An optimal control approach for exploratory actions in active tactile object recognition

In this paper, we treat the active tactile object recognition problem using an anthropomorphic robotic hand. Regarding the exploratory action design, to avoid such undesirable situations that the robot might break the object or might get a damage, the compliance of the robot behaviors is important as well as the informativeness of the resulting sensor data. However, most previous studies cannot consider these characteristics simultaneously since they treat the planning problem of exploratory actions separately from the robot control problem. We propose to design the exploratory actions using the formulation of the optimal control problem with the robot dynamics. Our cost function is composed of two terms: the informativeness and the energy consumption that can promote resulting actions to be compliant. The effectiveness of the proposed method is validated for the task of tactile object recognition through physical simulations and experiments in real environment.