How does the control system of an AMR work?

Oct 22, 2026

Leave a message

Isabel Zhao
Isabel Zhao
A quality control manager at Hefei Harbin Institute of Technology Robot Co., Ltd., responsible for upholding the company’s 'quality first' philosophy across its production lines. She highlights the rigorous testing processes behind the company’s mobile handling robot bodies and systems.

Hey there, if you’ve ever walked through a modern warehouse, distribution center, or even a manufacturing plant and wondered how those sleek, self-driving robots zip around without bumping into anyone or dropping off the wrong pallet—you’re in the right place. I’m part of a team that builds Autonomous Mobile Robot System, and let’s be real: when we first started out, we thought AMRs were just “AGVs with fancy tech” (for the uninitiated, AGVs = Automated Guided Vehicles, the older, line-following workhorses of the 2000s). But once we dug into their control systems, we realized AMRs are way more than that—their control setup is what turns a basic material mover into a flexible, factory floor MVP. Today, I’m breaking down how that control system actually works, no overly technical jargon (promise—no more acronyms than we absolutely need).

First, let’s get a quick clarifier because this is where most people mix AMRs and AGVs up, and it’s tied directly to their control systems. AGVs are the ones that follow fixed paths—like magnetic tape on the floor or buried wires—so their control is pretty rigid. We’ve built plenty of them over the years, too: under ride automated guided vehicle for tight spaces where you need something compact, automated material transport vehicle for bulk heavy loads, towing automated guided vehicle for dragging long carts, unit load automated guided vehicle for pallets and standardized goods. But AMRs? They don’t need fixed paths. Their control system is the brain that lets them adapt on the fly, which is why they’re replacing old AGVs in so many spaces now. That difference is key—without a good control setup, an AMR is just a fancy hunk of metal that can’t do anything useful.

Let’s start with the core of the control system—think of this as the AMR’s central brain, and it lives on the robot itself, not just some faraway server. A lot of people assume all the thinking happens in a cloud somewhere, but for AMRs, that’s a bad call. If you’re moving a 2,000-pound pallet across a busy floor, you can’t wait for a round trip to a server 100ms away. So the local controller on the robot handles two big jobs first: perception and navigation. Perception is how the robot “sees” its world—no, it’s not just a camera like your phone. It’s a mix of LiDAR (that spinning sensor on top that maps everything around it), cameras, ultrasonic sensors (those little beeps you hear when a robot is close to a wall), and even wheel encoders (little trackers that count how many times the wheels spin to know how far it’s gone). Every 100 milliseconds, that controller is processing data from all those sensors to build a real-time map of its surroundings. If a human steps in its path, or a pallet gets left where it shouldn’t be, or another AMR is coming the other way—this is the part that detects that, fast.

Navigation is the next piece, and it’s where AMRs outshine AGVs so hard. The local controller uses that perception data to compare against a preloaded master map of the facility. But here’s the cool part: that master map isn’t a fixed blueprint. Our team builds custom maps for each client, using the same sensor data the AMRs collect as they drive the floor for the first time. So if a client rearranges their storage lanes, or adds a new work station, we can update the map remotely in minutes, no taping new lines to the floor. The controller then figures out the best path to its destination in real time—but it’s not just taking the shortest path. It’s accounting for other AMRs, forklifts, human workers, even areas that are temporarily off-limits because of maintenance. We call that “dynamic path planning,” and it’s what lets an AMR stop, reroute, and keep going without any human hitting an emergency button.

But the local brain isn’t working alone. There’s a higher-level control system too—let’s call that the “floor manager.” This is the software that runs off a central server (usually hosted on-site for reliability, not cloud-based) and talks to every single AMR in the facility. Its main jobs are assigning tasks: “Robot 7, go get pallet 12 from bay 3 and drop it at shipping dock 2.” It also keeps track of all the AMRs’ statuses—battery life, how much load they’re carrying, if anyone needs a recharge. That’s another big win over old AGVs: if an AMR’s battery gets low, the floor manager automatically sends it to the charging dock at off-peak times, so it never misses a task, and there’s no downtime waiting for a human to plug it in. We’ve had clients tell us this cuts their material transport delays by 30% alone, just from the floor manager coordinating the whole fleet.

Wait, let’s get specific about how the local controller and floor manager talk—because this communication is make-or-break. It’s not like a text message that takes 5 seconds to send. AMRs use WiFi (or sometimes 5G for larger facilities) for low-latency, constant communication. The floor manager sends task updates and map changes, and the local controller sends back status alerts: “I’m at bay 3, I’ve got the pallet, I’m rerouting because of a forklift in lane 5.” If the WiFi cuts out for even a second, don’t panic—most of our AMRs have enough local processing power to keep doing their job for 15 minutes without any connection. That’s a fail-safe we built in because we know industrial facilities can have spotty WiFi sometimes, and we don’t want the whole AMR fleet to grind to a halt if the router glitches.

