AI-Driven Vehicle Data Logger with Live Telemetry for BAJA SAE Vehicles

Student BAJA SAE vehicles rarely carry instrumen-tation, so slow failures such as CVT overheating and brake fadeare diagnosed from driver memory after a run. This paperpresents a low-cost data logger built for the Team RedShiftRacing India vehicle and reports how far each part has beenvalidated. Two Arduino Uno R4 sensor nodes and a receivershare a 500 kbps CAN bus carrying eight identifiers; a RaspberryPi 4 logs the stream, drives a cockpit display, runs an offlinecleaning and machine-learning pipeline and uploads data througha SIM7600G-H LTE modem to a web dashboard. On three vehiclesessions the first logger recorded 3.31, 3.32 and 1.99 Hz with atmost 0.68 % rate error, 99.31 to 100 % completeness and under0.7 MB/h storage growth, but its 1.45 to 1.8 ms jitter failed the1 ms limit; the revised logger ran at 10.0 Hz with at most 0.12 msjitter. The calculated worst-case bus load is 1.19 %. In bench teststhe two brake-pressure channels read −0.27 to −0.38 bar at restand agreed within 2.2 % over 23 applications (r = 0.9983). Of28 tests fixed in advance, 8 pass, 4 pass in part, 1 fails and 15remain open, chiefly live streaming from the moving car andaccuracy against reference instruments.