How fast is the autonomy industry growing?

Why use publications?

Industry revenue sounds like the obvious answer, but public market-size estimates are usually proprietary, definition-dependent, and updated irregularly. A live webpage cannot responsibly manufacture the missing months between reports.

Research output offers a defensible alternative. Autonomy remains a research-intensive field: advances in perception, planning, controls, simulation, verification, human factors, and safety cases flow through technical literature before—and after—they become products. Publication count therefore measures the rate at which the field is producing codified knowledge. It does not measure revenue, profitability, deployments, or safety.

What the graph actually does

On every visit, the browser takes its current calendar date and asks the OpenAlex Works API for yearly counts from January 1, 2015 through that date. OpenAlex supports full-text search, date filters, and grouping by publication year; its API response is converted directly into the graph above.¹ The latest bar is deliberately labeled year-to-date so a partial year is not silently compared with twelve complete months.

The query is intentionally short and inspectable:

search: "autonomous vehicle" OR "autonomous robot"
date:   2015-01-01 through [reader's current date]
group:  publication_year

This design means “now” advances automatically with the reader. It also means historical values can move slightly as OpenAlex adds records, merges duplicates, or improves metadata.

What this metric includes—and misses

The two phrases capture a broad core of ground-vehicle and robotic autonomy while avoiding the much noisier word autonomous by itself. They will still include academic work with no commercial path. Conversely, they miss relevant papers that use terms such as unmanned aircraft, self-driving trucks, autonomous navigation, embodied AI, or field robotics without either chosen phrase.

There are also database effects. OpenAlex is a fully open scholarly index assembled from sources including Crossref, PubMed, repositories, and publisher data.² Independent assessments find it valuable and unusually broad, while also documenting gaps and inconsistencies in fields such as language and document type.³ Search-term counts should therefore be read as a repeatable sample of the literature, not a census.

The larger measurement problem

“Autonomy” is not one conventional industry classification. SAE’s driving-automation taxonomy alone distinguishes six levels for road vehicles,⁴ while autonomy work in aviation, industrial robotics, agriculture, and defense follows different regulatory and commercial structures. Any single financial total will change dramatically depending on whether it includes advanced driver assistance, sensors, simulation software, conventional industrial automation, or only systems that operate without an onboard human.

A fuller dashboard would pair this research signal with several independently refreshed measures: commercial driverless rides, deployed autonomous fleets, regulatory permits, autonomy-focused job postings, venture investment, and revenue disclosed by public companies. Until those inputs are available through stable public interfaces, the literature series is useful precisely because its limits are visible.