Vehicle driving intention recognition algorithm in intersections based on Ergodic Hidden Markov Model

To address the problem of Vehicle Driving Intention (VDI) recognition in intersection scenes, with the aim of reducing complexity, improving real-time performance and universality, a VDI recognition algorithm based on Ergodic Hidden Markov Model (EHMM) is proposed for Intelligent Connected Vehicles (ICVs) to achieve VDI recognition of surrounding target vehicles. This algorithm does not constrain the target vehicles to be ICVs and is suitable for the current transitional period when intelligent connected technology is not fully popularized. In this algorithm, the structural differences of intersections are first fully considered. Therefore, a relative measurement method is proposed to extract the relative behavioral characteristics of the target vehicles to ensure good universality between intersections with different structures. Furthermore, the relative behavioral characteristics of human driver population at intersections is chosen as the reference. By analyzing the similarity between the relative behavioral characteristics of the target vehicle and the reference, VDI recognition based on EHMM is achieved. Finally, through comparative verification experiments with two types of algorithms on two different intersection scene datasets in terms of complexity, real-time performance and universality, it has been proven that the proposed EHMM based VDI recognition algorithm has fewer model parameters, faster calculation speed, and better universality when the VDI recognition accuracy is equivalent.