Product Sorting Automation using UR5 and Deep Learning

Industry > LogisticsApplication > ClassificationSector > Manufacturing

This next-generation sorting solution leverages 3D Vision, Deep Learning (Instance Segmentation), Reinforcement Learning, and Motion Planning (OMPL) to automate complex tasks such as product classification, assembly, and packing.

By utilizing cost-effective 3D vision, the system recognizes the type and location of products and bins in real-time. This allows the robot to automatically compensate for shifts in the position of items or boxes without manual intervention. The integration of Motion Planning drastically reduces human teaching time, while Reinforcement Learning optimizes movements to minimize overall cycle times.

Components

Robot	Robot: Universal Robots UR5e Specs: 6-axis articulated, 5kg payload, 850mm reach, 20.6kg weight. Alternative Options: Also compatible with Robostar 6-axis robots (payloads ranging from 1kg to 50kg). AI Vision System: High-performance 3D Vision for real-time spatial and object recognition. Software Algorithms: Instance Segmentation: Distinguishes individual items even when overlapping or mixed. Reinforcement Learning: Continuous optimization of the Pick & Place cycle. Motion Planning (OMPL): Autonomous path generation to avoid obstacles and reduce downtime.

Robot

Robot: Universal Robots UR5e
- Specs: 6-axis articulated, 5kg payload, 850mm reach, 20.6kg weight.
- Alternative Options: Also compatible with Robostar 6-axis robots (payloads ranging from 1kg to 50kg).
AI Vision System: High-performance 3D Vision for real-time spatial and object recognition.
Software Algorithms:
- Instance Segmentation: Distinguishes individual items even when overlapping or mixed.
- Reinforcement Learning: Continuous optimization of the Pick & Place cycle.
- Motion Planning (OMPL): Autonomous path generation to avoid obstacles and reduce downtime.

Workflow

STEP 1.	Recognition: The system identifies and classifies individual items from a "random bin" containing various mixed products.
STEP 2.	Sorting & Stacking: The robot picks the recognized item and places it into the designated discharge box in a precise position.
STEP 3.	Continuous Operation: The robot repeats the task until all items in the input box are sorted. (The vision system automatically tracks and updates Pick & Place coordinates even if the boxes are moved during operation.)

※ The content provided by Bigwave Robotics is protected under U.S. and Canadian copyright and intellectual property laws. Unauthorized reproduction, distribution, or use of this content is strictly prohibited.

Features

Simple & Effortless Setup

Automated Robot Teaching: Deep learning-based vision calibration and automated teaching significantly reduce initial setup time.

Calibration-Free Operation: No need for manual re-calibration or re-teaching even when work conditions or environmental factors change.

Advanced 3D Object Recognition: Superior recognition capabilities for diverse objects using deep learning-based 3D vision.

Self-Learning Pick Optimization: The system automatically learns the optimal picking coordinates based on the specific shape and orientation of each product.

Infrastructure Minimalism: Eliminates the need for fixed-position jigs or custom feeders, as the system dynamically recognizes input and discharge boxes.

Seamless Integration: The robot can be installed directly into existing manual workspaces without extensive line modifications.

Rapid ROI (Return on Investment)

Operational Optimization: Continuous learning algorithms optimize robot postures and movement paths to consistently shorten cycle times.

Low Maintenance Costs: Automated compensation ensures stable operation even if the layout shifts or the robot experiences mechanical wear over time.

Maximum Productivity: Ensures steady production output by minimizing downtime through robust AI-driven self-correction.

Results