Machine learning for e-mobility in the AWS Cloud
ServiceXpert supports a long-standing client in the area of applied machine learning or artificial intelligence.
The task is to create and optimise several regression and artificial neural network based predictor models to achieve or exceed required accuracy levels. The models developed are intended to predict the required energy consumption over a given driving distance of electric vehicles. The necessary vehicle data for training and testing purposes was gathered from several electric vehicles over a defined period of time.
ServiceXpert’s tasks included the following:
- Data cleanup, data preparation, data pre-processing, data distribution validation, sorting, separation of data into trips, segments and driving modes
- Visualisation of data in 2D maps
- Feature engineering of new parameters to improve the accuracy of the model, e.g. power, state of charge, etc.
- Training / validation of several standard machine learning regression algorithms including Support Vector Machines and other project-specific tools
- Training / validation / optimisation of several multilayer neural network architectures
- Hyperparameter tuning using techniques such as Grid Search and RandomCV Search
- Setting up the repositories in Amazon Web Services (AWS) code commit
- Setting up the machine learning infrastructure in AWS Sagemaker
- Optimisation of the AWS Sagemaker infrastructure to reduce costs
- Documenting the test results of each model at each iteration.
- Deployment of the chosen models in the form of a web app (Python Flask / Plotly)
ServiceXpert engineers bring several years of extensive know-how in the evaluation of vehicle data and the implementation of these data analyses in the further development and training of artificial intelligence instances.
ServiceXpert, a company of the ESG Mobility Group, employs over 85 staff in its Hamburg and Munich locations. ServiceXpert is a Europe-wide operating system and software house with a focused service portfolio for Lifecycle management of EE information for leading manufacturers of commercial vehicles and their supplier industry.