🏠IT之家•Stalecollected in 1m
Li Auto teases advanced parking AI in Livis event
💡Li Auto is doubling down on embodied AI; see how they plan to solve complex autonomous parking challenges.
⚡ 30-Second TL;DR
What Changed
Li Auto held the Livis Day software and embodied AI conference.
Why It Matters
This signals Li Auto's continued investment in embodied AI for automotive scenarios. It highlights a strategic push to improve precision in complex parking environments.
What To Do Next
Monitor Li Auto's technical whitepapers on embodied AI to understand their approach to vision-based parking navigation.
Who should care:Developers & AI Engineers
🧠 Deep Insight
Web-grounded analysis with 12 cited sources.
🔑 Enhanced Key Takeaways
- •The Livis Day event on June 15, 2026, is centered on "embodied intelligence" and Li Auto's broader AI and software roadmap, with CEO Li Xiang asserting that 2026 is the final window for companies to become AI leaders and Level 4 autonomous driving will be deployed by 2028.
- •The new-generation Li L9 Livis SUV, launched last month, features two self-developed Mach M100 smart driving chips, each providing 1,280 TOPS of computing power, and integrates a 3D Vision Transformer (ViT) perception model that fuses LiDAR and visual data.
- •Li Auto's autonomous driving business has been restructured into an independent department, and the company is actively developing humanoid robots, including a two-wheeled robot expected to launch in mid-2024 for factory use, indicating a broader AI ecosystem strategy beyond vehicles.
- •An OTA update for Livis AI Glasses now allows users to initiate automated parking commands directly and includes a high-risk alert for Sentry Mode, demonstrating integration of AI across devices.
🛠️ Technical Deep Dive
- Mach M100 Chip: Li Auto's self-developed 5nm automotive-grade process chip, delivering 1,280 TOPS of computing power per chip, with the L9 Livis utilizing a dual-chip configuration for a total of 2560 TOPS. It is described as the first edge inference chip using a data flow architecture.
- Perception Model: The L9 Livis integrates a 3D Vision Transformer (ViT) perception model that fuses LiDAR and visual data, extending visible range by 50%.
- Maha VLA 2.1 System: An upgraded system that achieves a tenfold increase in multimodal computational capacity, enhancing autonomous driving, automated parking, and intelligent interaction, alongside improved risk prediction and intent understanding.
- MindVLA-o1 Foundation Model: Unveiled at NVIDIA GTC 2026, this next-generation autonomous driving foundation model features a native 3D ViT, a true 3D vision encoder that processes information directly in a three-dimensional world, unifying spatial geometry and semantics at the encoding stage. LiDAR functions as a high-precision ruler for geometric calibration within this framework.
- AD Max 3.0 (Previous Iteration): Featured a transition to a BEV (Bird's Eye View) perception algorithm and introduced an Occupancy algorithm utilizing four fisheye cameras to eliminate blind spots. It employed a heuristic search planning algorithm for optimal parking path identification and a low-latency control algorithm for swift maneuvers.
- Arteris FlexNoC 5 IP: This system IP technology is deployed in Li Auto's proprietary SoCs, such as the Mach M100, to manage AI compute data movement and integration automation, optimized for the 5nm process node.
- Chassis Technology: The L9 Livis features a fully drive-by-wire chassis, including active suspension, steer-by-wire, and brake-by-wire systems, designed to deliver millisecond-level response across the entire control chain.
🔮 Future ImplicationsAI analysis grounded in cited sources
Li Auto will likely accelerate its transition into a broader AI technology company beyond just automotive manufacturing.
The spin-off of the autonomous driving business, the focus on "embodied intelligence," and plans for humanoid robots indicate a strategic shift towards a wider AI ecosystem.
The integration of AI glasses for vehicle control, including parking, suggests a future where human-machine interaction with vehicles becomes more seamless and multimodal.
The recent OTA update enabling automated parking commands via Livis AI Glasses demonstrates Li Auto's commitment to expanding control interfaces beyond traditional in-car systems.
Li Auto's emphasis on self-developed chips and advanced 3D perception models will likely lead to more robust and sophisticated autonomous driving and parking capabilities.
The Mach M100 chip with high TOPS and the MindVLA-o1 foundation model with native 3D ViT represent significant investments in core AI hardware and software, aiming for human-level intelligence in spatial understanding.
⏳ Timeline
2015-07
Li Auto founded by Li Xiang.
2023-11
Introduced AD Max 3.0 parking feature, utilizing BEV perception and an Occupancy algorithm with fisheye cameras.
2025-03
Unveiled MindVLA, a high-level assisted driving solution based on a Vision-Language-Action (VLA) large model.
2025-12
Livis AI glasses went on sale, later receiving an update for automated parking commands.
2026-03
Unveiled MindVLA-o1, its next-generation autonomous driving foundation model, at NVIDIA GTC 2026, featuring a native 3D ViT.
2026-05
Launched the new-generation Li L9, including the Livis trim, equipped with self-developed Mach M100 chips and an 800V active suspension.
📎 Sources (12)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: IT之家 ↗