Battery Range Prediction using Federated Learning on Edge

Learn to develop a battery range prediction system for electric vehicles using federated learning techniques in this 14-minute conference talk. Explore how federated learning addresses the critical need for periodic model updates in battery range prediction by accommodating changes in battery parameters over time and variations in driving dynamics. Discover the dual advantages of federated learning: aggregating insights from fleet-wide data patterns to create sophisticated models trained on diverse scenarios, while maintaining user privacy by avoiding raw data transmission to central repositories. Follow the implementation of a range prediction solution using simulated vehicle data and the Flower FL framework, designed for easy deployment on embedded Texas Instruments edge platforms. Understand the architecture where edge components operate as quality managed (QM) components while central model aggregation runs as containerized applications on-premises or in the cloud, with communication established through gRPC protocols.