A Coalitional Autonomous Guiding Model Considering Traffic and Non-traffic Participants

In the field of autonomous driving, there is a lot of work focused on how to more accurately predict the trajectory of traffic participants and make reasonable and safe decisions based on the behavior of the surrounding agents. However, non-traffic participants are also a very important part of the traffic interaction network, the traffic lights have a non-negligible impact on the prediction module and the decision-making module of autonomous driving. In this paper, the focus of attention is on the impact of traffic lights on the performance of autonomous driving. We propose a new decision-making model GTL that include a trajectory prediction module, it can effectively utilize the non-traffic elements. The multi-objective trajectory prediction model HEAT_E improved the prediction accuracy by 43.07% and 23.79% on the SinD and CitySim datasets compare with baseline, respectively. GTL shared the latent representation of trajectory predict context between HEAT_E and decision-making module. GTL outperforms the baseline model across the board.