Abstract
This paper introduces a kinodynamic motion planning algorithm for Unmanned Aircraft Systems (UAS), called MP-RRT#. MP-RRT# joins the potentialities of RRT# with a strategy based on Model Predictive Control to efficiently solve motion planning problems under differential constraints. Similar to other RRT-based algorithms, MP-RRT# explores the map constructing an asymptotically optimal graph. In each iteration the graph is extended with a new vertex in the reference state of the UAS. Then, a forward simulation is performed using a Model Predictive Control strategy to evaluate the motion between two adjacent vertices, and a trajectory in the state space is computed. As a result, the MP-RRT# algorithm eventually generates a feasible trajectory for the UAS satisfying dynamic constraints. Simulation results obtained with a simulated drone controlled with the PX4 autopilot corroborate the validity of the MP-RRT# approach.
Highlights
The use of Unmanned Aircraft Systems (UAS) has increased progressively across a wide range of applications such as remote sensing, search and rescue, security and surveillance, precision agriculture, infrastructure inspection and urban planning, to name a few [35]
This paper presents a kinodynamic motion planning algorithm called MP-Rapidly-exploring Random Tree (RRT)#, where RRT# [2] is enhanced by the use of Model Predictive Control (MPC) [5] to compute the cost between vertices evaluated in the rewiring of the new sample
The proposed strategy is implemented in C++ using the Robot Operating System (ROS) [34] framework and the Open Motion Planning Library (OMPL) [38], which provides many state-of-the-art sampling-based algorithms and many additional functionalities to facilitate the development of new algorithms
Summary
The motion planning problem is ubiquitous in most of autonomous robotics applications, beside UAS. The motion planning problem is subdivided into two subproblems: path planning and path tracking. The authors in [14] adopted a potential field approach for planning, followed by a multi-constrained Model Predictive Control (MPC) strategy for tracking. The authors in [7] proposed a two-stage approach where path planning is computed by leveraging the Rapidly-exploring Random Tree (RRT) algorithm, associated with a Linear Quadratic Regulator (LQR) controller for the tracking of the resulting reference trajectory. None of the mentioned two-stage approaches guarantee the dynamic feasibility of the computed path. Another classical approach for path planning breaks the problem into two phases: a continuous
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have