A CENTRALIZED COOPERATIVE DRIVING ALGORITHM FOR NON-SIGNALIZED INTERSECTIONS

Open Access
Author:
Xu, Ting
Area of Honors:
Mechanical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
  • Sean N Brennan, Thesis Supervisor
  • Daniel Humberto Cortes Correales, Honors Advisor
Keywords:
  • Connected and Autonomous Vehicles
  • Intersection control
  • Centralized cooperative driving
Abstract:
Connected and Autonomous Vehicles (CAVs) provide the opportunity for signal-free intersection navigation. This thesis introduces and demonstrates a centralized cooperative driving algorithm that considers two vehicles approaching a non-signalized multi-way intersection where the safe traversal can be negotiated. It is assumed that the incoming and outgoing directions are known, and individual vehicle velocities are controllable within a specified range of acceleration and for a specified range from the intersection. The proposed algorithm is developed by first considering the time-space interval of possible intersections between the vehicles. This leads to the development of a set of collision patterns that predict intersection situations that do not need to be negotiated. It is shown that these patterns extend readily from two-way intersections to eight-way intersections. In cases where path conflicts are detected within the intersection, the algorithm seeks to minimize the complexity of multi-vehicle coordination by preventing any speed deviation of the first vehicle passing through the intersection. The proposed solution in the algorithm is to redesign velocity profiles of the second vehicle arriving at the intersection, thereby avoiding any interference in the planned trajectory of the first vehicle. The algorithm is agnostic to the number of directions in/out of the intersection, and is readily generalized for ranges in acceleration limits and interaction ranges between vehicles. Based on the different cases where two vehicles’ original trajectories can cause potential collisions, simulation results show the effectiveness of the algorithm under different approaches, such as allowable velocity ranges, accelerations, and minimum algorithm starting distances.