Technology 01
Multi-sensor fusion — LiDAR, depth camera, ultrasonic — for mapping, localization, and obstacle avoidance around people, racking, and narrow routes. Shared across the cleaning, AMR, and outdoor robot lines.
Capabilities
Sensor comparison
Different sites need different sensing. A quiet office corridor, a forklift aisle and a public road each fail in different ways, so the sensor suite is chosen per product family rather than one stack for everything.
| Model | Primary navigation | Obstacle and safety layers | Typical environment |
|---|---|---|---|
| ACR-0290 | Multi-camera vision, one-key mapping | Front, side and top cameras; ranging sensors; IR illuminators; 6 cliff sensors; floor-material sensor | Shops, offices, restaurants |
| ACR-0440C | 2D LiDAR (3D optional) | 2D obstacle radar, ultrasonic radar, anti-drop sensors, HD camera | Indoor commercial floors |
| ACR-0370 | 2D single-line or 3D multi-line LiDAR | 21 sensors for 360° protection | Workshops, hospitals, metro stations |
| ACR-0520P | 3D LiDAR | Wide-angle depth camera, AI vision for 20+ object and stain classes | Malls, office towers, transport hubs |
| AMR-0300 | VSLAM + LiDAR SLAM fusion | Built to ISO 3691-4 | Factories and warehouses |
| AMR-0600R | Three LiDARs (360°, 0.1–40 m) | 215° RGB camera, optional depth camera, front and rear bumpers | Heavy rack transfer, campus outdoor |
| ACR-1380 | 3D LiDAR + 2 single-line LiDARs | 4 × 3D cameras, 4 surveillance cameras, IMU, bumper strip | Plazas, parks, roads |
| ACR-1330 | LiDAR with 254 TOPS on-board computing | Surround-view cameras and AEB automatic emergency braking | Patrol sweeping on public routes |
Sensor lists from each product page. Exact sensor models and firmware versions are confirmed in the quotation datasheet.
How autonomy works on site
Every deployment goes through the same sequence. Knowing it helps facility teams plan the time and access the commissioning engineer needs.
The robot is driven or walked through the site once and builds a map from LiDAR and camera data. The ACR-1380 commissions a new site in about 10 minutes.
Cleaning zones, delivery points, no-go areas and speed limits are drawn on the map in software, not marked on the floor.
During work the robot matches live sensor data to the map to know where it is, even when shelves, pallets or furniture have moved.
Layered sensors detect people, pallets, glass, low objects and drops. The robot slows, re-plans or stops depending on distance and object type.
Repeatable docking to charging, water and conveyor stations lets robots refill, drain or hand over loads without help.
Design choices
Each sensor type has blind spots. Fusion means one layer covers what another misses.
Accurate distance and mapping in any light, but can miss glass, very low objects and black surfaces.
Recognise objects, stains and people and see low obstacles, but need light and are affected by glare.
Detect glass and close obstacles that optical sensors miss; short range only.
Last-line protection against drops, stairs and contact.
Tracks motion and tilt between sensor updates, which matters on ramps and uneven outdoor surfaces.
AI recognition runs on the robot, so decisions do not wait for a network round trip.
FAQ
No. Robots navigate from a map built with LiDAR and cameras. Routes and zones are edited in software, so layout changes do not require floor markers.
Yes. Layered sensors detect people and obstacles at different heights, and the robots slow down, re-plan or stop. Outdoor models such as the ACR-1330 add automatic emergency braking.
Localisation matches live sensor data to the stored map, so moved objects are treated as obstacles rather than causing the robot to get lost. Large permanent changes are handled by updating the map.
Glass is hard for LiDAR alone, which is why indoor models add ultrasonic, radar or depth-camera layers. Large glass areas can also be marked as virtual walls during commissioning.
It depends on size and complexity. The ACR-1380 outdoor sweeper commissions a new site in about 10 minutes; indoor buildings with many zones take longer because routes and schedules are set up as well.
The AMR-0600 supports the VDA5050 protocol, which is designed for mixed-vendor fleets. Other models connect through our dispatch platform and APIs.