Multi-Cameras based Decision Making at Mini-Roundabouts for Autonomous Vehicles
Title: Multi-Cameras based Decision Making at Mini-Roundabouts for Autonomous Vehicles
Authors: Weichao Wang (Computer Science, Loughborough University); Quang A. Nguyen (Computer Science, Loughborough University); Paul W. H. Chung (Computer Science, Loughborough University); Qinggang Meng (Computer Science, Loughborough University);
Year: 2019
Citation: Wang, W., Nguyen, Q. A., Chung, P. H., Meng, Q., (2019). Multi-Cameras based Decision Making at Mini-Roundabouts for Autonomous Vehicles. UK-RAS19 Conference: “Embedded Intelligence: Enabling & Supporting RAS Technologies” Proceedings, 75-79. doi: 10.31256/UKRAS19.20
Abstract:
Safety driving in complicated traffic situations such as at roundabouts is crucial in autonomous vehicle design. Utilising multiple cameras for approaching vehicles detection in real-time has been determined the key challenge in this research area. This paper proposes a grid-based image processing approach that effectively learns the movement, position and direction of approaching vehicles, thus supporting an autonomous vehicle to make a human-like decision at mini- roundabouts. 230 video clips recorded in the UK were examined using three Machine Learning models (i.e. Support Vector Machines, Artificial Neural Network, k-Nearest Neighbours). Experiments indicated that SVM was outstanding with 91.32% accuracy rate in 0.7 seconds. This result suggests that the proposed system is reliable for autonomous vehicles to enter mini-roundabouts safely and smoothly.