A roof- and end-of-bay LiDAR cluster tracks bridge, trolley and hook position in real time and writes slow-down and stop interlocks into your crane PLC before the collision happens — with structure, with stockpile, with neighbouring cranes. OPC UA, Profinet or hardwired DI, vendor-agnostic and built for your bay.
LiDAR crane anti-collision is portal-mounted perception software that tracks the position of industrial cranes (bridge, trolley, hook) in real time and drives interlocks into the crane PLC before collisions occur with structure, stockpile, or neighbouring cranes.
The pain is the same in every bay where a crane shares its rails with something it can damage. A misjudged grab descent puts the hook into a stockpile shoulder. A tandem-crane bay loses awareness of who is where and the two bridges close on each other. End-stops get hit at speed because the operator was looking somewhere else, or because the bay was operated semi-automatically and the position estimate drifted. Each event is an hour of downtime at best, a wrecked grab or a safety incident at worst, and an insurance conversation either way.
Our approach is a LiDAR cluster — typically two to four sensors at the end-of-bay portals and on the roof — fused into a single redundant position estimate of the crane bridge, the trolley and the hook. Safety bands are defined around structure, walls, stockpiles and (where applicable) the other crane. When the position estimate enters a band, the system writes a slow-down or stop signal into the crane PLC over OPC UA, Profinet or hardwired DI — within a control cycle. Redundant sensors mean a single failure does not silently disable the safety function. The productised flavour of this for waste bunkers already runs in the field — see our OWL EYE® waste-bunker product — but every steel mill, paper mill or scrap yard has its own bay geometry, its own crane vendor and its own existing safety architecture, and that is what the custom build is for.
This is not an off-the-shelf box. Every engagement descends from our hub service Industrial Perception AI as a discovery + fixed-scope build — typically four to twelve weeks from contract to live interlock, with code, model parameters and integration documentation handed over to you at the end.
Same team and same stack that delivers our LiDAR wagon-detection pipeline and the productised OWL EYE® waste-bunker anti-collision system.
Two to four industrial LiDAR sensors mounted at end-of-bay portals and on the roof, covering the bridge rails, the trolley travel and the hook envelope. IP65+ housings, dust-tolerant optics, runs in steam, dust and night light — the kind of conditions where steel-mill, paper-mill and waste-bunker bays actually live. Sensor placement is engineered for your bay, not a generic template.
A real-time fusion stage takes raw point clouds from the cluster and returns a single redundant position estimate for the bridge, the trolley and the hook — typically under 5 cm error and updated at 10–25 Hz. Redundancy means each axis is observed by at least two sensors; one sensor failure degrades accuracy, it does not silently disable the safety logic.
The fused position drives configurable safety bands around structure, stockpile and neighbouring cranes. When a band is entered, a slow-down or stop signal is written into the crane PLC over OPC UA, Profinet or hardwired DI — within a control cycle. Every interlock event is logged with timestamp, sensor snapshot and position track for the safety review.
We keep the architecture boring on purpose. Three loosely coupled stages, each one independently testable, each one swappable when the sensor stack or the crane vendor changes. Built on the same stack we use across all our perception work: PCL, Open3D, OpenCV, PyTorch.
The cluster — typically two to four LiDAR per bay — delivers synchronised 3D point clouds covering the bridge rails, the trolley path and the hook envelope. Sensors are time-synchronised at the network layer; raw frames are stamped and locally buffered for the audit trail and for post-incident replay.
Each frame goes through ground removal, voxel down-sampling and crane-component segmentation; the bridge, the trolley and the hook are extracted and tracked over time. A Kalman-style fusion stage produces a single redundant position estimate per axis — typically <5 cm error, updated at 10–25 Hz — and evaluates it against the configured safety bands around structure, stockpile and neighbouring cranes.
When a safety band is entered, the system writes a slow-down or stop signal into the crane PLC over OPC UA, Profinet or hardwired DI — within a control cycle. Where the plant runs a safety PLC, the integration is FS-aware and can be wired into the existing safety chain. Every event is logged with timestamp, sensor snapshot and position track.
All three stages run on an industrial PC in the crane-cabinet or control-room rack. No cloud dependency, no external API, no licence dial-home. The code is yours at handover.
Configurable slow-down and stop signals written into the crane PLC over OPC UA, Profinet or hardwired DI. Safety bands are defined per obstacle — structure, walls, stockpiles, tandem-bridge buffer — and tuned together with the crane operators and the safety engineer.
A continuous bridge / trolley / hook coordinate stream into your historian, MES or yard-management system — typical update rate 10–25 Hz. The same data we use to drive interlocks is available for production analytics, semi-automated grab routing, or feeding into our stockpile-monitoring pipeline.
A control-room display showing live position, active safety bands and recent interlock events, plus a post-incident replay tool — sensor snapshots, position tracks and PLC outputs aligned on a single timeline. Built for the way crane operators and safety engineers actually do reviews.
Every crane bay has its own obstacles, its own crane vendor and its own existing safety architecture. A generic anti-collision box solves the generic case; your bay is not the generic case.
Every bay has different obstacles — pillars, walls, stockpile shoulders, neighbouring cranes, end-stops at non-standard distances. We survey the bay, place the sensor cluster for the geometry you actually have, and tune safety bands together with the operators and the safety engineer. We re-calibrate when the bay layout changes.
The interlock layer talks OPC UA, Profinet and hardwired DI, so it works with Demag, ABUS, Konecranes, Stahl, Kone, ZPMC and the older nameplates nobody quite remembers — and with mixed fleets in the same bay. The perception layer doesn't care which vendor built the crane.
Where the plant runs a safety PLC, the integration is FS-aware: the LiDAR layer provides the position estimate, the safety PLC owns the trip. Where the bay runs a standard PLC only, we wire a hardwired DI shutdown alongside the OPC UA path so the trip survives a network outage. We work with your safety engineer, not around them.
You own the source code, the calibration data and the integration documentation at handover. We document the system, train your crane and safety teams, and walk away clean. No black box, no monthly per-crane licence, no service contract you can't exit. See our FAQs on IP and engagement model for the standard terms.
The engagement-model questions we hear from every plant considering a custom anti-collision build. Need something more specific to your bay and your crane vendor? Ask us.
Bay layout, crane vendor, PLC type, the obstacles you need to protect against, any existing safety architecture — anything you have. We come back within two business days with an honest first assessment and a fixed-price scope for the Discovery workshop.