Is end-to-end intelligent driving definitely better than mapless driving? What problems of mapless driving can it solve?
Xiaomi EV is one of the first automotive brands in China to apply end-to-end technology in parking scenarios.
In terms of end-to-end capabilities, Xiaomi is about to achieve end-to-end coverage across all driving and parking scenarios, offering a higher experience ceiling with more coherent and consistent performance.
The accumulation of mapless NOA is the foundation of end-to-end driving. End-to-end introduces user driving data for training. Building on the mapless foundation, it offers more human-like trajectories, more efficient navigation, and safer driving. With the addition of the VLM (Vision-Language Model), it can handle more complex scenarios.
More human-like driving: The trajectory during driving is smoother and more natural, with a significant reduction in sudden acceleration and hard braking. It can intelligently anticipate and plan the optimal driving path, making more reasonable driving decisions in advance.
More efficient navigation: Ready to use as soon as you get in the car from the parking space, with stronger negotiation capabilities. It flexibly handles oncoming traffic, pedestrians, and static obstacles, and can navigate through ETC and barrier checkpoints, achieving a seamless journey from parking space to parking space.
Safer driving: The end-to-end model's free multi-dimensional evaluation mechanism conducts real-time safety assessments of trajectories, engraving the safety baseline into its DNA. It reacts faster in complex road conditions where two-wheelers and pedestrians are mixed.
VLM handles more complex scenarios: It analyzes road conditions in real time, identifying and alerting the driver to potholed roads, construction zones, T-junctions, etc.