New AI framework lets self-driving cars handle intersections without traffic lights
A unified decision and trajectory planner using time-varying potential fields avoids collisions at unsignalized intersections.
Deep Dive
A novel framework integrates decision-making and trajectory planning for automated vehicles at unsignalized intersections. Using a Finite Horizon Optimal Control Problem with Time-Varying Artificial Potential Fields and a conflict-zone occupancy coefficient, it generates safe, feasible reference trajectories. Simulations in multi-vehicle traffic scenarios demonstrate effectiveness.
Key Points
- Unifies decision-making and trajectory planning in a single Finite Horizon Optimal Control Problem (FHOCP).
- Uses Time-Varying Artificial Potential Fields (TV-APF) with short-horizon motion prediction for dynamic obstacle avoidance.
- Introduces a conflict-zone occupancy coefficient to explicitly model collision risks at unsignalized intersections.
Why It Matters
This approach could eliminate the need for traffic lights by enabling safe, autonomous navigation at complex intersections.