Project Summary
This project will research and develop a Human-like Pedestrian Avoidance System. The proposed pedestrian-aware automated vehicle decision making and control system will enable the automated vehicles to avoid pedestrians like human driven vehicles, making the automated vehicles’ performance more human like, natural and acceptable from both on-board drivers and the other road users.
Project Achievements
Activities 1. Data collection: We developed a Driving-Pedestrian Distributed Simulator based on Unreal Engine 4.26. The driver and pedestrian interaction data were collected in the virtual urban road environment. 2. Development of a pedestrian model: To predict the future behaviour of the pedestrian based on our car behaviour, we developed a pedestrian model based on a probabilistic graphical model. 3. Optimisation of automated vehicle controller: With the developed pedestrian model, we optimised our automated vehicle control system.
Conclusions
1. Conclusions: The developed pedestrian model and vehicle controller can predict pedestrian motion, quantify the uncertainty in their future trajectories and improve automated vehicle’s motion behaviour when interacting human road users. 2. Potential impact: The developed automated vehicle control algorithm has the potential to improve human road users’ acceptance and significantly contribute to road safety by improving motion behaviour of the automated vehicle when interacting with road users.
Next Steps
1. Improve the human-vehicle interaction test system 2. Multi-agents interaction test, i.e. multi-vehicle and multi-pedestrians 3. Engage with some industrial partners to explore the potential deployment of our control algorithm 4. Many technological challenges are existing for the interaction modelling, interactive decision-making and control as well as human factor analysis, we will continue to address the existing challenges.
