An official website of the United States government
A framework for autonomous waypoint planning, trajectory generation through waypoints, and trajectory tracking for multi-rotor unmanned aerial vehicles (UAVs) is proposed in this work. Safe and effective operations of these UAVs is a problem that demands obstacle avoidance strategies and advanced trajectory planning and control schemes for stability and energy efficiency. To address this problem, a two-level optimization strategy is used for trajectory generation, then the trajectory is tracked in a stable manner. The framework given here consists of the following components: (a) a deep reinforcement learning (DRL)-based algorithm for optimal waypoint planning while minimizing control energy and avoiding obstacles in a given environment; (b) an optimal, smooth trajectory generation algorithm through waypoints, that minimizes a combinaton of velocity, acceleration, jerk and snap; and (c) a stable tracking control law that determines a control thrust force for an UAV to track the generated trajectory.
more »
« less
Trajectory Generation on SE(3) for an Underactuated Vehicle with Pointing Direction Constraints
This paper addresses the problem of generating a position trajectory with pointing direction constraints at given waypoints for underactuated unmanned vehicles. The problem is initially posed on the configuration space ℝ 3 × ℝ 2 and thereafter, upon suitable modifications, is re-posed as a problem on the Lie group SE(3). This is done by determining a vector orthogonal to the pointing direction and using it as the vehicle's thrust direction. This translates to converting reduced attitude constraints to full attitude constraints at the waypoints. For the position trajectory, in addition to position constraints, this modification adds acceleration constraints at the waypoints. For real-time implementation with low computational expenses, a linear-quadratic regulator (LQR) approach is adopted to determine the position trajectory with smoothness upto the fourth time derivative of position (snap). For the attitude trajectory, the thrust direction extracted from the position trajectory is used to first propagate the attitude to the subsequent waypoint and then correct it over time to achieve the desired attitude at this waypoint. Finally, numerical simulation results are presented to validate the trajectory generation scheme.
more »
« less
- Award ID(s):
- 1739748
- PAR ID:
- 10195616
- Date Published:
- Journal Name:
- 2019 American Control Conference (ACC)
- Page Range / eLocation ID:
- 1930 to 1935
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
This paper addresses the problem of generating a continuous and differentiable trajectory on the Lie group of rigid body motions, SE(3), for a class of underactuated vehicles modeled as rigid bodies. The three rotational degrees of freedom (DOF) are independently actuated, while only one translational DOF is actuated by a body-fixed thrust vector. This model is applicable to a large set of unmanned vehicles, including fixed-wing and rotorcraft unmanned aerial vehicles (UAVs). The formulation utilizes exponential coordinates to express the underactuation constraint as an intrinsic part of the problem. It provides steps to generate a rest-to-rest trajectory after obtaining conditions that guarantee controllability. An attitude trajectory is selected to satisfy the given initial and final attitude state. The position trajectory generation is subsequently posed as an optimal control problem expressed as a linear quadratic regulator (LQR) in the exponential coordinates corresponding to position. As a result, an optimal position trajectory is obtained which ensures that the trajectory generated is feasible with realistic velocities and with given initial pose and final pose, while satisfying the underactuation constraint. Numerical simulation results are obtained that validate this trajectory generation scheme.more » « less
-
Robot arms should be able to learn new tasks. One framework here is reinforcement learning, where the robot is given a reward function that encodes the task, and the robot autonomously learns actions to maximize its reward. Existing approaches to reinforcement learning often frame this problem as a Markov decision process, and learn a policy (or a hierarchy of policies) to complete the task. These policies reason over hundreds of fine-grained actions that the robot arm needs to take: e.g., moving slightly to the right or rotating the end-effector a few degrees. But the manipulation tasks that we want robots to perform can often be broken down into a small number of high-level motions: e.g., reaching an object or turning a handle. In this paper we therefore propose a waypoint-based approach for model-free reinforcement learning. Instead of learning a low-level policy, the robot now learns a trajectory of waypoints, and then interpolates between those waypoints using existing controllers. Our key novelty is framing this waypoint-based setting as a sequence of multi-armed bandits: each bandit problem corresponds to one waypoint along the robot’s motion. We theoretically show that an ideal solution to this reformulation has lower regret bounds than standard frameworks. We also introduce an approximate posterior sampling solution that builds the robot’s motion one waypoint at a time. Results across benchmark simulations and two real-world experiments suggest that this proposed approach learns new tasks more quickly than state-of-the-art baselines. See our website here: https://collab.me.vt.edu/rl-waypoints/more » « less
-
Some locomotor tasks involve steering at high speeds through multiple waypoints within cluttered environments. Although in principle actors could treat each individual waypoint in isolation, skillful performance would seem to require them to adapt their trajectory to the most immediate waypoint in anticipation of subsequent waypoints. To date, there have been few studies of such behavior, and the evidence that does exist is inconclusive about whether steering is affected by multiple future waypoints. The present study was designed to address the need for a clearer understanding of how humans adapt their steering movements in anticipation of future goals. Subjects performed a simulated drone flying task in a forest-like virtual environment that was presented on a monitor while their eye movements were tracked. They were instructed to steer through a series of gates while the distance at which gates first became visible (i.e., lookahead distance) was manipulated between trials. When gates became visible at least 1-1/2 segments in advance, subjects successfully flew through a high percentage of gates, rarely collided with obstacles, and maintained a consistent speed. They also approached the most immediate gate in a way that depended on the angular position of the subsequent gate. However, when the lookahead distance was less than 1-1/2 segments, subjects followed longer paths and flew at slower, more variable speeds. The findings demonstrate that the control of steering through multiple waypoints does indeed depend on information from beyond the most immediate waypoint. Discussion focuses on the possible control strategies for steering through multiple waypoints.more » « less
-
Abstract This paper presents a geometric adaptive position tracking control system for a quadrotor unmanned aerial vehicle. In particular, the attitude control system is designed on the product of the two-dimensional unit sphere and the one-dimensional circle such that the direction of the thrust that is critical for position tracking is controlled independently from the yawing direction that is irrelevant to the position dynamics. Compared against the prior work with coupled attitude controls on the special orthogonal group, the proposed controller prevents large yaw errors from causing an undesirable performance degradation in tracking a position command. Further, the control input is augmented with adaptive control terms to mitigate the effects of disturbances, and it is formulated globally on the spheres to avoid singularities and complexities of local coordinates. The efficacy of the proposed control system is illustrated by both numerical examples and indoor/outdoor flight experiments.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.23919/ACC.2019.8815238
