Alessandro Amici

The research tackles the complex challenge of dexterous multi-robot manipulation of delicate, soft entities, specifically insects. Deformable Object Manipulation (DOM) is difficult due to the softness and indefinite shapes of objects like biological tissues, soft polymers, seeds, and insects. Existing methods focus on manipulating a single type of known object, whereas this research aims to handle a variety of insects with different physical properties.

The goal is to develop robotic systems capable of detecting and adapting to changes in stiffness and deformation during manipulation. The research will explore the use of control algorithms and machine learning techniques to ensure stable and precise handling of soft objects with unpredictable deformations. Additionally, it will investigate how multiple robots can collaborate in manipulating these objects, and how to simulate and benchmark such tasks, considering factors like friction and elasticity.  

The approach will integrate design, sensing, modeling, planning, and control to address the challenges of DOM. Multi-robot systems with advanced sensing capabilities and specialized grippers will be employed, supported by algorithms for smooth handling. Initial experiments will take place in simulated environments, followed by real-world manipulation of insect-like objects. Techniques such as Deep Reinforcement Learning (DRL) and transfer learning will be explored to enable dexterous manipulation of soft, deformable entities.  

This research could greatly advance the discovery and identification of new insect species through fast, accurate and precise robotic manipulation. Additionally, it has the potential to make a significant impact on industries that require delicate and precise work, such as agriculture, biotechnology, and electronics by automating tedious and intricate tasks, thereby enhancing efficiency and precision.