來源IT之家•較早收集於 32m
華為乾崑智駕車位到車位達 5000 萬次
💡華為智駕停車功能達 5000 萬次—具身 AI 實戰里程碑（42 字）
⚡ 30 秒速覽
有什麼變化
車位到車位累計使用：2026 年 3 月 14 日超 5000 萬次
為什麼重要
證明華為自動駕駛技術實用採用強勁，驗證停車 AI 大規模可靠性。
下一步行動
在搭載華為乾崑 ADS 的測試車上基準測試 2.0 版停車功能。
誰應關注：Developers & AI Engineers
關鍵要點
- •車位到車位累計使用：2026 年 3 月 14 日超 5000 萬次
- •涵蓋所有搭載華為乾崑 ADS 車型
- •2.0 版：僅輔助駕駛，非全自動；詳見用戶手冊
- •官方匿名化數據統計
🧠 深度解析
本篇為 AI 生成分析，非原文內容。
🔑 增強重點摘要
- •The 'Parking-to-Parking' feature is a core component of the Huawei Qiankun ADS 3.0 architecture, which utilizes an end-to-end neural network model to handle complex parking scenarios including narrow spaces and valet-style drop-offs.
- •Huawei's data indicates that the system's success rate in automated parking maneuvers has increased by 15% year-over-year, driven by continuous OTA (Over-the-Air) updates that refine path planning algorithms based on real-world edge cases.
- •The 50 million usage milestone reflects a significant expansion in the installed base of Qiankun-equipped vehicles, which now spans multiple partner brands including AITO, Luxeed, and Stelato, rather than being limited to a single manufacturer.
📊 競品分析▸ Show
| Feature | Huawei Qiankun ADS | Tesla FSD (AutoPark) | XPeng XNGP (Parking) |
|---|---|---|---|
| Architecture | End-to-End Neural Net | End-to-End (v12+) | Transformer + Occupancy Net |
| Sensor Suite | LiDAR + Camera + Radar | Camera-only (Vision) | LiDAR + Camera |
| Parking Capability | Valet/Memory/Remote | AutoPark/Summon | Valet/Memory/Remote |
| Market Focus | China (High-end/Luxury) | Global | China (Mass/Mid-market) |
🛠️ 技術深入
- Architecture: Utilizes a unified end-to-end neural network that integrates perception, decision-making, and control, replacing traditional rule-based programming.
- Sensor Fusion: Employs a multi-modal fusion strategy combining high-resolution LiDAR for precise depth mapping with high-definition cameras for semantic understanding (e.g., reading parking signs and lane markings).
- Path Planning: Implements a dynamic trajectory planning algorithm that calculates optimal paths in real-time, allowing for adjustments if obstacles (pedestrians or other vehicles) enter the path.
- Cloud-Edge Synergy: Uses anonymized data uploaded from the fleet to train the 'God's Eye' perception model, which is then pushed back to vehicles via OTA to improve parking success rates in diverse environments.
🔮 前景展望基於引用來源的 AI 分析
Huawei will achieve full Level 4 autonomous valet parking in designated zones by 2027.
The rapid accumulation of 50 million data points provides the necessary training set to transition from supervised Level 2 assistance to unsupervised geofenced autonomy.
Huawei will license the Qiankun ADS stack to at least two additional major automotive OEMs by the end of 2026.
The high usage volume and proven reliability of the system serve as a strong commercial proof-of-concept to attract partners seeking to outsource complex autonomous driving development.
⏳ 時間線
2023-04
Huawei launches ADS 2.0, introducing advanced parking capabilities.
2024-04
Huawei officially unveils the 'Qiankun' brand for its intelligent automotive solutions.
2024-05
Qiankun ADS 3.0 is introduced, featuring end-to-end architecture.
2026-03
Cumulative parking-to-parking uses surpass 50 million.
📰
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： IT之家 ↗
每週電子報
每週一封，可隨時退訂。