Hierarchical Design of Connected Cruise Control: Perception, Planning, and Executio

The emerging wireless Vehicle-to-Vehicle (V2V) communication technologies can be exploited to monitor the motion of distant vehicles, even those beyond the line of sight. Incorporating the data provided by V2V communication into vehicle control systems has great potentials for enhancing vehicle safety, improving traffic mobility, and reducing fuel consumption. In this dissertation, Connected Cruise Control (CCC) is proposed to regulate the longitudinal motion of vehicles by incorporating motion data received from multiple vehicles ahead via V2V communication. CCC allows the incorporation of human-driven vehicles that do not broadcast information. Moreover, it needs neither a designated leader nor a prescribed connectivity topology. Such flexibility makes CCC practical for implementation in real traffic, leading to a Connected Vehicle Network (CVN) that is comprised of CCC vehicles and conventional human-driven vehicles. The design of CCC is challenging since V2V communication leads to complex connectivity topologies and may have significant information delays. Moreover, uncertainties arising from the vehicle dynamics and the varying traffic environment lead to additional complexity for CCC design.;To reduce design complexity, a hierarchical framework is utilized for systematically designing CCC that remains scalable for complex vehicle networks. This framework is comprised of three levels: perception level, planning level, and execution level. At the perception level, a causality detector is proposed to determine whether the information received from V2V communication is relevant to the CCC vehicle. Then, we design a linklength estimator to identify the number of vehicles between the broadcasting vehicle and the receiving vehicle. Based on the output of the link length estimator, we also design a network-dynamics identifier to approximate the nonlinear time-delayed dynamics of vehicle networks, which can be used to predict the motion of the vehicle immediately ahead by using the information received from distant vehicles. At the planning level, a general controller is presented to generate the desired longitudinal dynamics by incorporating information delays and connectivity topologies. We derive conditions for choosing control gains which can ensure the asymptotic stability of the equilibrium and can also attenuate perturbations from vehicles ahead. A motif-based approach is proposed for modular design of complex vehicle networks that is scalable when the number of vehicles increases. Simulation results show the advantages of V2V communication in improving traffic dynamics by attenuating disturbances. At the execution level, we consider a physics-based vehicle model that includes uncertain vehicle parameters and external disturbances such as aerodynamic drag. An adaptive sliding-mode controller is designed to regulate the engine torque, in order to make the vehicle state track the desired longitudinal dynamics.