POMDP Approach to Robotized Clothes Sorting

Pol Monsó, G. Alenyà, A. Peñate-Sánchez, E. Serradell, C. Torras

The object manipulation with robots has mainly relied on precise, expensive models and deterministic executions. Given the great complexity of modeling deformable objects accurately, their manipulation remains an open research challenge. This thesis proposes a probabilistic approach to deformable object manipulation based on Partially Observed Markov Decision Processes (POMDP) where the action and perception deficiencies are compensated through interaction planning that efficiently leads to uncertainty reduction. Hence, we will prove that it is possible to achieve proposed goals through a simple, inexpensive set of actions and perceptions. The results of the thesis have been applied to a cloth sorting task in a real case scenario with a depth and color sensor and a robotic arm.

POMDP-based planning with many objects

POMDP-based common planning with three objects

Observation Model Dataset Download