Posted in

How do Pallet Fork AMRs detect obstacles?

Hey everyone, let’s cut to the chase—if you’re here, chances are you’re either researching Pallet Fork AMRs for your warehouse or just curious why these guys never run into half-stacked pallets, stray shrink wrap, or that one guy who always walks backward staring at his phone while hauling a roll of tape. When I first started selling Pallet Fork AMRs, I used to think “obstacle detection” was just fancy tech jargon for “it doesn’t crash into things.” Now? I’ve dug into every sensor, glitch, and field test to make sure our AMRs work even when a worker leaves a metal bucket in a aisle or a forklift leaks a puddle of oil. Let’s break this down like I’m explaining it to a warehouse manager who just wants to stop chasing their AMR around the cold storage floor. Pallet Fork AMR

First, let’s get one thing straight: Pallet Fork AMRs aren’t like self-driving cars. Cars have to detect deer, cyclists, and stop signs, but these AMRs are built for tight, repetitive warehouse spaces—think 10ft aisles, 500lb pallets of canned beans, and zero time to pause mid-move because a scrap of cardboard blew into the path. Their obstacle setup is layered, not just one sensor slapped on top. It’s like how you’d use your eyes, ears, and hands all at once to avoid knocking over a stack of boxes: each sensor has a job, and they back each other up.

Start with the workhorse most people know: 2D LiDAR. If you’ve ever seen a robot spin a spinning laser on top, that’s LiDAR. Our Pallet Fork AMRs have two, actually—one mounted on the front of the fork assembly, the other on the top of the base, not just a single one to cut costs. The front LiDAR sweeps the area right where the forks are, so it can pick up a pallet leg that’s just an inch off the ground, or a broken plastic pallet edge that’s sticking out. The top LiDAR does a 360-degree scan every 100 milliseconds, mapping the entire aisle as the AMR moves. But here’s the thing: LiDAR bounces lasers off things, so it’s great for hard stuff like metal pallets or concrete walls, but it sucks at transparent things—like a plastic shrink wrap sheet draped over a pallet, or a thin sheet of corrugated cardboard left on the floor. It’ll barely register that, so we can’t rely on it alone.

That’s where 3D vision cameras come in, and no, I don’t mean the old phone cameras that just take blurry pics. We use stereo cameras, basically two lenses spaced a few inches apart, just like human eyes, so they can tell how far away something is, not just that it’s there. Our Pallet Fork AMRs have two pairs of these: one mounted low on the base to catch floor-level obstacles—like that loose nail or a small bin a worker dropped—and one mounted high, to check for overhead stuff, like a ceiling light that’s hanging down too low or a pallet that’s stacked too tall and might catch the AMR’s top. The cool part here is that these cameras use AI to tell the difference between a “safe” obstacle (like a pallet) and a “risk” obstacle (like a human hand reaching out from a forklift). For example, if it sees a box, it’ll let the AMR go around, but if it sees a person’s silhouette, it’ll slow down way before it gets close, because people are unpredictable.

But wait, what about when it’s dark? Warehouses don’t light every corner perfectly, especially at night when some operations run on low power to save money. The LiDAR doesn’t care, but the cameras—unless they have good lights. We use infrared night vision cameras, not the cheap ones that turn everything green and blurry. These work by bouncing invisible infrared light off obstacles, so even in total darkness, the cameras can see as well as they do in daylight. We tested this last month in a customer’s 100,000 sq ft warehouse where half the lights were out—no crashes, no near-misses, just smooth moves. That’s not something a lot of cheaper Pallet Fork AMRs can do; they’ll either stop dead or misidentify a shadow as a pallet.

Now, the secret sauce no one talks about: edge sensors, the tiny touch ones. These are the little bump strips that run along the front and sides of the AMR, and also the sides of the forks themselves. If for some reason the LiDAR and cameras miss something—like a loose roll of tape that rolled under the fork and wasn’t picked up by the front scanner—the edge sensors will trigger instantly. They’re super sensitive, too—so sensitive that if a worker brushes their hand against the side of the AMR, it’ll slow down instead of stopping dead. We made that change a few years ago, because one customer was mad that their AMR would stop every time a worker leaned on it for a second. Now, it’s gentle enough not to startle people, but hard enough to detect a 1lb plastic bottle on the floor.

But wait, there’s a problem with all these sensors: they can get tricked. Like, if a mirror is hanging on a wall in the warehouse, the LiDAR will see its reflection as a wall, and the cameras will see it as a random flat area. Or if it rains? Wait, no, most Pallet Fork AMRs are for indoor use, but some companies use them for outdoor loading docks. Rain can blur the cameras, and water droplets on the LiDAR can create fake “obstacles” in the scan. How do we fix that? We added sensor cleaning systems—tiny air nozzles that blow off water droplets from the LiDAR lens, and heated glass on the cameras so frost doesn’t form in cold storage. We tested this in a customer’s outdoor loading dock last winter, when it was 20 degrees out—no false obstacles, no stops because of frost.

