Most comparisons of AMRs and AGVs stop at a definition: an AGV follows a fixed path, an AMR navigates dynamically. Accurate, and not much help when you're deciding what to put on your floor. It doesn't tell you what changes in your throughput math, or why a facility with working AGVs might still be looking at AMRs.
At HESCO we work with controls engineers evaluating material handling automation as part of a broader Rockwell architecture. The short version: an AGV knows where it is because the facility tells it. An AMR knows because it localizes against a map it maintains itself. That difference drives infrastructure cost, obstacle behavior, throughput, and controls integration.
| Characteristic | AGV | AMR |
|---|---|---|
| Path definition | Fixed, defined by physical infrastructure | Generated dynamically per trip |
| Route change | Physical infrastructure modification | Software configuration |
| Obstacle in path | Stops until the obstruction is cleared | Re-plans and drives around |
| Floor modification | Required | None |
| Traffic handling | Typically zone-blocking | Fleet-coordinated, robots share position data |
An AGV installation means putting something in or on the floor along every path the vehicle travels. That's a one-time cost at commissioning, usually budgeted. What often isn't budgeted is the second installation: when a line moves or a drop point relocates, the path has to be physically rebuilt, and the floor work has to happen during a window when the area isn't running.
An AMR route change is a change to the map and the traffic rules. Speed limits and exclusion zones are configuration rather than physical constraint, so the robots operate within new rules without anything installed, cut or removed.
The practical question: how many times will this path change over the life of the system? If never, the AGV infrastructure cost is a one-time line item. If annually or more often, the economics shift — and the downtime per change may matter more than the material cost.
An AGV on a fixed path meets an obstruction — a pallet in the aisle, a cart left out at shift change — and stops, because the path is the only route it knows. It waits until someone clears it, and each occurrence is unplanned downtime for whatever that vehicle was feeding.
An AMR treats the obstruction as a routing input, detecting objects above and below the LiDAR plane and navigating around them without a protective stop. The mechanism is adaptive fieldsets — PL-d rated switching protective fields that change shape with direction and speed. A single static field would either be sized for worst-case speed and stop constantly in tight spaces, or sized for tight spaces and cap speed everywhere.
Worth being precise: modern AGVs are not unsafe. Both use safety-rated sensing and both stop for people. The difference is what happens to a non-hazardous obstruction — an AGV treats it as a stop condition, an AMR as a routing problem.
This is the least-discussed difference and often the largest, because it doesn't appear in a speed or payload spec. Traditional AGV deployments often charge to depletion: run until low, park at a station, return to work. Opportunity charging inverts that — the robot tops up between jobs during idle time, and across the OTTO lineup it's the default.
Mauser Packaging Solutions' public case study shows the effect, and it's a direct swap of one technology for the other in the same application. Their AGV system completed three to four missions before charging, spent 25 to 30 minutes at the charger, and dropped to around 15% battery. A single OTTO AMR completes up to 36 missions and goes to the charger at about 70%.
That produced a 6x throughput improvement at Oakville. Worth noting what the figure measures: OTTO is explicit that the 600% is based on missions completed before charging, not maximum runtime.
Why this matters for your math: size an AMR fleet using AGV assumptions — vehicles unavailable for fixed charging blocks — and you will overbuy. The question isn't top speed or even runtime; it's how many missions a robot completes between charges.
For a controls engineer, integration usually matters more than navigation. Material handling that can't talk to the rest of the plant is just a faster way to move things between islands.
AGVs assume the path is permanent and buy predictability in exchange for infrastructure. AMRs assume the floor changes and buy flexibility in exchange for more onboard sensing and software.
AMRs aren't strictly better. An AGV is still reasonable where the route is permanent, the path is segregated, the payload exceeds the AMR envelope, or an existing system already meets throughput. The clearest signal a facility has outgrown them is the Mauser pattern: the vehicles work, but they can't keep pace, and every layout change means going back into the floor.
A quick read on your own operation: count the layout changes in the last three years, then walk the route at shift change rather than mid-morning. Staged pallets and parked carts decide whether fixed-path stopping is an annoyance or a daily throughput problem.
Trying to work out which fits your operation? The HESCO team can walk through your workflows, layout constraints and controls architecture. Get in touch at hesconet.com/contact-us.