Traditionally, the creation of robotic handling applications, such as pick-and-place and assembly operations, has been a laborious task that demands the expertise of individuals to meticulously define each step of the robot’s sequences. This involves considering numerous factors, including the specific part to be handled, its environment, the required forces, the optimal path to follow, and the type of gripper to use. Unfortunately, this process is not only time-consuming but also costly, making it unattainable for many industries dealing with a diverse range of products and relatively low production volumes.

However, the advent of Artificial Intelligence (AI) opens up the possibility of radically changing this process. By combining the power of AI with simulation tools, HARTU is studying an approach to create robotic models that can autonomously learn how to handle previously unseen objects without causing damage or risking them falling. Furthermore, these models, based on Deep Reinforcement Learning, can even determine the optimal part to manipulate within cluttered scenes featuring multiple items.

Likewise, the process of configuring assembly operations can now be accelerated thanks to learning from demonstration techniques. In this innovative approach, a human operator demonstrates the assembly procedure, and the robot intelligently learns to replicate it, controlling not only the sequence but also the contact between the different parts being assembled.

This introduction of Artificial Intelligence and new robotic handling promises to reduce costs, enhance efficiency, and expand the horizons of automation across various industries. With the adoption of these transformation technologies, production lines can open up to the future where adaptability and precision will be a norm, supporting operators in the most complex practices and developing new jobs and skills.