ANN-Based Hybrid Framework to Evaluate the Ride Comfort Standards in Autonomous and Conventional Vehicles

This study focuses on a novel hybrid framework for evaluating ride comfort standards in next-generation vehicles. An Artificial Neural Network (ANN) is trained to predict ride comfort matrices based on physics-based models. In both autonomous and conventional vehicles, the driver and/or occupants are subjected to road or vehicle-induced vibrations. These vibrations are then transmitted to passengers through different vehicle components, causing discomfort to the passengers. Based on existing 4 Degrees of Freedom (DOF) lumped mass models and different unique configurations of 4 DOF lumped mass models, the ANN is developed with 27 distinct models. Subsequently, the implemented ANN model was tested and validated using various experimental and mathematical models available in the literature. The ANN model predicts critical biodynamic evaluation parameters, such as Seat to Head Transmissibility (STHT), Driving Point Mechanical Impedance (DPMI), and Apparent Mass (AM), with exceptional accuracy. This physics-based ML model can be used in next-generation mobility systems to advance the ride comfort standards.