Comprehensive Control for Autonomous Vehicles: From Vehicle- to Human-Centric or From Integrated to Holistic?

The anticipated rise in the adoption of autonomous vehicles (AVs) has highlighted the need to redesign vehicle control architectures. This involves compromises guaranteeing performance and safety while addressing passenger comfort, which becomes increasingly critical as passengers shift from active drivers to passive occupants, making them vulnerable to motion sickness (MS). Despite the significant number of research studies examining MS mechanisms, quantification methods, and mitigation strategies, current control approaches seldom incorporate passenger states into the vehicle control loop. Nevertheless, a review of the literature reveals that existing definitions of a “holistic controller” are either vague or fragmented. Acknowledging the necessity to harmonize diverse methodologies and delineate a consolidated definition of a holistic framework, this review initiates with a thorough exposition of vehicle-related state estimation and control methodologies, accentuating the proposed literature solutions on holistic approaches. A critical distinction emerges between traditional integrated control, relying on separate, loosely-coupled modules with limited inter-module communication and vehicle-centric optimization, and the proposed holistic control featuring a multi-level architecture with bidirectional information flow, adaptive parameter weighting, and simultaneous consideration of vehicle and passenger subsystems to achieve multi-objective optimization encompassing safety, comfort, and overall efficiency. Additionally, a comprehensive discourse on the necessary additional module regarding the passenger state estimation techniques is presented, with a particular emphasis on those targeting head motion, which is closely associated with the onset of MS, with the following discussion focused on the sensing strategies employed in relation to the underlying vehicle estimation frameworks. In light of the aforementioned insights, the paper proposes the concept of a holistic controller, defined as a multi-level structure that collectively considers heterogeneous subsystems to achieve multi-objective optimization by managing trade-offs between competing objectives. Finally, the requirements and feasibility of such a framework in real-world applications are discussed, outlining how current research is evolving, defining the incoming demand for modular frameworks, high-performance computing, and shared solutions.