Interaction-Aware Motion Planning Framework Integrating Internal Driver Models with External Predictions for Lane Merging Automation

This paper analyzes the performance of decentralized motion planning algorithms for autonomous vehicles, when complementing them with future movement predictions of other vehicles. When multiple predictions are available (e.g., future states as predicted by behavior models (BM) used as prediction model within the controller or external predictions (EP) from prediction algorithms), the challenge is to take advantage of all predictions and find an optimal trajectory and control strategy. To this end, this paper presents a model predictive control (MPC)-framework that combines EP with BM used as prediction model within the MPC and analyzes the effect of combining multiple predictions. The framework is applied to and analyzed for a two-vehicle, multilane lane merging (LM) scenario. The proposed framework is compared with motion planners that use only EP or BM. It is shown that the framework inherits reactive performance properties from the EP and can better anticipate future motions of other road users, which is lacking in the MPC with only BM (e.g., the target vehicle cannot be predicted to do lane changes, but EP can). The integrated framework is also able to use the BM target vehicle proactive behavior, which is lacking when only using EP. Therefore, the integrated framework can handle a wider variety of scenarios and improve traffic behavior.