"A Real-World Dataset for SOC and SOE Estimation of Electric Cycle Lithium-Ion Batteries on Indian National Highways"

"Lightweight electric vehicles (EVs) with lithium-ion batteries (LIBs) are also becoming increasingly useful technologies for sustainable urban mobility in densely populated and polluted nations, including India, where short-distance travel dominates daily commuting. The pedal-assisted electric cycles are some of them, which cause low energy consumption, traffic congestion, and zero tailpipe emissions, thus promoting minimal physical activity that is healthy to the EV user. Howbeit, the extremely dynamic nature of operating conditions of e-cycles due to the behaviour of riders, their weight, interaction with traffic, and the state of the roads makes it difficult to predict the battery State of Charge (SOC) and State of Energy (SOE) under real-driving conditions. To address this, an E-Motorad pedal-assisted electric cycle was instrumented with a compact Bluetooth-enabled SOC and SOE monitoring kit for real-time field data acquisition. The SOC and SOE monitoring system integrates an isolated Hall-effect power measurement front end, an ESP32 microcontroller with IEEE 802.15.1-compliant wireless communication, ambient temperature and humidity sensors, and a high-voltage DC-DC converter supplied directly from the traction battery. During real driving trips, the system continuously records time-stamped battery pack voltage, current, power, temperature, humidity, and estimated SOC and SOE, which are transmitted through Bluetooth to a smartphone for visualization and data logging. The obtained dataset includes high-resolution measurements of electrical and environmental variables, providing a stable foundation for developing machine learning\/deep learning models that can be implemented on low-cost microcontrollers. A real-world dataset was gathered using the developed platform in 30 real driving trips performed on the Indian National Highways between September 2025 and December 2025, and covered rainy and winter environmental conditions. The experimental design specifically specifies rider-specific variability, which is paramount to lightweight EVs. Riders are in the weight range of 55 kg to 105 kg, representing low, medium, and high-weight riders and calm and aggressive riding behaviours. The differences explain the variation in acceleration, pedalling effort, and throttle use, which have a direct impact on energy consumption and the range that could be achieved during pedal-assisted e-cycles. The chosen e-cycle platform has a nominal riding range of about 30 km, which makes it quite convenient to use for short to medium distances in congested urban and peri-urban areas. The resulting dataset is realistic and depicts actual road conditions, traffic dynamics, rider interventions, and environmental influences. It is then utilized to develop and validate machine learning- and deep learning-based SOC and SOE estimation models suitable for low-cost microcontroller implementation in EV applications."