Soloctrix SXIVE logo banner image RealSense
RealSense » Case Studies » Scaling Rugged Vision for Mobile Industrial Machines

Scaling Rugged Vision for Mobile Industrial Machines


Solectrix’s SXVPU and the RealSense Depth Camera D457 deliver low-latency, edge AI perception for forklifts, robots, and industrial vehicles in harsh environments.

Download Case Study

The challenge

Today’s mobile machines operate in environments where failure isn’t an option. Dust, moisture, vibration, temperature swings, and shock are normal operating conditions, not edge cases. Perception systems must detect people, obstacles, and hazards in milliseconds to support emergency stops, speed reductions, or virtual safety zones.

Cameras may be mounted far from the compute unit — on masts, rear bumpers, booms, or articulated arms — driving stringent requirements for long cable runs, robust connectors, and EMI resilience. At the same time, OEMs need architectures they can reuse across fleets of forklifts, industrial trucks, agricultural machines, and construction vehicles without redesigning the perception stack for every model.

Solectrix needed a vision platform that combined:

  • Low-latency, high-resolution processing and AI at the edge
  • Rugged, cost-effective cameras that can be placed wherever safety demands
  • Reliable, automotive-grade connectivity over long distances

The solution

Solectrix designed the SXVPU as a semi-centralized vision subsystem. Instead of putting full intelligence into every camera or routing all raw data into a monolithic ECU, the SXVPU concentrates image processing, AI, and sensor fusion into a rugged, compact box while keeping cameras straightforward and robust.

On the sensor side, Solectrix selected the RealSense Depth Camera D457 as a key building block for industrial 3D perception.

The results

For Solectrix, pairing SXVPU with RealSense depth cameras is about building a scalable, rugged perception platform that customers can rely on for years to come—one that helps them move step by step from today’s assistance systems toward safer, more capable, AI-enabled machines. 

“RealSense depth cameras and SXVPU give us a robust core to build on,” Stefan Schütz concluded. “From here, we can help our customers scale from assisted operation today to more autonomous, AI-powered machines tomorrow.”

SXVPU: Zonal Vision“Brain” for Mobile Machines

The SXVPU combines an AMD Kria™ SOM and FPGA fabric with a dedicated AI accelerator, delivering tens of TOPS of performance for neural network inference and high-throughput image processing. The data path is tuned to keep end-to-end latency under 10 ms in typical configurations — a critical threshold for safety-related functions like collision avoidance, speed limiting, and zone monitoring.

Supporting up to four GMSL/FAKRA RealSense D457 cameras in its standard configuration—and scalable to as many as eight cameras in extended setups—plus Ethernet, USB, GPIO, and CAN, a single SXVPU can aggregate multiple viewpoints, fuse additional sensors such as radar or ultrasonic inputs, and stream results to the vehicle controller, HMI, or backend systems.

Rugged 3D Vision over GMSL/FAKRA

The D457 is RealSense’s first GMSL/FAKRA depth camera, built specifically for long cable runs and automotive-style locking connectors. Its IP65-rated enclosure is dust-tight and protected against water jets, making it suitable for both indoor industrial and outdoor mobile robotics.

Why

it Matters

Safer, Smarter, More Scalable Machines

By standardizing on the SXVPU and RealSense D457, Solectrix and its customers gain:

  • Real-time safety – Sub-10 ms perception pipelines support rapid reactions for collision avoidance, driver assistance, and virtual safety zones, powered by high-quality depth and global-shutter RGB even in dynamic scenes.
  • Rugged operation – IP65/67-level protection on the SXVPU and IP65 on the D457, plus automotive-grade GMSL/FAKRA cabling, support 24/7 operation in dust, moisture, vibration, and temperature extremes.
  • Lower system complexity and cost – A single SXVPU can replace multiple smart cameras and a separate switch, while using robust but straightforward cameras instead of fully embedded vision sensors.
  • Flexibility across platforms – The same architecture can be reused for forklifts and industrial trucks, construction machinery, agricultural equipment, mobile robots, and AMRs, shortening the path from prototype to series production.

“We can apply one scalable building block to many different machine platforms,” Schütz said. “That’s powerful for our customers — they get consistent performance and a shorter path from prototype to series production.”

— Stefan Schütz,
CEO of Solectrix GmbH

What’s

next?

Learn how RightHand Robotics is using RealSense™ computer vision technology to revolutionize automated warehouse order fulfillment.

Robotics

overview

See how RealSense™ cameras can revolutionize the robotics industry with the latest in depth‑sensing technology.