What If Vehicles are Controlled by the Road?

Recently, self-driving vehicles have been actively studied and have begun to operate on actual roads. Most of the current self-driving vehicles rely on a number of sensors to individually judge the situation and take action. However, from a road perspective, a collection of individual vehicles' decisions is inevitably sub-optimal, and it is essential that roads directly intervene (control/assist) in vehicle operation for optimal traffic management. This paper proposes Software-Defined Traffic Management (SDTM) that incorporates the concept of Software-Defined Networking (SDN) into the transportation environment. In SDTM, each vehicle is handled like a packet in SDN and its movement can be assisted/controlled using a global view of the road. To this end, SDTM enables reliable interaction between vehicles and infrastructure (RSU/controller) by 1) allocating collision-free channels to vehicles, 2) reducing vehicle's messaging interval by fully utilizing available channel resources, and 3) redistributing sensitive information from the infrastructure. Based on the traffic information collected and reported by RSU, the controller induces vehicle to form clusters and controls/assists the movement of vehicles on a cluster-by-cluster basis, implementing safe and efficient traffic flows. According to the performance evaluation using the ns-3 simulator, under SDTM, vehicles experience an average travel time reduced by 30% (more than 97% of all vehicles experience shorter travel time) compared to the real traffic traces, while maintaining a safety distance.