Championship in the Argoverse 2 Multi-agent Motion Forecasting Challenge

QCNet, a modelling framework that pushes the boundaries of trajectory prediction for autonomous vehicles to enhance safety and reliability, wins the championship in the Argoverse 2 Multi-agent Motion Forecasting Challenge at the Workshop on Autonomous Driving of the IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2023. Among 100+ submissions from top teams in the previous Argoverse forecasting challenges, QCNet achieved the best performance on all metrics by a significant margin. Besides the Challenge, QCNet also ranked first on both Argoverse 1 and Argoverse 2 single-agent motion forecasting benchmarks. This breakthrough AI technology was conducted by Prof Jianping Wang and Mr Zikang Zhou, a CS PhD student, together with collaborators from Hon Hai Research Institute and Carnegie Mellon University. The paper, entitled “Query Centric Trajectory Prediction,” was published in the proceedings of the CVPR held in June 2023 in Vancouver, Canada and was featured in the Conference’s Workshop on Autonomous Driving.