Open Collections

Motion planning based on uncertain robot states in dynamic environments : a receding horizon control approach

University of British Columbia

This thesis is concerned with trajectory generation for robots in dynamic environments with relatively narrow passages. In particular, this thesis aims at developing motion planning schemes using receding horizon control (RHC) and mixed-integer linear programming (MILP). The thesis is constructed of two phases. In the first phase, a general nonlinear RHC framework is developed utilizing existing algorithms for motion planning of a robot with uncertain states. In this phase, the motion planning problem in the presence of arbitrary shaped obstacles is tackled using nonlinear system and measurement equations. This method is then adopted to solve the problem of cooperating aerial and ground vehicles. An important application of the latter is automated wildfire suppression that is investigated as a case study in this thesis. Simulation results demonstrate that the nonlinear RHC algorithm can effectively compute an optimal trajectory while handling the system uncertainties and constraints. In the second phase, motion planning schemes for robots with uncertain states are developed under more specific assumptions (i.e., linear system and measurement equations and convex polyhedral obstacles) using receding horizon MILP (RHMILP). This phase builds upon the existing RHMILP motion planning methods by introducing two additional features including a state estimation algorithm and an obstacle avoidance constraint. First, the state estimation includes a partially closed-loop strategy to estimate the future states of the robot based on the anticipated future state measurements that can in turn help to reduce the uncertainty of the future states. Second, the obstacle avoidance constraint is designed to secure a line connecting each pair of sequential robot positions inside the configuration space (outside the obstacles) at all discrete times, analogous to maintaining line-of-sight (LOS) between each pair of sequential positions of the robot. The LOS-based obstacle avoidance is advantageous in that it can typically provide the same level of safety at time steps larger than those of the conventional approaches; hence creating trajectories which are more efficient at both planning and execution stages. Numerical simulations and experiments indicate that proper inclusion of the future anticipated measurements contributes to a less conservative (more feasible) path.

Text

2014-10-21

Attribution-NonCommercial-NoDerivs 2.5 Canada

http://hdl.handle.net/2429/50806

Mechanical Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/2.5/ca/