Abstract Electric Vehicle (EV) platforms require tight integration of energy management and perception to operate safely and efficiently on low-cost hardware. This work presents the prototype for an embedded, Artificial Intelligence (AI) driven framework that co-designs a microcontroller-based battery management system with a lightweight Convolutional Neural Network (CNN) for object detection and closed-loop actuation. The Battery Management System (BMS) performs real-time voltage, current, and temperature acquisition with differential sensing and implements a Recursive Least Squares (RLS) – Open Circuit Voltage (OCV) hybrid estimator for state-of-charge (SoC), cell-balancing control, and thermal safeguards. The perception module executes on an embedded processor and feeds an actuation layer that enables adaptive braking and speed control. Bench and on-vehicle tests demonstrate the SoC estimation error ≤ 2% under dynamic drive profiles, effective thermal regulation with cell-to-cell voltage dispersion constrained to ≤ 0.05 V during balancing, and object-detection accuracy of ~ 95% with typical end-to-end inference latency of 40–60 ms. The integrated system reduces computational and cost overheads while maintaining robustness across varying lighting and target distances. By improving battery safety, energy utilization, and perception-to-actuation timing on affordable hardware, the framework advances practical, sustainable EV prototyping aligned with Sustainable Development Goals (SDG) 7, 9, 11, and 13.
Keshyagol et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: