TraEDITS: Diversity and Irregularity-Aware Traffic Trajectory Editing

Abstract

We present TraEDITS, a novel traffic trajectory editing framework for autonomous vehicle testing, which can generate new traffic behaviors by controlling each vehicle interactively to increase the diversity or irregularity of traffic testing data. Given a traffic flow with its original trajectories, user’s edits, environmental constraints, and physical constraints as input, our framework is able to generate an edited traffic flow through a global path planning module and a data-driven microscopic traffic flow simulation module. With the way-points from the user, our global planning module generates lane-level navigation by heuristic-based path planning in discrete space. By taking internal properties of the vehicle, velocity continuity, reference path, and collision avoidance into consideration, our simulation module generates vehicles’ motions based on energy optimization driven by real traffic data. Given edits of the desired speed, the lateral deviation, or the reference path, our approach can generate a new trajectory of the selected vehicle, and adjust the surrounding vehicles’ trajectories accordingly. As a result, our framework is able to simulate irregular or even rare events existing in real traffic. Meanwhile, our framework can enhance the diversity and irregularity of traffic behaviors and interactions by creating challenging scenarios like swerve, nudge, and U-turn. We validate the usability and plausibility of our framework through extensive experiments and user studies.