A Federated Learning Framework for Collaborative Ageing and State of Health Prediction in E-Rickshaws

TOTO E-rickshaws, widely used in Kolkata, West Bengal, provide an affordable and eco-friendly transportation option for low-income families, relying on lead-acid batteries that degrade rapidly and incur high replacement costs. This paper proposes a federated learning framework integrated with Internet of Things (IoT) technology to enable collaborative aging analysis and state of health (SOH) prediction for these batteries. Utilizing low-cost IoT devices and owners’ smartphones, the system ensures real-time monitoring while preserving data privacy by sharing only model updates, avoiding transmission of raw battery data. A novel Temporal Attention Lightweight Network (TALNet) is introduced to enhance SOH prediction accuracy by capturing temporal degradation trends. Simulations demonstrate that TALNet achieves a mean absolute percentage error (MAPE) of 0.8% and extends battery lifespan by 20%, reducing annual replacement costs by approximately 4,000 rupee per vehicle. The approach enhances the economic sustainability of E-rickshaw operations, addressing challenges such as computational constraints and connectivity issues for low-income users.