HPC for motion planning
Title: HPC for motion planning
SNIC Project: SNIC 2020/13-67
Project Type: SNIC Small Compute
Principal Investigator: Kristoffer Bergman <kristoffer.bergman@liu.se>
Affiliation: Linköpings universitet
Duration: 2020-08-17 – 2021-09-01
Classification: 20202
Keywords:

Abstract

Motion planning is defined as the problem finding a feasible motion plan that brings a system from its current state to a desired goal state, without colliding with surrounding obstacles. In our previous work, we have developed a motion planner that uses so called motion primitives to solve the motion planning problem. These motion primitives are computed by solving a number of optimal control problems offline to reduce the workload online. In our new work, we want to be able to compute a function approximator for each motion primitive. The function approximator is intended to guide the selection of motion primitives when re-planning is performed online. To compute these function approximations, the amount of optimal control problems that needs to be solved offline increases drastically. However, all problems that needs to be solved are independent, which opens up the possibility for parallel computing. Hence, the idea is to use SNIC to solve these optimal control problems offline.