Obstacle perception
Detect objects above/below the plane of a conventional 2D scanner and provide richer scene information for navigation logic.
ifm · robotic perception
ifm O3R Perception Platform
Centralise and synchronise multiple 2D/3D camera heads and additional sensor information for mobile robotics and multi-camera perception. O3R combines industrial camera heads with an edge-compute VPU so developers can run IFM functions or their own ROS, Docker, Python/C++ and CUDA-based workloads.
Up to 6 camera heads2D + 3D ToFYocto Linux + DockerROS / Python / C++ / CUDA
Camera headsUp to 6Centralised multi-camera processing
3D range0.2–4 mCurrent O3R head specification
3D FoV60° / 105°60×45 or 105×78
ComputeEdge VPUYocto Linux + NVIDIA platform
Use cases
Designed for multi-camera and multi-modal perception
O3R is not a single replacement for a safety scanner. It is a perception platform used to add richer environmental awareness and application functions around a robot or multi-camera system.
Pallet detection
Use 3D camera data and available application functions to support pallet engagement and positioning.
360° environmental awareness
Arrange several heads around the vehicle or machine and process their information centrally.
Position & environment
Fuse multiple perception inputs to support localisation and navigation development.
Multi-view dimensioning
Use several cameras for large or occluded objects such as pallets, packages or logs.
Own algorithms
Run custom perception code in familiar software environments on the VPU.
Safety note: O3R perception functions must not automatically be treated as certified personnel-protection functions. Functional-safety sensing must be selected and validated separately where required.
Hardware architecture
Camera heads + central VPU + optional external sensors
The O3R centralises camera and sensor processing. Camera placement becomes part of the perception design: choose head field of view and location around the required coverage, then size VPU and software architecture around the algorithms.
012D/3D heads
→
02O3R VPU
→
03Algorithms / sensor fusion
→
04Robot controller
VPUCentral processing unit with camera ports, Ethernet, USB, CAN and industrial 24 V power.
Camera heads2D/3D heads connect via dedicated camera interfaces and provide synchronised depth/RGB data.
Additional sensorsRadar, lidar, ultrasonic or other Ethernet sensors can complement the perception stack.
Edge computeOffload perception from the vehicle controller and run containerised algorithms close to the sensors.
Camera heads
Current O3R head specification
IFM offers different camera-head fields of view so close-range scene coverage can be adapted to vehicle geometry and the application.
2D + 3D in one head
O3R camera heads combine PMD time-of-flight depth sensing with RGB information for multimodal perception.
3D image sensorpmd 3D ToF
3D resolution224 × 172 pixels
3D field of view60° × 45° / 105° × 78°
3D light source940 nm infrared
2D resolution1280 × 800 pixels
Range0.2…4 m
Max frame rate20 fps
Open development
The platform as an industrial perception computer
The VPU runs Yocto Linux and a Docker architecture. IFM documents support for development environments such as Python, C++, CUDA and ROS, letting developers combine IFM functions with their own perception stack.
Docker
Package and deploy application components in familiar container workflows.
ROS
Integrate into common robotics middleware and development stacks.
CUDA / NVIDIA
Use GPU-capable processing for suitable perception workloads.
Developer resources
Use IFM developer resources for protocol, code examples and integration support.
Project fit
Define coverage and algorithms before choosing hardware
O3R selection is a system-engineering task. Camera count is not the first question; define the blind spots, required depth accuracy, vehicle speed, compute load and robot interfaces first.
Robot / machineAMR, AGV or stationary multi-camera use case.
CoverageRequired zones, blind spots and mounting locations.
Detection taskObstacle, pallet, navigation, dimensioning or custom perception.
EnvironmentIndoor/outdoor light, floor, reflective/transparent objects, dust and contamination.
SoftwareIFM application function versus custom ROS/Docker/C++/Python workload.
InterfacesVehicle controller, Ethernet, CAN, other sensors and required data rate.