- Supervised Learning – labelled data
- Use: Traffic sign & pedestrian recognition.
- Example: Traffic Sign Recognition (TSR) systems in many cars (e.g., Nissan Traffic Sign Recognition, BMW, Mercedes, Tesla), trained on labelled sign images. Wikipedia+1
- Unsupervised Learning – no labels
- Use: Driver behaviour clustering, anomaly detection in telematics or in-vehicle networks.
- Example: Unsupervised models for anomalous driving pattern detection in telematics / CAN data (used by insurers and OEM analytics backends rather than as a visible in-car feature). arXiv+1
- Semi-Supervised Learning – mix of labelled & unlabelled road data
- Status: Common in research & internal development for perception and anomaly detection where labelling is expensive; not usually marketed in brochures, so specific car-level deployments are not publicly named.
- Reinforcement Learning – trial-and-error optimisation
- Use: Energy management, HVAC optimisation, eco-driving strategies.
- Status: Widely studied for EV energy optimisation and autonomous driving, but OEMs rarely say “we use RL” in public docs; most examples are research prototypes and pilots, not clearly tied to one production model.
- Regression / Prediction – RUL & predictive maintenance
- Example: AI-enhanced Battery Management Systems in EVs that estimate state of health and remaining useful life to optimise range and schedule service (e.g., BMS solutions highlighted for major EV makers like Panasonic / CATL in current EVs). evmagazine.com+1
- Clustering – grouping drivers / trips
- Use: Driving-style segmentation, risk scoring, UX recommendation segments.
- Example: ML-based driving style clustering using telematics data (research used by insurers and fleet operators, feeding into usage-based insurance and risk analytics). ScienceDirect+1