VISION-LANGUAGE-ACTION MODELS FOR ROBOTIC MANIPULATION OF DEFORMABLE OBJECTS

The research activity aims to study and develop Vision-Language-Action (VLA) models to enable robots to manipulate deformable objects, such as fabrics, clothing, ropes, and flexible materials, in complex domestic environments. The work will focus on integrating visual perception, understanding natural language instructions, and generating robotic actions, with the goal of improving the robots' ability to perform robust and generalizable manipulation tasks. Particular attention will be paid to adapting and evaluating VLA models in scenarios characterized by uncertainty, complex deformations, and prolonged physical interactions with the environment.