Balancing Safety and Efficiency for Connected and Automated Vehicles: A Shared Control Approach

The emerging technologies of communication and vehicle automation empower connected and automated vehicles (CAVs) to execute intelligent trajectory planning, which can enhance safety, energy efficiency, and overall mobility. In particular, safety has been the top priority in autonomous driving (AD) industries, as it determines if an AD product can be trusted, deployed, and scaled up. Moreover, the safety performance of CAVs guarantees the welfare of human society. However, recent studies point out that a fundamental tradeoff exists between safety and other performance metrics (e.g., traffic and energy efficiency, comfort). This indicates that the over-conservative behavior of safety-oriented CAVs will inevitably jeopardize normal traffic operations, increasing traffic congestion and energy consumption. For instance, a CAV might (i) reserve large inter-vehicle spacing (reducing highway throughput); (ii) brake frequently (disturb traffic and introduce traffic congestion); and (iii) not dare to make lane changes near on/off-ramps or turnings at intersections (stuck in wrong routes and triggering travel delay). Thus, the significance of safety should not deteriorate other performance metrics inappropriately. It is imperative to modulate safety and other performance tasks to intelligently decide which control objective should be emphasized during CAV operations.

To ensure a smooth integration of different control objectives while preserving optimality and safety, this study proposes a shared control-based method. The proposed method consists of three components: (i) a model predictive controller (MPC) emphasizing fuel economy and comfort in lane-change and car-following operations (labeled MPC-ECOCOM); (ii) a safety-assurance MPC for collision avoidance; and (iii) a dynamic weighting allocation (DWA) mechanism to synthesize the control actions from two MPCs by smoothly adjusting the weightings of safety, energy efficiency, and comfort during operations. The DWA mechanism uses potential functions to monitor driving risk and other performance metrics. If the potential function of driving risk is much greater than other performance metrics, the DWA will produce a higher shared control weighting for the control action computed from the safety-assurance MPC. By contrast, during normal operations, the shared control weighting of MPC-ECOCOM will be larger, and the synthesized control action will be dominated by fuel-efficiency and comfort-oriented control action. With interpretable parameters, the proposed method is also flexible in adjusting the bias between two control modes, which can promote personalizable AD for different road users (e.g., medium/heavy-duty trucks, passenger vehicles).

The proposed method is first validated using a hardware-in-loop (HIL) experiment, which uses a passenger vehicle mounted on a wheel dynamometer to measure actual wheel speed and torque, delivering a realistic evaluation on the performance of the proposed method. The results indicate that the proposed method can ensure desired tracking, comfort and energy efficiency in normal car-following scenes while preserving collision-free responses under emergency harsh-braking and lane-change events. Next, a traffic simulation performed in SUMO compares the proposed method to other baseline control methods, quantifying the improvements in traffic throughput, stability, energy efficiency, and safety.