Tactile recognition and localization using object models: the case of polyhedra on a plane

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Object recognition and localization are crucial in robotics.
• Tactile sensing offers rich information for object manipulation.
• Existing methods often require global information or complex feature extraction.

Purpose of the Study:

• To develop a method for object identification and localization using only local tactile sensor data.
• To enable robots to recognize and determine the position of objects from a known set.
• To leverage contact point and surface normal information for robust perception.

Main Methods:

• Utilizing local data from multiple tactile sensors, including contact point positions and surface normal ranges.
• Structuring the recognition and localization as a hypothesis generation and pruning process.
• Employing a tree-based search to manage consistent pairings between sensor contacts and object surfaces.

Main Results:

• The proposed algorithm demonstrates the feasibility of object recognition and localization using limited tactile information.
• Simulations show the effectiveness of the hypothesis pruning approach.
• The method is validated for polyhedral objects with three degrees of freedom on a known plane.

Conclusions:

• Local tactile sensor data is sufficient for effective object identification and localization.
• The hypothesis-driven approach provides a structured framework for tactile perception.
• This method contributes to more capable and adaptable robotic systems.