Power Cable Bin Picking using UR5e, Deep Learning, and Reinforcement Learning

This solution automates product sorting, assembly, and packing by utilizing 3D Vision, Deep Learning (Instance Segmentation), Reinforcement Learning (RL), and Motion Planning (OMPL) algorithms.

By using 3D vision to recognize product types, product locations, and box positions in real-time, the system can automatically compensate for changes in positioning. The integration of deep learning and reinforcement learning for object identification allows the solution to be applied even to irregularly shaped objects.

Because the system calibrates itself through continuous learning—even when conditions change, such as vision/robot misalignment or switching to different products—it ensures easy operation, simplified maintenance, and minimized downtime.

Components

Robot	Robot: UR5e 6-axis articulated robot, 5kg payload, 850mm reach, 20.6kg weight.

Workflow

STEP 1.	Input of workpieces.
STEP 2.	Optimal picking planning for the workpieces.
STEP 3.	Picking and aligning the workpieces.
STEP 4.	Aligning and stacking the pieces into the discharge box.

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

Key Features

Effortless Robot Teaching

Automated Teaching: Reduces teaching time by automating vision calibration and teaching processes through deep learning.
Condition Flexibility: No manual recalibration or re-teaching required even when operating conditions change.
Superior Recognition: High recognition capability for diverse objects using deep learning-based 3D object perception.
Self-Learning Optimization: Automatically learns the optimal pick position according to the product shape.
Rapid Training: Utilizes simulation data to significantly shorten the AI training period.
Active Picking: If an object is in an unsuitable state for gripping, the robot can knock it over or move it to enable successful picking.
High-Speed Detection: Rapid 6D pose estimation (object posture recognition) in under 0.2 seconds.
Quick Installation: Applicable to non-standard, irregular objects, reducing initial setup time.

Compact and Simple Installation

No Specialized Feeders: Eliminates the need for separate feeders or fixed positioning for input boxes.
Space Efficiency: The robot can be added directly to existing manual workspaces without extensive layout changes.

Rapid ROI - Operational Optimization and Reduced Maintenance Costs

Cycle Time Reduction: Continuously optimizes work postures and paths through persistent learning.
Automatic Compensation: Performs automatic correction and continues operations even if the layout shifts or the robot's accuracy deviates due to aging.
Consistent Productivity: Ensures stable production by minimizing system downtime.

Implementation Results

ROI Period
Key Metrics
Client Feedback

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

Please log in to view

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

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

Components

Workflow

Key Features

Implementation Results

Power Cable Bin Picking using UR5e, Deep Learning, and Reinforcement Learning

Recommended Use Cases