Platoon Coordination under Signal Temporal Logic Specifications

Sammanfattning: Autonomous Driving has been hailed as one of the biggest game-changers for solving a lots of problems of the future, such as traffic congestion, efficiency of transportation, carbon emissions, road safety etc. Vehicle platooning has been a well studied idea in this field for efficient highway driving, which has been around since a few decades now. This work aims at planning for tasks that will be undertaken by an autonomous vehicle in a computationally efficient manner instead of usual methods that are calculation heavy, like MPC, and present how they can enable the vehicle to move around highway on its own or in coordination with a platoon. Using computationally simple planners leaves enough bandwidth for processors to look for unprecedented changes in a dynamic environment. The major contribution of this thesis is in demonstrating the formulation of specification for such tasks in a dynamic and heterogeneous environment using Signal Temporal Logic (STL). Control Barrier Functions (CBF) are then used to represent them as constraints for the system which are in turn fulfilled by a quadratic optimization-based controller. The STL definition allows more insight into planning for the system from a time-based perspective, while employing CBFs as constraints guarantees the system never leaves its specified set of safe states. Two different maneuvers, namely lane changing and overtaking-merging with a platoon, are performed. The merging task also demonstrates cooperation between vehicles using the same method, which shows its suitability for platoon coordination tasks. These results are presented from a computer based simulation which, along with the system and controller, also models different traffic behaviours. Particularly, the average speed of the traffic and the traffic density are modelled to verify the suitability of the controller in varying conditions and understand its limits. Also, experiments with real robots were performed in laboratory environment, which were done with the single-integrator holonomic vehicle model to assess the feasibility of the method in real world.