绝影Lite3

Deep Robotics' industrial robots (X20, X30, Lynx M20/M20S/S10) operate fully autonomously — navigating routes, collecting sensor data, uploading it in real time, and returning to auto-charging stations without human involvement in the inspection task itself.

Vendor press releases and official videos assert full autonomous closed-loop inspection [1][3][30], and a research paper validates autonomous navigation over ~2.8 km outdoor routes on a comparable legged platform [21], but no independent third-party field audit confirms sustained real-world autonomous performance at the claimed accuracy levels; the European logistics deployment figures are explicitly flagged as unverified [14].

Deep Robotics has achieved real-world international deployments, including a Singapore power grid tunnel inspection robot delivery.

The Singapore power grid deployment is cited in an official Deep Robotics announcement [9][19], and Hannover Messe 2026 attendance with EU-compliant units is reported by trade press [14], but no independent customer statement, operational report, or third-party journalist has verified the Singapore deployment's scope or ongoing performance.

The Lynx M20S wheeled-legged robot achieves a top speed of 9 m/s and operates in temperatures as low as -30°C with IP67 protection.

These specs are reported by robotics trade press (Robotics & Automation News) covering the product launch [13][16], which constitutes secondary reporting of vendor-supplied data rather than independent testing — no teardown, benchmark test, or field trial by a neutral party has verified these figures.

Deep Robotics grew unit sales from 529 units (2023) to 3,936 units (2025), with 2025 revenue of CNY 337M (~$49.6M) and its first profitable year (net income CNY 28.68M).

These figures are drawn from Deep Robotics' own IPO prospectus as reported by multiple news outlets [10][11][29], which gives them higher credibility than a press release alone, but they remain unaudited by an independent public accountant in the sources provided and are subject to regulatory review as part of the STAR Market IPO process.

Deep Robotics' DeepVLA 1.0 end-to-end embodied AI platform represents a transition from classical MPC+WBC control to a foundation-model-based approach for its robots.

DeepVLA 1.0 is mentioned only in a video source reporting on the company's technology roadmap [29] with confidence rated 0.78 in the dossier; no research paper, independent benchmark, or third-party evaluation of DeepVLA 1.0's real-world performance has been identified.