Autonomous Mobile Robots

Run the Traffic Test

An autonomous mobile robot navigates dynamically rather than following a fixed path, which is what distinguishes it from a traditional guided vehicle running on wire or tape. That flexibility is the selling point and the source of the difficulty.

The market has clear leaders by job. For pallet and cart transport, OTTO Motors (Rockwell Automation) and MiR build autonomous vehicles rated by payload. For piece-picking fulfilment, Locus Robotics and Zebra's Fetch robots lead the follow-me and goods-to-person model. Sort candidates by whether they move pallets, totes, or pickers before comparing specifications.

The Bot Scout traffic test evaluates the fleet rather than the robot: what happens when two AMRs meet in a full aisle, when a pallet blocks a route, when three units queue for one charger, and when a person stops to talk in a doorway. A single unit in an empty building tells you nothing about any of it.

Throughput in practice is set by waiting, not by speed. Utilisation collapses well before reliability does, and utilisation is usually what the business case was built on.

The Software Is the Purchase

Fleet management assigns jobs, routes units, resolves conflicts, and reports on the whole system. Its integration with the existing warehouse or manufacturing execution system is routinely scoped late and routinely sets the schedule, a pattern also described in the warehouse automation guide.

Ask who arbitrates when two jobs claim the same unit, how priorities are configured, and whether operations staff can change them without a vendor visit. A fleet nobody on site can reconfigure becomes a fixed installation.

Wireless handoff is the layer underneath the fleet software that rarely gets its own line item. A robot that stalls crossing from a private 5G cell into a Wi-Fi zone looks identical to a navigation bug from the floor; see our breakdown of Celona Orion for how that handoff gap gets managed on mixed-network sites.

For inspection and data collection rather than transport, a different class of mobile robot applies; Boston Dynamics Spot illustrates that distinction, and it should not be compared with transport AMRs as if they were interchangeable.

• Test with the full fleet, not a single unit.
• Block a route deliberately and time the recovery.
• Check charger contention at peak, not at shift start.
• Confirm who can change priorities without a vendor visit.
• Scope the WMS or MES interface before signing for hardware.

What Makes a Robot Autonomous

A robot is autonomous when it decides how to achieve a goal rather than replaying a fixed path. Autonomous robots sense their surroundings, build or update a model of them, plan a route, and adjust that plan when the world changes.

The distinction that matters commercially is autonomous versus automated. An automated machine repeats a defined motion and stops when something blocks it; an autonomous robot is given a destination and works out the route itself, including around an obstacle that was not there yesterday.

Three capabilities have to be present. Perception builds the picture from lidar, cameras, and odometry; planning turns a goal into a route; and control executes that route while correcting for drift and slip.

This is why an autonomous robot needs no floor magnets, wire, or reflective tape. Guided vehicles follow infrastructure laid into the building, and removing that infrastructure is the whole economic argument for the autonomous version.

Autonomy is a spectrum rather than a switch. Most deployed autonomous robots run supervised, with a human handling the exceptions the robot escalates, and the exception rate is the number that decides whether a fleet actually saves labour.

Warehouse and yard AMRs operate in a mapped, controlled space with reliable connectivity, which is the easy end of that spectrum. Our look at the TALUS military logistics platform covers the harder case: unmapped terrain and degraded or denied communications, where perception and planning have to work with far less certainty about the environment.

• Perception — lidar, cameras, and odometry combined into a live map
• Localisation — knowing where the robot is on that map, continuously
• Planning — turning a destination into a route around current obstacles
• Control — executing the route while correcting for wheel slip and drift
• Exception handling — escalating to a human when the situation is outside the robot's competence

Floors, Networks, and People

AMRs need floor conditions, network coverage, and defined traffic rules. Ramps, dock plates, thresholds, and reflective surfaces all affect navigation, and network dead zones produce stoppages that look like robot faults.

Shared space is the operating environment, not an exception. OSHA notes that many robot incidents occur during non-routine work such as setup, testing, and maintenance, which in a live building happens alongside pedestrians and forklifts.

Staffing changes shape rather than disappearing. Fleet recovery, exception handling, and someone who owns the data are new roles, and a plan that omits them will find them anyway.

Recent AMR fleet deployments and mobile-manipulation deals are tracked in warehouse robotics news.

A data hall is another mapped indoor space for these fleets, with thermal patrol and cable handling covered in robots that patrol data centers.

AMR vs AGV Compared

An automated guided vehicle follows a fixed path laid into the building, and an autonomous mobile robot plans its own route. That single difference sets the cost, the flexibility, and the failure mode of everything downstream.

An AGV tracks a wire, a magnetic tape, or a reflector grid. The path is infrastructure, so it is installed once, surveyed, and changed by a contractor. An AMR carries a map and a sensor set, plans around a pallet left in an aisle, and gets rerouted from software. Buying an AGV means buying a route. Buying an AMR means buying a vehicle and the fleet software that decides where it goes.

The sticker price understates the gap. An AGV vehicle is usually cheaper and its installation is not, because tape, wire, or reflectors have to go into the floor and be resurveyed whenever the layout moves. An AMR costs more per vehicle and installs against a map instead of a trench, so a rack move is a remap rather than a rebuild.

The two also sit under different US safety standards. A driverless guided vehicle is covered by ANSI/ITSDF B56.5, written around vehicles that follow a predetermined guidepath. A mobile robot that plans its own route falls under ANSI/A3 R15.08. That standard exists because neither B56.5 nor the stationary-robot standard covered a machine that navigates on its own. ITSDF publishes B56.5 and A3 publishes R15.08. A site running both vehicle types is covered by both, and ANSI sells the two together as an industrial mobile robot safety set.

Neither wins outright. A fixed route running the same load between two points all day is exactly what an AGV is for, and paying for autonomy that never gets used is waste. An AMR is worth its higher price in a building whose layout changes, whose aisles carry people and forklifts, or whose routes are not settled a year ahead.