Autonomous Driving in Highway Scenarios through Artificial Potential Fields and Model Predictive Control

An approach for automated driving in highway scenarios in the context of a two levels hierarchical architecture is proposed. In particular, we define suitable artificial potential functions (APF) combinations that can effectively handle the most relevant maneuvers of highway driving, such as speed and distance tracking, lane keeping, overtaking and returning. Parameters of the APF functions are dynamically tuned according to the acquired scenario. The defined APF are included in the cost function of a Model Predictive Control (MPC) control problem to generate the path trajectory. A behavioral logic described by a finite state machine (FSM), based on sensor acquired data and suitable dynamic conditions is defined to select the most appropriate maneuver to realize. Extensive simulation tests are introduced to show the effectiveness of the proposed approach.