Interaction-aware Prediction of Urban Traffic Scenarios
Andreas Philipp, Daniel Göhring – 2022
This work presents a novel rule-based interaction-aware multi-modal prediction method for urban traffic scenarios. The method takes into account the most common classes of traffic participants and handles all relevant types of motion behaviors. The potential trajectories of the traffic participants are rolled out resulting in multi-modal probability distributions for the states of all agents for each prediction time step. The analysis of collision risks between these trajectories is the basis for the interaction-awareness. The prediction is fully interaction-aware by considering also the interactions between the obstacles. The system is able to predict complex urban scenarios with numerous different agents in real-time. The approach is evaluated using real-world scenarios and in a simulated environments.