Quadruped Robots for Security Patrol: What They Actually Do

Legged robots are being pitched hard to security and facilities teams right now, and most of the pitches skip the part that matters. So here is the honest version, from someone who sells them.

What a Patrol Quadruped Actually Is

Strip away the marketing and you have a battery-powered platform that carries sensors along a route, on legs instead of wheels, and streams what it sees back to somebody. The legs are the whole point: they let it take stairs, curbs, catwalks, gravel, and cable trays that stop a wheeled robot dead.

What it is not: it does not intervene, it does not detain, it does not make judgment calls, and it does not remove the need for a human in the loop. Every credible deployment has a person watching the feed. The robot changes how often a route gets covered and what gets recorded. It does not change who is responsible.

Where They Genuinely Earn Their Keep

• Routes nobody wants to walk at 3 a.m. Substations, tank farms, water treatment plants, remote yards. The robot does not get bored on the two hundredth lap, and it logs every one of them.
• Environments that are unpleasant or hazardous for people. Post-incident checks, confined areas, hot or noisy plant, anywhere you would rather send a machine first.
• Routes with stairs and level changes. This is the single strongest case for legs over wheels. If your route is flat concrete, a wheeled robot is cheaper and runs longer.
• Meter reading and equipment checks on a fixed round. Consistent camera position and consistent timing make automated comparison possible in a way a clipboard never does.
• Thermal anomaly detection. A robot carrying a thermal payload past the same panel every night builds a baseline. Deviations become obvious.
• Documented coverage. For some organizations, the timestamped proof that a route was walked is worth as much as the walking.

Where They Disappoint People

These are the deployments that go badly, and they go badly for predictable reasons.

• Open, flat, fixed areas. If the space does not change and there are no level changes, fixed cameras cost less, run forever, and need no charging.
• Anything requiring response. The robot sees the open door. Someone still has to go close it.
• Crowded public spaces. The robot becomes the attraction. People stop, film it, and follow it. That is fine for a marketing activation and terrible for surveillance.
• Deployments with nobody assigned to it. A patrol robot needs an owner: someone who charges it, updates routes, watches feeds, and handles it when it gets stuck. Without that person it becomes an expensive shelf ornament within about two months.
• Severe weather, if you bought the wrong protection rating. See below.

Five Questions to Ask Before You Buy

1. What is the ingress protection rating, really?

This is where most outdoor deployments fail. IP54 means dust and splashing water. It does not mean rain for six hours, it does not mean pressure washing, and it does not mean snow. If the robot lives outdoors, you want an IP66 or IP67 class machine, and you should confirm the rating in writing rather than accepting "weatherproof."

Also check the operating temperature range. Many robots in this category are rated 0°C to 40°C. That rules out a large part of the country for several months a year.

And note: a mounted third-party sensor carries its own rating. An IP67 robot with an IP54 camera bolted to it is an IP54 system.

2. What is the real endurance at your payload?

Quoted battery life is usually unloaded and on flat ground. Add a thermal camera, a LiDAR and a compute box, then send it up stairs, and the number drops. Ask for endurance at rated payload, and ask for the distance figure, not just the hours. A robot that runs two hours but covers 3 km is a different tool than one that runs two hours and covers 6 km.

Then ask about hot-swappable batteries and autonomous charging docks. On a real patrol schedule, the recharge cycle is the constraint, not the run time.

3. Who writes the route, and can you change it yourself?

Some platforms let you walk a route once and have the robot repeat it. Others require SDK work. If your team cannot modify a patrol route without calling a vendor, the deployment will calcify.

Related: confirm whether the model you are quoting supports secondary development at all. On several platforms the education and enterprise variants can be programmed and the consumer variant cannot, and it is not a field upgrade.

4. Is it eligible for the way you are paying for it?

Two separate compliance questions get conflated constantly.

FCC equipment authorization is a radio-emissions question. It says the device's radios are authorized to operate in the United States. You can verify any FCC ID yourself, free, at apps.fcc.gov/oetcf/eas. Do that rather than trusting a spec sheet.

NDAA procurement eligibility is an entirely different regime governing whether federal agencies, or organizations spending federal funds, may buy a given piece of equipment based on country of origin and supply chain. A device can hold a valid FCC grant and still be ineligible for federal purchase.

If any part of your budget is federal, or your grant conditions reference country-of-origin restrictions, settle this before you spec a robot. It is the most common way these projects die late.

5. What happens when it falls over or gets stuck?

Ask whether the robot self-rights. Ask what the recovery procedure is when it wedges itself somewhere. Ask where it gets repaired and how long that takes. A robot that ships overseas for service is a robot you do not have for two months.

Rough Budget Bands

A Sensible Way to Start

Buy one unit in the middle band, pick one route, and give it an owner. Run it for ninety days and measure two things: how many laps it actually completed, and how many findings it produced that a person would have missed. If both numbers are good, you have a business case for a fleet. If the robot spent six weeks on a charger, you have saved yourself a much larger mistake.

Resist buying the industrial machine first. The protection rating and payload are real, but if the deployment fails it will fail for organizational reasons, not mechanical ones, and it is cheaper to discover that at $8,000 than at $40,000.

Quadrupeds We Stock

For evaluation and pilot work, the Unitree Go2 Pro is the entry point, with the Go2 X and Go2 Edu adding capability and development access as you move toward a real deployment. The Go2 Ent targets enterprise use. For production industrial patrol work, the Unitree A2 is the heavier platform.

Browse all Unitree quadruped robots, or send us the route and we will tell you honestly whether a robot is the right answer for it.

Frequently Asked Questions

Can a robot replace a security guard?
No. It replaces some of the walking, and it produces a better record of the walking. Response, judgment and intervention stay human.

How long do patrol quadrupeds run on a charge?
Commonly one to two hours in the pilot bands and three or more hours at rated payload on industrial machines. Always ask for the figure at your payload, not unloaded.

Can they work outdoors in rain and snow?
Only if the protection rating and temperature range support it. IP54 machines are fair-weather. Check the rating in writing, and remember that mounted sensors carry their own ratings.

Are quadruped robots NDAA compliant?
That depends entirely on the manufacturer and model, and it is a separate question from FCC authorization. If federal funds are involved, confirm eligibility with your procurement office before you spec anything.

Do we need a programmer?
For basic operation, usually not. For custom routes, integrations, or mounting your own sensors, you need a model that supports secondary development, and those are specific variants, not a field upgrade.

Have a route in mind? Email daniel@globaldronehq.com or call (858) 906-5185 and we will spec it with you. Purchase orders and tax-exempt purchasing welcome.