Case Study: Nara Institute of Science and Technology achieves automated tactile object recognition and faster development with MathWorks' MATLAB and Robotics System Toolbox

Nara Institute of Science and Technology Researchers Develop Tactile Object Recognition Algorithms for a Robotic Hand

Researchers at Nara Institute of Science and Technology set out to give a ROS-enabled Shadow Dexterous Hand the ability to recognize objects by touch—using pressure, vibration, and temperature—rather than vision. To avoid lengthy C++ development and speed up testing, the team used MathWorks products, primarily MATLAB and Robotics System Toolbox, alongside toolboxes for optimization and parallel computing to connect their algorithms directly to the robot and focus on research.

Using MathWorks’ MATLAB with Robotics System Toolbox, Global Optimization Toolbox, Optimization Toolbox, and Parallel Computing Toolbox, NAIST developed active-learning tactile recognition algorithms (based on a Gaussian Process Latent Variable Model), simulated them, and then executed them on the physical robot via ROS. The MathWorks solution automated a previously 24-hour, manually intensive workflow—eliminating hundreds of manual steps—enabled the system to identify 10 objects (with plans for 100+), increased opportunities to try new algorithms, and made it easier to share expertise with students and other researchers.

Nara Institute of Science and Technology

Daisuke Tanaka

Researcher and Ph.D. Candidate