Autonomous vehicle following : a virtual trailer link approach

• Main Author: Ng, Teck Chew
• Other Authors: Martin David Adams
• Format: Theses and Dissertations
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation
• Online Access: https://hdl.handle.net/10356/18901
• Institution: Nanyang Technological University

This thesis addresses the automation of the vehicle following function in an urban city environment, i.e., travelling under heavy traffic conditions or in a ‘stop-and-go’ motion. A virtual trailer link model for vehicle following has been proposed. With this perspective, the leader is represented as a tractor pulling the follower, which is modelled as a trailer, in the form of a virtual link. The optimum configuration and the length of the virtual trailer link model have been determined by taking into consideration the safe following distance as well as general car-like vehicle dynamics and constraints. In implementing the virtual trailer link model for vehicle following, sensors are required for the estimation of the relative pose and velocity of the lead vehicle in relation to the follower. However, inherent sensor noise, as well as limitations on their fields of view and resolution can affect the performance of the vehicle following function. A Bayesian formulation is thus proposed to model the process and sensor noise in the system. The key to a tractable solution for this formulation is based on the justified assumption that the pose of the follower vehicle is statistically independent of that of the leader. By estimating the poses of both vehicles, together with the uncertainties of the system, it is possible to minimize the path deviations between them. Moreover, as a result of uncertainties in the system, the computed driving commands based on the virtual trailer link model need to be optimized. Hence, a metric is required to evaluate and optimize the driving commands for the follower vehicle. An information theoretic framework is proposed. The aim of this framework is to select an optimal control input to the follower so as to minimize the pose error between the vehicles. Under this framework, the relative information has been used as a metric to evaluate a sequence of controlling actions, which act as inputs to the follower vehicle.