Train and Deploy an MLP-based, Battery Electrochemical Parameter Estimator into a Microcontroller

To-Dos (MUST be completed by 8 August 2026): 1) Train a multi-layer perceptron (MLP) model on the provided dataset (generated by PyBamm) for electrochemical parameter estimation of Li-Ion batteries. 2) Use "ONNX2C Accelerated" to generate embedded C code based on the trained MLP model. 3) Integrate the obtained C-scripts with the PPU and CPU (TriCore) of the Infineon AURIX TC4D7 Lite Kit, for real-time deployment. 4) Perform testing of the deployed MLP and verify its performance (estimation accuracy / RMSE%, execution time, memory size, etc.) in PPU and CPU, respectively. The outputs of the test should be printed in real-time in ADS-L's Terminal / Console (e.g., winIDEA). ** Save all result files and also record the comparison results (PPU vs CPU) in clear and concise format in a MS Word document. ** You must also record the training-to-deployment steps (workflow), settings, SW tools (versions), and best hyperparameters / architecture of the MLP model in the document. This document serves as a User Guide for future users who are new to the project. ** The main SW tools for this project are as follows: Python (in VS Code), ONNX2C Accelerated, AURIX Development Studio Limited (ADS-L, v 1. 10. 30-L), GCC (for CPU builds), MetaWare 2.1 (for PPU builds), winIDEA (debugger / flasher). ** Note: GCC and winIDEA are free tools, comes pre-integrated within ADS-L. Among all the tools listed above, only MetaWare 2.1 requires a license. ** Deliverables: Working Python scripts, C-scripts for PPU and CPU, Recorded Result Files (e.g., Excel / .csv), and Documentation Report (User Guide). ** Please keep the naming conventions of all submission files tidy and well-organized, with clear / concise folder structure! ** You will be provided with all the necessary files needed to begin this project.