Vision Guide Industry Robot for Random Bin Picking

Traditional automation involves robots that faithfully carry out specific actions repeatedly based on programs meticulously crafted by highly skilled engineers. These robots are inflexible and ill-fitted for production of goods with short product lifecycle. In high-productivity manufacturing line, parts detection and posture estimation must be performed in real-time.

• Data Generation Engine: Generate artificial AI training data through 3D engine. No manual labelling is required.
• Deep Learning: Extract object features by AI to train an appropriate model without the need of manual parameter tuning for new object.
• Adaptive Optical Design: Adopts coded phase shift light pattern and dual-polarized optical design, the system is automatically adapted to different production line environment & multiple measured objects.
• GPU Accelerated Point Cloud Processing: Apply GPU accelerated algorithm on object recognition to ensure speed (< 0.5 s 3D recognition time) and accuracy (0.1 mm @ 0.5m).

• Enhancement of the flexibility of manufacturing line by enabling efficient reconfiguration for multiple products.
• Increasing the throughput of the manufacturing line in low-cost.
• Expediting the upgrading of local industries with flexible cutting-edge 3D robot cognition to perform complex tasks in advanced manufacturing.

• Random Bin Picking
• Defect Inspection
• Assembly
• Autonomous robot navigation

• US Patent No. 11,287,626; CN App No. 201980000804.2
• US Patent No. 11,023,770; CN App No. 201980002023.7 and HK App No. 62020022213.6