Data-Driven Model for State of Charge Estimation of Lithium-Ion Battery Pack in Autonomous Mobile Robots

Autonomous mobile robots (AMRs) require efficient battery management to enhance operational performance and extend battery life. Lithium-ion battery (LIB), known for their high energy density and stability, are the primary power source for AMRs. However, the complexity of sensors, motor drivers, and control algorithms necessitates precise state of charge (SOC) estimation within the battery management system (BMS) to optimize efficiency and reduce charging frequency. This study proposes a data-driven SOC estimation method using a single-cell equivalent model combined with a 1D-CNN, enabling fast and accurate SOC estimation while simplifying BMS hardware design. The 1D-CNN model is trained on the University of Wisconsin-Madison's 25° C dynamic stress test dataset, with self-collected battery pack data used for testing. The proposed approach enhances battery management and offers a scalable solution for high-performance AMR applications.