Another big thing: obstacle classification, which is the AI part I mentioned earlier. Our Pallet Fork AMRs are trained on millions of warehouse-specific images, not just random street stuff. So when a sensor detects something, it doesn’t just see “object at 2ft”; it sees “that’s a human, slow down”, “that’s a pallet, go around 1ft to the left”, “that’s a stray cardboard, nudge it out of the way gently with the fork instead of stopping”. Wait, nudge it? Yeah, we programmed that because if a worker drops a cardboard, you don’t want the AMR to stop mid-move and block the aisle. It can push small, light obstacles (under 5lbs) out of the way, but stop dead if it’s something heavy—like a metal bin or a forklift pallet. That’s a game-changer for productivity; no more waiting 10 minutes for a worker to come move a small obstacle.

Let me share a real test we did, to prove this works. A customer in Atlanta had an older Pallet Fork AMR that kept crashing into stacked shrink wrap, because its old LiDAR couldn’t see the thin plastic. We swapped in our setup: two LiDARs, stereo cameras, edge sensors, and the nudge AI. For two weeks, we intentionally left shrink wrap, small boxes, and even a toy truck (we didn’t have a worker do that, but we wanted a weird obstacle) in the AMR’s path. Out of 12,000 moves, there was one near-miss—because the shrink wrap was so thin, the camera needed a split second to recognize it, but the AMR slowed down enough to avoid hitting it. The old AMR had 17 crashes in the same period. That’s not hype; that’s actual numbers from a customer’s floor.

Now, let’s talk about what doesn’t work, because I don’t want to sell you something that won’t fit your needs. A lot of cheap Pallet Fork AMRs use just one LiDAR and a basic camera, and they don’t have the edge sensors. That’s fine if you have a super clean, organized warehouse with no stray obstacles, but if your workers are human and make mistakes, that’s a recipe for crashes. We also don’t use radar—wait, radar is good for cars, but for tight warehouse spaces, it’s too slow. It takes radar 200 milliseconds to detect an obstacle, which is way too long for an AMR moving at 5mph in a 10ft aisle. So we stick with the mix of LiDAR, cameras, and edge sensors because they’re fast, accurate, and built for small spaces.

Another thing people ask: what about moving obstacles? Like, another AMR, or a forklift coming down the aisle. The 360-degree LiDAR tracks moving objects, and the AI predicts where they’ll be in the next second. So if another AMR is 10ft away moving toward it, our Pallet Fork AMR will either stop in its lane, or slow down and pull over into a small empty space, so both can keep moving. We programmed that to avoid traffic jams, which is a big problem in warehouses with multiple AMRs.

Look, I get it—obstacle detection isn’t the most exciting part of a Pallet Fork AMR. You care more about how fast it can move pallets, how much it saves on labor, how long the battery lasts. But if the obstacle detection is garbage, all that other stuff doesn’t matter. A $50k AMR that crashes into things every week is a waste of money, but a $70k AMR that never stops, never crashes, and keeps your warehouse running smooth? That’s an investment that pays for itself in 6 months.

If you’re tired of chasing your old Pallet Fork AMR around the warehouse, dealing with repair bills from crashes, or watching productivity drop because of constant stops, we can help. We’ve customized our obstacle detection system for every type of warehouse—cold storage, e-commerce fulfillment, manufacturing, even smaller warehouses with narrow aisles. We can do a free on-site assessment to see how our Pallet Fork AMRs would work in your space, no pressure. Just reach out to us, and we’ll walk you through everything, no fancy jargon, just straight talk about what works for your team.

(ECS) Electronic Control System References:

  1. Robotics Industries Association. (2022). "Safety Standards for Industrial Mobile Robots"
  2. International Federation of Robotics. (2023). "Pallet Fork AMR Deployment Case Studies"
  3. Journal of Field Robotics. (2021). "Obstacle Detection for Indoor Industrial Robots: LiDAR vs. Camera Integration"
  4. Warehouse Innovation Magazine. (2024). "Overcoming Common AMR Obstacle Challenges in Mixed Workforce Environments"

Deliecn
DELIECN is one of the most reliable pallet fork amr manufacturers and suppliers in China, our products have good reputation in the market. Please rest assured to wholesale high quality pallet fork amr at competitive price from our factory. Contact us for more details.
Address: Room 602, Building G2, Artificial Intelligence Industrial Park, No.88 Jinjihu Avenue, Suzhou Industrial Park, Suzhou City, China
E-mail: marketing@delilogitech.com
WebSite: https://www.deliecn.com/