Deep Robotics Unveils Five Embodied AI Research Advances for Its Brain Architecture

DEEP Robotics unveiled five research advances supporting DEEPRobotics-Brain, its embodied intelligence architecture for real-world robot operation, the company said on October 24, 2026. The work spans physical prediction, navigation and manipulation, policy optimization, predictive memory, and data-driven evaluation, and includes DR-WM-1.0, DR-VLA-2.0, DR-APO, DR-Agent, and DR-Benchmark. GlobeNewswire published the announcement.

What the Five Deep Robotics Advances Cover

The five research areas explore how robots can anticipate action outcomes, translate instructions into motion, improve learned policies, revise outdated spatial beliefs, and measure progress using real-world data. The research involves collaborations with researchers from the Massachusetts Institute of Technology, the University of California San Diego, the University of Illinois Urbana-Champaign, Tsinghua University, Zhejiang University, the Hong Kong University of Science and Technology, and the Beijing Institute of Technology.

DR-WM-1.0 incorporates robot-terrain contact dynamics into action consequence prediction, drawing on 2.58 million image-mask pairs and more than 100,000 real-robot clips. DR-VLA-2.0 combines action-free pretraining on human video with adaptation using robot demonstrations, and its EgoLynx pipeline comprises 10,000 hours of curated egocentric web video and 2,000 teleoperated robot navigation clips. On the held-out EgoLynx-Robot test split, DR-VLA improves the Quadruped Navigation Score by approximately 5% and reduces open-loop final heading error by 14.6%.

DR-APO raises pretrained NavDP's simulated success rate from 70.33% to 89.55%, an increase of 19.22 percentage points. In real-world navigation tests on the DEEP Robotics LYNX M20 wheel-legged robot, success rises from 25% to 65%, a gain of 40 percentage points. DR-Agent draws on EvolvingNav, a navigation method built around predictive 4D belief, and its EvoWorld-Bench benchmark comprises 54 simulated scenes, 803,680 task instances, and approximately 239,000 dynamic state changes.

Deep Robotics' Funding and Valuation Records

According to akta.pro company records, Deep Robotics closed a Series C financing of over 500 million RMB, about $70 million, in December 2025, co-led by CMB International and ChinaAMC with strategic participation from China Telecom and China Unicom funds, at a reported pre-money valuation of approximately 8 billion RMB, about $1.1 billion. The same records list total funding of $140.5 million and a valuation estimate of $1 billion to $5 billion. Deep Robotics was founded in 2017 and is headquartered in Hangzhou.

Earlier Deep Robotics Deployments

DEEP Robotics deployed its Lynx M20S wheeled-legged robot in the Ailao Mountains in October 2026 to assist researchers with route scouting, wildlife observation, and equipment transport, FinancialContent Business Page reported. In September 2026, the company deployed patrol and firefighting robot dogs at Haining Leather City, a 1.3-million-square-meter market with over 6,000 businesses, where the robots use LiDAR, cameras, and gas sensors to detect anomalies. DEEP Robotics also deployed its Lynx M20S to patrol a desertification control plantation in China's Tengger Desert in September 2026, carrying 35 kg of supplies and using thermal imaging to drive away animals at night.

Where the Research Preprints Are Available

The research preprints are available in the research section of the DEEP Robotics Developer Center, the company said.