Skip to Content

PhD Opportunities

Interaction-aware Intelligent Vehicle Decision-making and Control

School of Mechanical and Aerospace Engineering | PHD
Funding
Funded
Reference Number
MAE2021/03
Application Deadline
None specified
Start Date
None specified

Overview

Automated Vehicles (AVs) and other type mobile robots are facing challenges at interactive environments, particularly human-involved environments. Decision-making and control algorithm designs of the AV need to consider both human’s response to it and the uncertainty of human motions. To this end, a human-centric interaction-aware AV decision-making and control algorithm need to be developed.

Automated driving is moving a step closer to reality, with increasing levels of vehicle automation technologies being involved in AVs. However, there are still some technical barriers that affect the deployment of automated vehicles. One important problem is that the behaviour of the automated vehicles is not human-like or overcautious when interacting other road users, particularly when interacting with pedestrians. At some interactive urban environment with multi-pedestrians, automated vehicles’ decision-making, motion planning and control algorithm design is more challenging. The reasons for this can be summarized as: 1. It is not clear about how the human-driven vehicles interact with the other road users and how the other road users interact with the vehicles (both traditional and automated). 2. The developed algorithms including decision-making, motion planning and control are not accurate and effective. To investigate the vehicle-pedestrian avoidance, some researchers and engineers tried to use the dummy experiment to mimic the behaviour of the pedestrians, but the motion of the dummy is very difficult to control, which affect the accuracy and reliability of data collection. On the other hand, virtual reality techniques provide a flexible test environment for driving data collection and automated system verification. At multiple vulnerable road users environment, the design of the decision-making algorithm need to consider the uncertainty and the interaction among the road users, including the ego automated vehicle. This study aims to develop a model-based reinforcement learning algorithm to mimic the human driver’s behaviour when interacting with vulnerable road users.

Funding Information

Project Summary
Supervisor

Dr Chongfeng Wei

More Information

c.wei@qub.ac.uk

Research Profile


Mode of Study

Full-time: 3 years


Apply now Register your interest