SafeAD provides an end-to-end perception stack that detects, tracks, and predicts the behavior of all traffic participants for autonomous driving. The system infers road layout and topology without external maps, enabling scalable operation in unseen environments. It supports flexible sensor configurations and offers fleet mapping tools to generate large‑scale maps from low‑fidelity data, extending sensing range and improving comfort and reliability.
Funding
Funding not disclosed
Founders
Product
Problem
Existing autonomous driving systems struggle with accurately interpreting complex real-world scenarios due to limitations in sensor fusion and scene understanding. Reliance on pre-defined maps restricts operation in unmapped or changing environments, hindering scalability and adaptability.
Solution
SafeAD develops an end-to-end AI perception system for autonomous driving that mimics human-like scene understanding. The system fuses data from camera, LiDAR, and radar to create a comprehensive 3D representation of the environment. Neural networks infer road elements and topology, generating centimeter-accurate high-definition maps. Continuous change detection triggers automatic map updates via the cloud, enabling scalable fleet learning and operation in dynamic environments.
Target Audience
The primary target audience includes automotive manufacturers and technology companies developing autonomous driving systems for next-generation vehicles.
Features
- AI-powered perception system for detecting and tracking all traffic participants
- Early sensor fusion approach for seamless integration of camera, LiDAR, and radar data
- Neural networks for inferring road elements, topology, and other semantic information
- Centimeter-accurate high-definition map generation with continuous cloud-based updates
- Fleet mapping solution for automated large-scale map creation from low-fidelity fleet data
- Flexible architecture supporting various sensor configurations, from camera-only to camera-radar-LiDAR setups
- Behavior prediction for safe navigation in complex scenarios