Data-Driven Motion Control of Autonomous Vehicles in GPS-Unreliable Environments

PI: Chuan Hu (UAF),, ORCID: 0000-0001-5379-1561

Co PIs: none

AMOUNT & MATCH: $50,000 from PacTrans; $50,000 Match

PERFORMANCE PERIOD: 3/16/2021 – 3/15/2022

STATUS: Active

CATEGORIES: Autonomous Vehicles



FINAL PROJECT REPORT: will be available once completed

PROJECT DATA: will be available once completed

DESCRIPTION: In this project, a novel data-driven strategy will be proposed for AV motion control when a GPS signal is not reliable. In recent years, data-driven approaches such as reinforcement learning (RL) and adaptive dynamic programming (ADP) algorithms have been widely adopted in solving dynamic programming problems. However, there is seldom any related application in AV control systems when a GPS signal is not reliable, where technical difficulties occur due to the unavailability of the vehicle location, orientation and certain critical vehicle states. An AV’s complex operation environment, external disturbances, system nonlinearities, modeling and non-structural uncertainties also lead to challenges for reliable motion control.

To this end, this project will develop an enhanced ADP approach for AV motion control when the GPS signal is not reliable, based on the estimation results for the sideslip angle and tire-road friction coefficient. The dependable inputs will be signals collected/measured from on-board sensing results. The innovations of this research are: 1) An adaptive and cost-efficient estimation scheme will be proposed to estimate the tire road friction coefficient and sideslip angle simultaneously based on on-board sensors; 2) A novel learning-based adaptive motion control strategy will be proposed based on the sensing results and obtained estimation results to solve the tracking control with guaranteed prescribed performance.

Research Project Progress Report #1 10/10/2021
No Cost Extension Request 1/15/2022
Draft Report 1/15/2022
Final Project Report 3/15/2022