AEE Project Work course 2018 Masters Programme in

AEE Project Work course 2018 Master’s Programme in Automation and Electrical Engineering L. Bondén (project manager), V. Pirsto, N. Luostarinen, S. Sirniö, M. Baranauskas V. Mukherjee (instructor) Objective Smart-bike A compact traction kit with smart user features has been built which is able to convert any regular bike to an artificial intelligent assisted electric bike. The bike kit is also compliant with Finnish legislation. Main modules 1. Converter • Three phase inverter commutating the battery DC into AC fed to the motor. • DC-DC converter for powering the Arduinos. 2. Motor controller • BLDC controller implemented on Arduino uno. • Reads motor current and position to control the converter with PWM signals. 3. Assist reference controller • Arduino uno reads sensor data such as speed, pedalling cadence and acceleration. • Predicts assist amount and forwards the reference to the motor controller. Figure 1. Main modules of the kit. 4. Motor and mechanics • BLDC-motor utilizing friction to drive the wheel. • Motor disengages from the wheel when not in use. Machine learning • Aims to learn the rider’s cycling behaviour and automate the motor usage optimally for the user. • Sensor data is exported to the app which computes a neural network model using Google’s AI Tensorflow. • Model is exported back to the bike, which uses it to predict the assist amount. Figure 3. Operation principle of the machine learning app. Figure 2. Block diagram of main modules and their interaction. Figure 4. Converter layout eea. aalto. fi Electromechanics research group