A Bayesian Framework for Active Tactile Object Recognition, Pose Estimation and Shape Transfer Learning

Researchers developed a Bayesian framework combining particle filters and Gaussian processes for robotic tactile object recognition and pose estimation. The system can identify known objects, detect novel objects, and transfer knowledge to learn new shapes through active touch exploration.

• →A unified Bayesian framework combines particle filters and Gaussian process implicit surfaces for tactile object recognition.
• →The system can simultaneously estimate object class and pose while detecting novel objects through active touch.
• →Knowledge transfer capabilities allow the framework to learn new shapes by grounding prior knowledge from known objects.
• →An exploration procedure guides active data acquisition and automatically terminates when sufficient information is gathered.
• →Simulation experiments demonstrate effectiveness in recognizing known objects and learning novel shapes reliably.