Abstract:

Microscopic traffic simulation plays a crucial role in transportation engineering by providing insights into individual vehicle behavior and overall traffic flow. However, creating a realistic simulator that accurately replicates human driving behaviors in various traffic conditions presents significant challenges. Traditional simulators relying on heuristic models often fail to deliver accurate simulations due to the complexity of real-world traffic environments. Due to the covariate shift issue, existing imitation learning-based simulators often fail to generate stable long-term simulations. In this paper, we propose a novel approach called learner-aware supervised imitation learning to address the covariate shift problem in multi-agent imitation learning. By leveraging a variational autoencoder simultaneously modeling the expert and learner state distribution, our approach augments expert states such that the augmented state is aware of learner state distribution. Our method, applied to urban traffic simulation, demonstrates significant improvements over existing state-of-the-art baselines in both short-term microscopic and long-term macroscopic realism when evaluated on the real-world dataset pNEUMA.

Model Overview:

We utilize a VAE to model both the expert and learner's context-conditioned past trajectory distribution. During training, we augment each expert data by generating a learner-aware augmented expert state through the VAE's reconstruction of its past trajectory. Using this augmented state along with the original future trajectory, we train the learner's policy network through supervised learning. During simulation, we roll out the policy network with several post-processing steps including sampling, projecting the trajectories onto the road, and smoothing the projected trajectories.

Videos from Our Simulator

There are some collision, off-road and traffic light violation cases because the dataset does not provide accurate vehicle shape, heading, high-definition map, and traffic light information.

Changing Road Shape

Our microscopic long-term traffic simulators can help transportation engineers and planners to analyze and predict the impact of microscopic adjustments on traffic patterns without disrupting real-world traffic. For example, it can help analyze how changing road shape affects traffic patterns. We present the mean road density and speed changes in our simulator after modifying several roads' shapes. We can see that a local microscopic modification in road network can causes traffic congestion or alleviation in distant areas.