Minerva Debuted Roger 1.0 for Ordnance Disposal. Oil and Gas Pilots Start This Fall; Field Deployments Are Slated for 2027.

California-based Minerva Humanoids recently debuted Roger 1.0, a semi-autonomous humanoid built for explosive ordnance disposal and hazardous-materials response, National Defense reported on November 4. A trained operator supervises the robot from a safe distance through a virtual-reality headset and haptic gloves, seeing through its cameras and feeling what its hands touch, according to a company fact sheet cited in the article. The magazine published an artist's rendering, not a photograph of a field unit. Minerva says oil and gas pilot projects with Roger run this fall, with operational deployments scheduled to start in 2027.

Our view: This is a humanoid pitch aimed at the jobs where tracked single-arm robots still send people into the blast radius, not at folding towels in a warehouse. Bomb technicians already use remote platforms. The failure mode Minerva keeps naming is the "long walk": when the robot cannot finish a two-handed task, the operator puts on the EOD suit and approaches the device alone. Roger's claim is that five-finger teleoperation plus onboard balance and navigation can keep the specialist's judgment on the device while their body stays at the command post. That is a different product thesis from most 2026 humanoid headlines, and it will live or die on named pilots and render-safe results, not on another funding announcement.

What Roger is supposed to do

National Defense says Roger was originally designed to support NATO operations and is meant for bomb disposal; chemical, biological, radiological, and nuclear threats; and emergency response and first aid. Minerva's public-safety page frames the same stack: EOD and counter-IED work, CBRN sampling and decontamination, critical-infrastructure overwatch, and remote casualty care. Treat those as company mission maps, not as published field results.

The control stack is teleoperation with an AI layer underneath. Felber told National Defense that Minerva Intelligence, the company's in-house system, can run without a live cloud connection and handles low-level body awareness and navigation while taking continuous input from the human operator. "Having a human in the loop is a lot easier when you have five fingers, and the thing that you're operating has five fingers," he said, contrasting that with teleoperating a non-humanoid platform.

The October 6 company release adds that Minerva Intelligence manages balance, fall recovery, and navigation, that operators use a VR headset to see through the humanoid's eyes, and that the company claims operators learn full-body control in minutes. Dominic Asquith, a British Special Forces veteran and Minerva's head of hazardous operations, put the customer problem in one line in that release: "The tracked, single-arm systems we've relied on for decades send a lot of people on that walk because the task needs two hands and real dexterity."

Who built it, and how fast

Minerva raised approximately $10 million in a pre-seed led by General Catalyst, with Long Journey Ventures and Credo Ventures as co-leads, the company said on October 6 as it came out of stealth. The round lists additional participants including Hugging Face and several smaller funds. Founders are CEO Sandor Felber and CTO Maurice Rahme. Felber previously worked on Tesla's Optimus program and on launching an MIT CSAIL research line on AI-driven humanoids; Rahme spent six years at Boston Dynamics as a technical lead on Stretch, taking it from early prototype to commercial deployment, according to the release.

The company says it went from an idea to a walking Roger prototype in five months, with software and AI developed primarily in the United States and first prototypes manufactured and assembled in Germany. Headquarters are in San Francisco, with offices in Cambridge, Massachusetts, and Sindelfingen, Germany. Robin Dechant, a General Catalyst partner, said in the release that shipping hardware on that timeline "takes physical AI out of the lab and into the world where it matters." General Catalyst's own October 6 interview with Felber leans on older teleoperation research — including a 2004 analysis that found human failures far more often than mechanical ones on rescue robots — to argue that smarter models plus haptics can cut operator load.

Minerva also says it showcased at the NATO Explosive Ordnance Disposal Centre of Excellence in Samorin, Slovakia; that it has research partnerships with Mohamed bin Zayed University of Artificial Intelligence for energy-sector humanoids and the University of Waterloo for perception in smoke, darkness, and sensor failure; and that it is exploring deployments at two of the world's ten busiest airports after on-site public-safety testing. None of those claims name a customer, airport, or energy operator in the release.

What is still unproven

National Defense's illustration is labeled an artist's rendering. No article in this source set publishes a named oil and gas pilot site, a unit count, a price, a DoD contract, or a render-safe success rate under live conditions. The company estimates that energy work kills about 7,000 people a year worldwide — roughly one death every 75 minutes — by applying an International Labour Organization fatal-accident rate to an International Energy Agency energy-workforce figure; that is Minerva's estimate, not an independent fatality census for the tasks Roger would replace.

The timeline Minerva has put on the record is concrete enough to check: paid pilots this fall, operational deployments starting in 2027. Until those land with named sites and metrics, Roger remains a well-financed prototype thesis about keeping bomb techs and energy specialists off the long walk.

Objections

"EOD already has robots." Yes. The open question is whether two-handed humanoid teleoperation cuts how often the suit still goes on. Asquith's "long walk" is the right metric, and it is not yet a published field result for Roger.

"Five months to a walker is marketing." The company and its lead investor both repeat the five-month claim. A walking prototype is not a certified EOD system. Watch whether 2027 deployments keep the same platform or a redesigned one.

"Hazardous teleop is a niche." Niches with certification, classified procedures, and sovereign-supply rules can still be large contracts. They also move slower than warehouse demos. That is a feature for safety cases and a risk for a pre-seed calendar that already points at 2027.

Three things to check

1. Named oil and gas pilots this fall. National Defense and the company both say pilots start now. Publish the site class (offshore platform, refinery, midstream) and the task list, not only that a pilot exists.
2. Any airport or EOD proof point with a public sponsor. The release says Minerva is exploring two of the world's busiest airports and has already done on-site testing. A named agency or airport authority would turn marketing into a deployable reference.
3. What "operational deployments" in 2027 means. Count of robots, hours under supervision, and whether a human still takes the long walk when Roger stalls are the numbers that separate a humanoid story from another teleop demo.