6533b7d6fe1ef96bd12657ab

RESEARCH PRODUCT

Traffic congestion reduction based on vehicle platoons and intelligent crossroads interactions

subject

description

Intersections are at the core of urban congestion. For more than a decade, new approaches based on autonomous and connected driving have been proposed. They aim to improve the performance of traffic control at intersections, by harnessing connectivity and driving automation (longitudinal control). These approaches have in common the fact that vehicles can negotiate together their right of way to use the conflicting space. However, they are different in terms of the way they share the space and optimization techniques. The challenge is to define the sequence of access of vehicles to the common space (which one goes first, which is the second, and so on) and the speed profile of vehicles to avoid, if possible, unnecessary stops. The literature shows that it is difficult to optimally solve both problems simultaneously in a dynamic context under strong real-time constraints.To solve the problem with respect to reality, the thesis explores the negotiation protocol as well as the policy that meets the safety requirements and respects the hard real-time constraints of the system. From the safety standpoint, vehicles access conflicting spaces by forming virtual platoons. In this way, they can maintain a sufficient safety gap to be ready to react safely in the case of danger. Regarding the real-time constraints, a rule-based system was chosen to form the sequences. In order to improve the performance of the intersection, two properties were exploited. The rules allow vehicles that follow each other (property 1) or those that can pass in parallel (property 2) to form groups. The group crosses the intersection together. A distributed right-of-way negotiation algorithm is proposed and compared to other policies of the literature. The simulation shows a significant gain in terms of intersection capacity.To further improve the performance of the proposed cooperative traffic control at intersections, the thesis focused on the longitudinal control issue. It defines an optimal output state achievable, using optimal control theory. Control based on quadratic programming shows the interest of the approach on an elementary intersection. On the one hand, the optimal output state minimizes the headway times between two conflicting vehicles. This improves the throughput of the intersection. On the other hand, it allows the modification of the sequences during the longitudinal control to improve the sequence dynamically according to the new incoming vehicles. The new approach was extended to a complex intersection. Several optimal output state-based sequence formation policies were simulated. The simulation shows that the policy based on distributed particle swarm optimization significantly improves the performance of the intersection in terms of capacity and speed. Distributed particle swarm optimization allows the formed group of platoons to be adapted to the dynamic traffic demand patterns.