Intelligent transportation systems: algorithms and formal verification

Abstract

The adoption of Connected and Automated Vehicles (CAV) on public roads along with of fering freedom to humans, offers immense possibilities for devising intelligent strategies for traffic management that can optimally utilize their precise control over movement, ability to communicate and cooperate, consistency, etc. By utilizing these qualities of CAVs en hancements in traffic management can be achieved with respect to efficiency, safety, and economy. This dissertation documents our efforts to utilize the capabilities of CAVs for their efficient management while performing collective (or individual) lane changing operations and at intersections. This dissertation presents contributions to the field of Intelligent Transportation Systems (ITS) by proposing novel algorithms for multiple ITS scenarios. First, we introduce a new ITS procedure called Lane Sorting, which refers to the rearrangement of vehicles such that every vehicle gets into its desired lane. We then propose the Cooperative Lane Sorting (CLS) algorithm, which is a cooperative algorithm generalized in terms of the number of lanes and traffic density. CLS process vehicles in independent batches called frames, and processing each frame involves solving a non-linear optimization problem reducible to a mixed-integer linear programming problem (MILP). After obtaining a solution to the MILP, vehicles adjust their position and perform the lane change operation. Being a generalized algorithm, CLS can be used for performing lane change operation of any number of vehicles wanting to do so in a given group of vehicles. Next, we propose the Heuristic Autonomous Intersection Management (HAIM) algo rithm for intersection management of autonomous vehicles. HAIM makes use of four levels of heuristics to schedule vehicles such that no two vehicles are present at the same space at the same time and the delay caused in this scheduling is minimum. The HAIM algorithm is proposed for a scenario that is equipped with a central controller along with roadside infras tructure for wireless communication. The central controller called Intersection Manager (IM) communicates with all incoming vehicles to gather information regarding their route, vehicle parameters, and motion parameters. The IM, which implements the HAIM algorithm, will then return each vehicle the velocity with which it has to travel. Vehicles, as soon as they enter the scenario, are required to transit to the given velocity and then maintain it throughout the rest of the scenario. The HAIM algorithm works on traffic that already contains vehicles on lanes corresponding to their destination direction at the intersection. Added with the fixed trajectory of vehicles at the intersection, the problem of intersection management reduces to the scheduling of vehicles at some fixed number of conflict points thus reducing the com plexity of the overall problem. A comparative study of the HAIM algorithm with two other autonomous intersection management algorithms and also with traffic light control is later performed. Results obtained show that of all considered intersection management schemes, the HAIM algorithm introduces the minimum delay in vehicle trips caused by scheduling. The next major contribution of this dissertation is to the field of formal verification of ITS algorithms. We perform verification of the design and operation of the HAIM algorithm and through this, we document an innovative approach to the verification of ITS algorithms that can be performed using available tools and techniques. For verification of HAIM, for mal modeling of the algorithm, traffic injection, vehicle behavior, and collision detection is performed and the models so obtained are scrutinized using a combination of techniques. Along with verifying the safety properties, sanity of the modeling, and verification of the expected behaviors of each of the four layers of HAIM are also performed. Techniques such as statistical model checking, simulation testing, internal verification, and artificial error in jection are used for this purpose. The approach used to verify the safety property of HAIM and sanity of modeling is unique in terms of the combination of techniques used.