Automotive supplier explores unprecedented energy consumption prediction
The outcomes
- A robust energy prediction model, leveraging real-world car data.
- A better view on technical feasibility of creating a next-gen navigation system.
The context
A global supplier of automotive components launched a research project to explore the feasibility of a next-generation navigation system in electric vehicles (EV). The goal: predict energy consumption with unprecedented accuracy while enabling smarter route planning and charging decisions.
The logic
Current navigation systems optimize for distance or time but fail to account for real-world energy dynamics. For EV drivers, this means uncertainty—charging stops, waiting times, and inefficient routes. To deliver a competitive edge, the system must combine predictive energy modeling with intelligent routing algorithms, integrating traffic, terrain, and charging infrastructure data.
The solution
The research project combined two key improvements:
Enhancing the existing planning capabilities: Automatic route optimization combined with informed choices—available charging stations, expected waiting times, and energy requirements.
Improving the routing algorithms: computing shortest paths while predicting energy consumption per segment.
To enable this, we secured data access, consolidated datasets and set up a coding environment in the AWS cloud environment. The implementation involved modular development: data collection, model training, model comparison, and best-model analysis.
A solid energy consumption prediction model has been determined. It takes multiple inputs, among which the speed profile of the car on a road segment. This speed profile has been computed from car data. The energy prediction model can now be enriched with road features (e.g. legal speed, traffic lights, slopes, road curvature, ...).
Next steps in applied excellence
The current model predicts energy consumption for a road segment using multiple inputs, including speed profiles derived from vehicle data. The next step is predicting speed profiles from road features—legal speed, traffic lights, slopes, and curvature—making the model fully autonomous. This final model will enable precise energy forecasting for every road segment, transforming EV navigation from reactive to predictive.