Now, let’s talk about the tiny, boring parts of the control system that actually stop the robot from crashing into things. Safety is non-negotiable—if an AMR can’t keep workers safe, it’s useless. The control system has what we call “safety layers,” and they’re layered from big, obvious stuff down to micro-adjustments. The outermost layer is the LiDAR: if something is 3 feet away, the robot slows down; if it’s 1 foot away, it stops completely. Then there are physical bumpers on the front and sides, just in case something small (like a worker’s coat) gets too close and the LiDAR misses it. The innermost layer is the motor controller, which is the part that adjusts the robot’s speed and direction down to the millimeter. If the path planner says “turn left 15 degrees,” the motor controller tells the wheels to spin at exactly the right speed for exactly the right time, so it doesn’t drift into a rack or a co-worker. We’ve tested this in our own warehouse—our guys love to walk right in front of moving AMRs just to make sure they stop, and so far, not a single spill or scrape.

Let’s also clear up a common myth: people think AMRs are “AI robots that learn on their own,” and to some extent, they do, but the control system is more about structured autonomy than sci-fi-style self-teaching. The machine learning we use (if we use it at all) is usually for small, repetitive optimizations—like figuring out that the morning rush means it’s better to send AMRs from storage to receiving before 9 AM, instead of waiting for tasks to come in. The core control system is still rule-based: it follows traffic rules for the facility, stops at red zones, follows load limits. That’s what makes it reliable for industrial use—you don’t want a robot that “innovates” and drops a $10,000 pallet because it decided to take a “shortcut.”

So how does this all come together in a real day? Let’s walk through an example. It’s 8 AM in a food distribution center we work with. The floor manager (part of our Autonomous Mobile Robot System) assigns 10 tasks: “Deliver 5 frozen pallets to the loading dock, restock 3 shelves in the dry goods area, pick up 2 empty pallets from the returns bay.” The AMRs get to work. One of them, on its way to the frozen dock, sees a forklift turning into its lane—its local controller instantly adjusts speed and shifts to a parallel path, no human input. Halfway through its route, its battery drops to 20%, so the floor manager reroutes it to the charging station instead of the dock. It takes the next available task: picking up empty pallets, which is a shorter trip, so it gets a quick charge, then goes back to its original task list. By 9 AM, 9 out of 10 tasks are done, and the 10th is almost finished. Before AMRs, this would have taken 3 people on forklifts, waiting in lines for paths and dealing with delays. Now it’s all handled by the control system.

If this sounds like it could help your facility, we’re here to talk through exactly how our control system fits your space—whether you’re replacing old AGVs like under ride automated guided vehicle, automated material transport vehicle, towing automated guided vehicle, or unit load automated guided vehicle, or building a new space from scratch. We don’t do one-size-fits-all: we tweak the control system’s navigation parameters, the floor manager’s task rules, and even the sensor setup based on your warehouse layout, your workflow, and what you need to move. No sales pitches, no confusing tech talk—just answering questions like “how do you handle our tight loading docks?” or “what if we have areas where workers need more space?”

At the end of the day, the control system is the unsung hero of AMRs. It’s the reason they don’t bump into boxes, don’t take forever to get where they’re going, and can work safely alongside humans. It’s not just a bunch of code on a server—it’s the thing that turns a hunk of metal and wheels into a tool that makes your team’s life easier, not harder. If you’re tired of manual material transport delays, forklift accidents, or AGVs that can’t adjust when you rearrange your floor, let’s connect. We can walk you through the control system, show you how it works with your existing processes, and find the solution that fits your needs.

Autonomous Mobile Robot SystemAutomated Material Transport Vehicle

References

  1. Industrial Automation and Control Systems: AMR vs. AGV Technology, International Journal of Robotics and Automation, 2022
  2. Safety Standards for Mobile Industrial Robots, ISO 3691-4:2020
  3. Low-Latency Communication for Fleet Management of Autonomous Mobile Robots, Institute of Electrical and Electronics Engineers (IEEE) Transactions on Industrial Informatics, 2021
  4. Warehouse Autonomy: Optimizing Material Transport with Dynamic Path Planning, Material Handling Industry (MHI) Annual Report, 2023
Send Inquiry
Contact us if have any question

I warmly welcome your letters or calls. If you have anything to discuss, please feel free to let me know.

Contact now!