A Unified Approach for Decision Making and Trajectory Planning for Automated Driving using Time-Varying Potential Fields in Intersections

Automated Driving (AD) has become one of the most active research areas in control engineering, owing to its potential to enhance road safety, improve traffic efficiency, and reduce environmental impact. Within this framework, the development of SAE Level 4 Autonomous Vehicles (AVs) poses the fundamental challenge of designing Motion Planning (MP) and Control strategies capable of generating safe, comfortable, and efficient trajectories in complex traffic environments. This thesis extends a previously developed control architecture for AD, originally conceived for standard maneuvers such as lane keeping and overtaking, to more complex and interactive contexts, namely unsignalized intersections, where mutual interactions and dependencies among vehicles play a critical role. The baseline framework adopts a hierarchical architecture integrating a Global Path Planner (GPP), a Local Trajectory Planner (LTP), and a Nonlinear Model Predictive Control (NMPC) layer. The GPP computes a global reference path to reach the destination based on GPS data, while the LTP refines it locally by formulating and solving a Non-linear Programming (NLP) problem whose cost function is based on Artificial Potential Fields (APFs) to model the surrounding road environment. In particular, static APFs represent fixed elements such as lanes and road boundaries, whereas Time-Varying APFs (TV-APFs) describe dynamic obstacles, adapting their shape and amplitude to the evolving traffic scenario. This formulation enables the LTP to compute smooth and safe trajectories that account for environmental changes. The NMPC is then employed to track the LTP trajectory while regulating both the lateral and longitudinal dynamics of the vehicle, as required by higher levels of driving automation such as SAE Level 4 and Level 5. The main contribution of this thesis is the introduction of an additional TV-APF specifically designed to handle unsignalized intersections without relying on discrete decision-making or rule-based logic. The proposed field dynamically adapts its amplitude and influence according to the predicted occupancy of the intersection area, acting as a dynamic safety barrier that influences the ego-vehicle’s behaviour. This mechanism causes the ego-vehicle to decelerate or stop when the intersection is occupied or unsafe and to proceed smoothly when the path is clear. To prevent unnecessary reactions, an actor selection mechanism ensures that only relevant vehicles affect the field, specifically those entitled to right of way or whose predicted paths intersect that of the ego-vehicle. As a result, the system reproduces human-like caution, allowing the ego-vehicle to adapt its motion to other traffic actors. The proposed architecture has been validated through simulations of multi-vehicle unsignalized intersections, designed using the Driving Scenario Designer tool of MATLAB’s Automated Driving Toolbox. The results demonstrate the framework’s ability to generate collision-free trajectories and to reproduce safe, efficient, and natural driving behaviour in complex traffic conditions.