Interaction-Aware Trajectory Prediction and Planning for Autonomous Vehicles in Forced Merge Scenarios

Merging is, in general, a challenging task for both human drivers and autonomous vehicles, especially dense traffic, because the merging vehicle typically needs to interact with other vehicles identify or create gap safely merge into. In this paper, we consider problem of control forced scenarios. We propose novel game-theoretic controller, called Leader-Follower Game Controller (LFGC), which interactions between ego priori uncertain driving intentions is modeled as partially observable leader-follower game. The LFGC estimates vehicles’ online based on observed trajectories, then predicts their future trajectories plans vehicle’s own trajectory using Model Predictive Control (MPC) simultaneously achieve probabilistically guaranteed safety objectives. To verify performance LFGC, test it simulations NGSIM data, where demonstrates high success rate 97.5% merging.