លំហសិក្សាធិការកម្ពុជា
V1.0
លំហសិក្សាធិការកម្ពុជា
V1.0
Electric vehicles introduce routing constraints that are different from those of conventional vehicles. A charging station can be close on a map yet still be unsuitable because the remaining battery is insufficient, the connector does not match, the station is inactive, or the charger is likely to be occupied at the driver's estimated arrival time. The research problem is therefore not simply shortest-path navigation; it is a prediction and ranking problem under safety constraints. The EV Fan prototype provides a practical application context for this Machine Learning study. Its current route planner already uses battery state, vehicle range, reserve target, connector compatibility, station status, and road distance to identify safe charging stops. The Phase 3 system also exposes station provenance and explicit unsafe-route warnings. These implemented rules form a deterministic baseline. The Machine Learning subject of this paper is the next layer: learning from charging-station usage data so the system can predict which safe station is most likely to be usable when the driver arrives. The central principle of the research is that a learned model should improve station selection without replacing hard safety rules. A prediction can influence ranking, but it should not override insufficient battery reserve or connector incompatibility. This hybrid design makes the ML component easier to evaluate and reduces the risk of treating uncertain predictions as safety guarantees.
