Dave Savostyanov
Robotics and physical AI Sections
AI research areas

Robotics and physical AI

Learning to act in the physical world through robot policies, navigation and autonomous driving.

By Dave Savostyanov. 2025 data · Sources: 26 Sep 2026

2025 figures. Papers are OpenAlex query matches; author ranges and investment allocations are scenarios. Author counts marked “observed” cover the full query result. Areas overlap and cannot be added together. About the data.

Bars compare publication counts within this page. Topic tags describe subtopics, not separately measured markets. Organization examples link to their sources; activity was checked in September 2026.

These areas include classical robotics. Their figures do not measure the physical AI market alone.

Robot learning and vision-language-action models

Imitation learning, vision-language-action models, manipulation, locomotion, sim-to-real, and safe control.

Embodied AI VLA Imitation learning
3,215 papers
2.1× vs. 2023
Authors · scenario
4.6k–16k
Investment · scenario
$1.96B–$7.84B
Sources and methodology

Publication activity

2019
602
2020
821
2021
1,079
2022
1,214
2023
1,509
2024
2,045
2025
3,215

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("robot learning" OR "vision language action" OR "robot manipulation" OR "imitation learning" OR "robot locomotion" OR "sim to real")

Retrieved 2026-09-26.

OpenAlex query results

Publishing authors

Base scenario: 5.4k publishing authors. Paper count × 5.00 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 4,552 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 3,215 papers (31.1%). 14.8% of authorship records lack an Author ID; 8.5% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 5.00; 95% bootstrap interval 4.79–5.21. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

OpenAlex sample query

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $3.92B.

  • Robotics: $7.84B × 50.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Stanford AI Index 2026 / Quid

Scope and related research

Focuses on physical actions. General reinforcement-learning methods are listed separately.

Industrial AI, logistics and energy
ICML 2026 — research topics AAAI 2026 — AI taxonomy

Organizations

Selected examples
  • Physical Intelligence Research

    Generalist robot policies learned across tasks and robot embodiments.

  • Google DeepMind Research

    Models, world models and robotics.

  • NVIDIA Models

    Isaac GR00T develops robot foundation models and policy-training workflows.

More organizations (14)
  • 1X Robotics product

    Humanoid robotics and world models.

  • AGIBOT Robotics product

    Embodied AI and robots.

  • Agility Robotics Robotics product

    Warehouse humanoids.

  • Apptronik Robotics product

    Humanoid robotics.

  • Boston Dynamics Robotics product

    Mobile and humanoid robots.

  • Figure Models

    Helix learns humanoid control with a vision-language-action model.

  • Galbot Robotics product

    Embodied robotics.

  • Intrinsic Product

    Robot software platform.

  • Nomagic Product

    Warehouse robotic manipulation.

  • Sanctuary AI Robotics product

    General-purpose robots.

  • Skild AI Product

    General-purpose robot intelligence.

  • Tesla Robotics product

    Optimus is a humanoid robotics program developing perception, navigation and control.

  • UBTECH Robotics product

    Humanoid robotics.

  • Unitree Robotics product

    Quadruped and humanoid robots.

Navigation and SLAM

Localization, mapping, state estimation, sensor fusion, motion planning, and embodied navigation.

SLAM Sensor fusion Robot navigation
3,185 papers
1.9× vs. 2023
Authors · scenario
3.7k–12k
Investment · scenario
$1.92B–$7.66B
Sources and methodology

Publication activity

2019
768
2020
983
2021
1,151
2022
1,323
2023
1,674
2024
2,170
2025
3,185

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("simultaneous localization and mapping" OR "visual navigation" OR "embodied navigation" OR "sensor fusion" OR "robot navigation") AND (("machine learning" OR "deep learning" OR "neural network" OR "artificial intelligence") OR "robot" OR "autonomous")

Retrieved 2026-09-26.

OpenAlex query results

Publishing authors

Base scenario: 4.1k publishing authors. Paper count × 3.82 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,700 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 3,185 papers (31.4%). 12.0% of authorship records lack an Author ID; 15.3% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.82; 95% bootstrap interval 3.63–4.01. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

OpenAlex sample query

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $3.83B.

  • Autonomous vehicles: $7.94B × 15.00%
  • Robotics: $7.84B × 15.00%
  • Internet of things: $14.6B × 10.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Stanford AI Index 2026 / Quid

Scope and related research

Covers physical autonomy beyond learned manipulation and control.

Industrial AI, logistics and energy
AAAI 2026 — AI taxonomy

Organizations

Selected examples
  • Skydio Product

    Autonomous drones.

  • Exyn Technologies Product

    Autonomous aerial mapping.

  • ANYbotics Product

    Autonomous inspection robots with localization and environment sensing.

More organizations (5)

Autonomous driving

Road-scene perception, behavior prediction, planning, and end-to-end driving.

Self-driving End-to-end driving
12,100 papers
1.5× vs. 2023
Authors · scenario
9.4k–47k
Investment · scenario
$2.98B–$11.9B
Sources and methodology

Publication activity

2019
3,597
2020
4,788
2021
5,476
2022
6,315
2023
7,964
2024
10,254
2025
12,100

Matches include papers applying these methods. Search precision and recall have not been systematically measured.

Title and abstract matches for articles, preprints and reviews; retracted work is excluded. Versions and overlapping areas may be counted more than once.

("autonomous driving" OR "self driving" OR "autonomous vehicle")

Retrieved 2026-09-26.

OpenAlex query results

Publishing authors

Base scenario: 16k publishing authors. Paper count × 3.87 known authors per paper ÷ 3 papers per author per year. The range assumes 1–5 papers per author; the base and both ends of the range are at least the 3,750 distinct Author IDs already observed in the sample.

The publication rate is an assumption, not a measured rate for this field. The range is a scenario, not a confidence interval.

Random sample: 1,000 of 12,100 papers (8.3%). 11.2% of authorship records lack an Author ID; 19.7% of retrieved works lack an abstract. Retrieved 2026-09-26.

Mean known authors per paper: 3.87; 95% bootstrap interval 3.71–4.04. This describes sampling variability, not search accuracy or uncertainty in the number of researchers.

Applied coauthors are included. Missing or incorrectly linked author records affect these figures. They do not measure research jobs or everyone working in a field.

OpenAlex sample query

Private investment

An assumed allocation of broader investment segments. The range is 0.5–2× the base case. Shares are editorial assumptions, not measured deals or research spending.

Base case: $5.96B.

  • Autonomous vehicles: $7.94B × 75.00%

Company investment, not revenue or research spending. Ranges are scenarios, not confidence intervals.

Stanford AI Index 2026 / Quid

Scope and related research

A system-level task that overlaps with vision, reinforcement learning, and robotics.

Industrial AI, logistics and energy
CVPR 2026 — research topics AAAI 2026 — AI taxonomy

Organizations

Selected examples
  • Waymo Product

    Autonomous driving.

  • Wayve Research

    Research on driving world models, 3D perception and reconstruction.

  • Tesla Product

    FSD (Supervised) driving assistance and the Robotaxi development program.

More organizations (9)
  • Applied Intuition Product

    Vehicle intelligence and simulation.

  • Aurora Product

    Autonomous trucking.

  • Baidu Product

    Apollo develops autonomous-driving technology for robotaxis.

  • Mobileye Product

    Driving perception and autonomy.

  • Nuro Product

    Autonomous driving technology.

  • Pony.ai Product

    Autonomous driving.

  • Waabi Product

    Autonomous trucking and simulation.

  • WeRide Product

    Autonomous driving.

  • Zoox Product

    Autonomous robotaxis.