Vention has introduced GRIIP, short for Generalized Robotic Industrial Intelligence Pipeline, an end-to-end physical artificial intelligence system designed to support autonomous robot cell deployment in unstructured manufacturing environments.
The system integrates Vention’s proprietary models with open models from NVIDIA, including NVIDIA FoundationStereo for stereo matching and NVIDIA FoundationPose for pose estimation. According to the company, the pipeline spans scene digitalization and calibration, object detection and segmentation, six-degree-of-freedom pose estimation, grasp point evaluation, and collision-free motion planning. The architecture is designed to adapt to changing production conditions without manual configuration.
GRIIP operates on Vention’s MachineMotion AI controller, powered by NVIDIA Jetson, and supports over-the-air updates via WiFi or built-in LTE connectivity. The company states that the system can incorporate updated physical AI models over time without requiring hardware changes.
Vention reports that the platform sustained autonomous 24/7 lights-out operation for three months while maintaining throughput of up to five parts per minute. The company also cites sub-millimeter pose estimation accuracy and indicates that cycle times remained stable throughout the operating period.
The pipeline is designed to function without task-specific programming or custom training datasets. Vention states that CAD-to-pick setup can be completed in approximately 15 minutes, with full robot cell deployment in under two days.
GRIIP is intended to support multiple manufacturing applications using a single generalized architecture, including bin picking, machine tending, conveyor pick-and-place, kitting, depalletizing, and sanding. The company says the platform can also be used to convert traditionally programmed robotic applications into autonomous operations within existing production environments.
