is to capture airborne targets in flight. We present an approach in which a simple dynamic model of a quadrotor/target interaction leads to the design of a gripper and associated velocity sufficiency region with a high probability of capture. A model of the interaction dynamics maps the gripper force sufficiency region to an envelope of relative velocities for which capture should be possible without exceeding the capabilities of the quadrotor controller. The approach motivates a gripper design that emphasizes compliance and is passively triggered for a fast response. The resulting gripper is lightweight (23 g) and closes within 12 ms. With this gripper, we demonstrate in-flight experiments that a 550 g drone can capture an 85 g target at various relative velocities between 1 m/s and 2.7 m/s.
Control models of three NASA Urban Air Mobility (UAM) reference vehicles (the quadrotor, octocopter, and Lift+Cruise (LPC)) were created and compared to determine the effect of rotor number and disk loading on control margin and design. The heave and yaw axes demand more actuator usage than the roll and pitch axes. Between heave and yaw, heave was the more demanding of the two because of the dependence of heave on the engine speed controller (ESC). When the feedback gains for all three vehicles were optimized to Level 1 handling qualities (HQs) specifications using CONDUIT, the ESC for the octocopter was the most stable and had the highest rise time (time for the rotor to respond to an input), while the LPC ESC was the least stable and had the smallest rise time. Rise time corresponds to the time required for rotor response. When actuator usage was translated to current margin, torque margin, and power margin, heave was the most demanding axis, followed by yaw, roll, and then pitch for all three vehicles. The results emphasize the importance of an accurate motor model within the control system architecture.
Autonomous drone racing is a challenging research problem at the intersection of computer vision, planning, state estimation, and control. We introduce AirSim Drone Racing Lab, a simulation framework for enabling fast prototyping of algorithms for autonomy and enabling machine learning research in this domain, with the goal of reducing the time, money, and risks associated with field robotics. Our framework enables generation of racing tracks in multiple photo-realistic environments, orchestration of drone races, comes with a suite of gate assets, allows for multiple sensor modalities (monocular, depth, neuromorphic events, optical flow), different camera models, and benchmarking of planning, control, computer vision, and learning-based algorithms. We used our framework to host a simulation based drone racing competition at NeurIPS 2019. The competition binaries are available at our github repository 1 .
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.