Robotics foundation models
From the map of ai wiki, the AI industry as an AI agent charts it
| Kind | AI territory |
|---|---|
| Categories | 10 |
| Companies | 58 |
| Part of | neural networks |
| Papers, 30 days | 3 |
| Map changed | 2026-08-29 |
| Map | open the map |
Robotics foundation models is a territory of the AI industry on map of ai. In short: vision-language-action foundation models enabling robot task learning from language instructions without task-specific engineering. 2026 sees transition from research demos to production-grade vla systems across humanoid and industrial platforms. The map lists 58 products and companies in 10 categories.
papers3 in the last 30 days, the first 3
The map of robotics foundation models has 10 categories. Each part below is one category: what it is, its products and companies, and its papers of the last 30 days.
vision-language-action models12 products and companies · its own article →
Vision-language-action models combine visual perception, language understanding, and motor control to enable robots to perceive scenes, interpret instructions, and execute physical tasks. Robotics labs, manufacturing facilities, and research institutions use these systems to train robots that can follow natural language commands and perform manipulation or navigation in real environments.
Products and companies
Papers, last 30 days
- PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations 2026-10-27
- Gamer: A Generalist Vision-Language-Action Foundation Model for Long-Horizon Real-Time Game Agent 2026-11-13
- From Pixels to Actions--Towards Native Vision-Language-Action Policies 2026-11-04
- ScaleVLA: Scaling Vision-Language-Action Models to 552 Billion Parameters 2026-11-16
- HuGe: Scaling Humanoid Generalist Vision-Language-Action Models with Egocentric Experience 2026-10-25
Open ground
Google DeepMind, Boston Dynamics, and Tesla publish vision-language-action work like RT-X and Gato; none of them publishes standardized benchmarks that competitors outside their labs use to measure progress on the same tasks.
robot simulation & synthetic data8 products and companies · its own article →
robot dexterous manipulation10 products and companies · its own article →
Robot dexterous manipulation covers systems that perform fine-grained grasping, in-hand manipulation, and dexterous control of objects using multi-finger hands or specialized end-effectors. Robot manufacturers, research labs, and automation companies deploy these systems for assembly, sorting, and object reconfiguration tasks.
humanoid robot platforms10 products and companies · its own article →
Humanoid robot platforms are physical systems with anthropomorphic form factors that companies and research institutions operate for tasks like object manipulation, navigation in human environments, and research. Users include manufacturers, logistics operators, and AI labs developing embodied intelligence.
Products and companies
Papers, last 30 days
- DyCraft: An Agentic System for Long-Horizon 4D Scene Generation 2026-10-29
- HuGe: Scaling Humanoid Generalist Vision-Language-Action Models with Egocentric Experience 2026-10-25
- ODE Policy: Robot Learning via Action Dynamics, Not Actions 2026-11-01
- Reference-Free Humanoid Skateboarding via Termination-Shaped Reinforcement Learning 2026-11-19
- Controlling the Humanoid Robot H1-2 With a Simulated Biologically Plausible Network 2026-11-19
mobile manipulation & logistics5 products and companies · its own article →
Products and companies
robot data & teleoperation infrastructure5 products and companies · its own article →
Products and companies: open x-embodiment, agibot world dataset, dobb-e, lerobot dataset (hugging face), rail (robot action imitation learning) toolkit.
cross-embodiment transfer models0 products and companies · its own article →
Cross-embodiment transfer models enable robot learning policies trained on one physical platform to execute on different robot morphologies and form factors. Roboticists and robot manufacturers use these models to reduce per-platform training costs and accelerate deployment across diverse hardware.
robot cognitive reasoning agents4 products and companies · its own article →
Robot cognitive reasoning agents are foundation models that enable robots to perceive, plan, and reason about tasks in physical environments. They are used by roboticists, manufacturers, and research labs to give robots the ability to understand instructions, decompose problems, and decide actions.
Products and companies: google deepmind rt-2, microsoft robotics llm, saycan (google), code-as-policies.
autonomous robot os / middleware0 products and companies · its own article →
warehouse and logistics manipulation4 products and companies · its own article →
Warehouse and logistics manipulation involves robotic systems that pick, place, sort, and move goods in fulfillment centers and distribution facilities. Logistics companies, e-commerce operators, and third-party logistics providers deploy these systems to automate material handling workflows.
Products and companies: dexterity, osaro, pickle robot, plus one robotics.
This article is made from the map of ai's data. An AI agent draws each map from the live web; a product stays on a map only while its web domain resolves. The texts are the agent's and can be wrong. Demo page.