Programming industrial robots is often time-consuming and requires specialised expertise. This is particularly challenging in flexible production, where companies manufacture small batches of varied products. At the same time, developments in AI, machine learning and robotic perception have created new opportunities for robots to adapt to different objects and production conditions. For these methods to be used in industry, however, AI models need to operate with high performance and predictability while meeting requirements for safety and reliability.
EDGE-SKILLS therefore focuses on AI-based robot programming for complex grasping tasks. Models are trained offline using real and simulated data and then optimised to run on industrial edge hardware. In this way, the project aims to bridge the gap between advanced AI methods and the requirements of real production environments.
Purpose
The purpose of EDGE-SKILLS is to enable edge-based, safe and predictable AI robot programming for complex grasping tasks, with increased industrial flexibility and efficiency.
Objectives
The project will:
- develop AI-based solutions for programming complex grasping and handling tasks
- use real and simulated data and reinforcement learning to train AI models
- compress and optimise the models for robust and reliable operation in industrial environments
- validate the solutions in production-near robotic cells
- develop guidelines for usability, robustness and industrial integration.
