https://www.mdu.se/

Recent advances in wireless technology facilitating Vehicle-to-Vehicle (V2V) communication have paved the way towards connected and more cooperative Intelligent Transportation Systems (ITSs), enhancing road safety and sustainability. Connected and Automated Vehicles (CAVs) can exchange information with one another and their surrounding infrastructure, thereby enabling cooperative automated maneuvering such as vehicle platooning. In platooning, a group of CAVs follows the longitudinal and lateral movements of a Lead Vehicle (LV) through V2V communication and onboard sensors to form a close-knit vehicle train. Collaborating CAVs hold the potential to revolutionize transportation by enabling, e.g., enhanced safety, fuel efficiency, road efficiency, and overall mobility. However, wireless communication, a key enabling technology for collaborating CAVs, is often subject to transient outages due to irregular packet losses, which are caused by factors such as attenuation, fading and interference. An automated platoon of CAVs must remain fail-operational during such transient communication outages to stay as safe as before the outage, even if the inter-vehicle distances are short. Furthermore, a platoon may encounter road hazards, requiring emergency braking to transition into a fail-safe state. Traditionally, communication outages have been treated as permanent faults or failures, which is a worst-case scenario that has little practical value. More recent attempts to model wireless communication as either on or off as a function of time are still too pessimistic and may lead to safety distances which are unnecessarily long. To this end, this thesis proposes to characterize the nature of transient wireless communication outages into finer granularities so that, e.g., the vehicles in a platoon can adapt to the currently available information, prioritize safety, and still remain as efficient as possible. Such characterization also enables formulating a state machine in which the states represent various fail-operational, emergency braking, and fail-safe states as a function of the instantaneous communication quality. An approach involving changing states by switching controllers and adjusting inter-vehicle gaps is developed to keep a platoon fail-operational during runtime. In addition, several emergency braking strategies are proposed to minimize the LV's stopping distance while avoiding inter-vehicle collisions when transitioning the platoon into a fail-safe state. The thesis also employs Machine Learning (ML) for real-time prediction of communication outages and collision risks during emergency braking, enabling proactive collision avoidance measures. A simulation tool named PlatoonSAFE, using realistic traffic mobility models, has been developed to evaluate the proposed algorithms. Rigorous simulation studies demonstrate that the characterization of communication outages into finer granularities enables fault tolerance and provides a more balanced trade-off between fuel efficiency, string stability, and LV tracking compared to the traditional way of modeling communication outages. Finally, the ML tool and the outage model enable fuel-efficient platooning at high speed while still being able to respond to road hazards by enabling fast emergency braking.