AI-Powered 3D Delta Robot for Pick & Place
- Client
- Factory automation & robotics
- Year
- 2025
- Category
- Industrial Automation
Delta robots are ideal for fast, precise part handling, but without vision they only work with pre-defined positions and expensive mechanical fixtures. NELSco built an AI-powered 3D delta robot that, by fusing an RGB-D depth camera with deep learning, identifies objects in any random orientation on the conveyor and picks them with no fixtures required.
An Intel RealSense depth camera for full RGB-D data and 3D object localization, plus instance/semantic segmentation with TensorFlow and Keras for pixel-level boundary detection and reliable grasping of irregular shapes.
A custom inverse-kinematics controller computing exact motor angles from the 3D grasp pose, a custom pneumatic end-effector tailored to the product geometry, and a retrainable AI model for new products.
The Challenge
The robot had to recognize irregularly shaped objects in fully random position and orientation on the belt, find their precise boundaries for a reliable grasp, and translate the 3D grasp coordinates into exact motor angles, all within a time budget compatible with line speed and without relying on expensive mechanical fixtures.
Our Approach
An Intel RealSense depth camera provides full RGB-D point-cloud data (X, Y, Z object localization). Instance/semantic image segmentation with TensorFlow and Keras detects object boundaries at the pixel level for reliable grasping of irregular shapes. A custom inverse-kinematics controller calculates the exact motor angles from the 3D grasp pose, and a custom mechanical/pneumatic end-effector tailored to the target product geometry was built. The AI model is retrainable, so the system can be adapted to new product types.
The Result
The result was a flexible automation cell that removes the need for mechanical alignment fixtures and handles randomly oriented parts, increasing production-line flexibility and throughput: machine vision, AI, and robotics combined for fixture-free picking.
Results
- Fixture-free picking of objects in fully random orientation on the belt
- Eliminates the cost of mechanical alignment fixtures
- Increases line flexibility and throughput by handling randomly oriented parts
Have a similar project?
Get in touch with NELSco's engineering team to discuss your project requirements.