IMTP: Intention-Matching Trajectory Prediction for Autonomous Vehicles

Wenzhi Bai, Luwen Yu, Andrew Weightman, Zhengtao Ding, Shengquan Xie, Zhenhong Li

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Trajectory prediction for surrounding vehicles is critical for ensuring the safety of autonomous driving. In this paper, we introduce a novel prediction framework named Intention-Matching Trajectory Prediction (IMTP). Different from existing results that predict trajectories based on only environmental information and historical trajectories, the proposed method initially identifies the possible intentions of surrounding vehicles based on the environment and generates intention-informed trajectories based on the physical vehicle model. Historical trajectories are then used to identify the intention and trajectory with the highest probability. The proposed framework effectively integrates the physical vehicle model, road-related environmental factors, and interactions among surrounding vehicles. A comparative study conducted on a public dataset demonstrates that our framework enhances both prediction accuracy and robustness.

Original languageEnglish
Title of host publication2023 29th International Conference on Mechatronics and Machine Vision in Practice, M2VIP 2023
PublisherIEEE
ISBN (Electronic)9798350325621
DOIs
Publication statusPublished - 2 Feb 2024
Event29th International Conference on Mechatronics and Machine Vision in Practice, M2VIP 2023 - Queenstown, New Zealand
Duration: 21 Nov 202324 Nov 2023

Publication series

Name2023 29th International Conference on Mechatronics and Machine Vision in Practice, M2VIP 2023

Conference

Conference29th International Conference on Mechatronics and Machine Vision in Practice, M2VIP 2023
Country/TerritoryNew Zealand
CityQueenstown
Period21/11/2324/11/23

Keywords

  • autonomous vehicles
  • trajectory prediction

Fingerprint

Dive into the research topics of 'IMTP: Intention-Matching Trajectory Prediction for Autonomous Vehicles'. Together they form a unique fingerprint.

Cite this