Randomised MPC-based motion-planning for mobile robot obstacle avoidance

Authors:
Alex Brooks;Tobias Kaupp;Alexei Makarenko
Affiliations:
Australian Centre for Field Robotics, University of Sydney, NSW, Australia;Australian Centre for Field Robotics, University of Sydney, NSW, Australia;Australian Centre for Field Robotics, University of Sydney, NSW, Australia
Venue:
ICRA'09 Proceedings of the 2009 IEEE international conference on Robotics and Automation
Year:
2009

Citing 4
Cited 2

Kinodynamic motion planning

Journal of the ACM (JACM)
Robot Motion Planning

Robot Motion Planning
Toward Reliable Off Road Autonomous Vehicles Operating in Challenging Environments

International Journal of Robotics Research
Planning Algorithms

Planning Algorithms

The problem of boundary following by a unicycle-like robot with rigidly mounted sensors

Robotics and Autonomous Systems
A Method of Boundary Following by a Wheeled Mobile Robot Based on Sampled Range Information

Journal of Intelligent and Robotic Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents an algorithm for real-time sensor-based motion planning under kinodynamic constraints, in unknown environments. The objective of the trajectory-generation algorithm is to optimise a cost function out to a limited time horizon. The space of control trajectories is searched by expanding a tree using randomised sampling, in a manner similar to an RRT. The algorithm is improved by seeding the tree using the best control trajectory from the previous iteration, and by pruning branches based on a bound to the cost function and the best trajectory found so far. Performance of the algorithm is analysed in simulation. In addition, the algorithm has been implemented on two kinds of vehicles: the Segway RMP and a four-wheel-drive. The algorithm has been used to drive autonomously for a combined total on the order of hundreds of hours.