Summary of the solution
e-con Systems collaborated with a leading Physical AI company to develop a custom RGB-D vision platform for high-volume robotic manipulation. The solution enabled synchronized depth and color capture, multi-camera operation, GMSL-to-PCIe streaming, SDK optimization, and cleaner 3D data. This empowered the client to improve robotic perception, depth accuracy, and long-term deployment stability.
Key challenges faced by the client
Off-the-shelf RGB-D cameras couldn’t meet depth accuracy and noise requirements
High CPU load and slow data transfer limited real-time depth processing
Multi-camera deployments
needed
synchronization for
up to 8 cameras
Real-world warehouse conditions required
on-the-fly tuning
and depth enhancement
Benefits of the solution
- Delivered a custom RGB-D camera board with a ToF depth sensor, an RGB image sensor with ISP, FPGA signal processing, GMSL serializer, and onboard IMU
- Achieved reliable depth performance with ±5 mm depth deviation from 0.5 m to 4 m for object detection, localization, and manipulation
- Supported synchronized operation for up to 8 cameras using a GMSL-to-PCIe Frame Grabber board and Quad FAKRA cable architecture
- Reduced CPU load and improved data transfer through SDK optimization, helping the robotics platform process depth data in real time
- Improved 3D scene quality through point-cloud noise reduction, giving the robot cleaner spatial data for manipulation tasks
- Strengthened long-term deployment stability with firmware hardening, memory stability, zero runtime allocation, and post-release software support


Mr. Thomas Yoon
+82-10-5380-0313










