Stanford Seminar - Improving Robotic Dexterity with Optical Tactile Sensor DenseTact
03 Feb 2024 · 2 years ago

DanceTech Sensor and Compliant Gripper
- The DanceTech sensor is a soft optical tactile sensor that can estimate the shape and force of grasped objects.
- A compliant gripper using the DanceTech sensor enables stable grasping of small objects in cluttered environments.
- An interfinger manipulation controller is developed using tactile feedback from the DanceTech sensor, allowing for in-hand orientation of grasped objects.
T-Tech Mini Sensor
- The T-Tech Mini sensor is a new sensor for grasping and classifying small objects, including those as small as 1 mm, from flat surfaces.
- The T-Tech Mini sensor features a smaller size, a faster frame rate camera, and an attachable/detachable fingernail design, allowing for various fingertip materials to be used.
- The sensor utilizes three different strategies for grasping: T-grasping, fingernail grasping, and fingertip grasping.
- The T-Tech Mini sensor achieved a high grasping success rate of over 90% for various objects.
Object Classification
- The final step of object classification is achieved by estimating the depth from the DanceTech sensor and using DB scan to clusterize the detected objects.
Key Takeaways
- The T-Tech Mini sensor can grasp, classify, and reorient objects in controlled environments by upgrading hardware and utilizing surface interaction forces and fingernail designs.
- The impact that occurs when grasping an object can be used to estimate the force of the sensor.
- Historical input is important for complicated motions as it provides more information about the object's condition.
- A single camera is used for detecting the object.
- The distance between the operator and the object can be estimated using the best camera.
- The orientation of the tiny object can be done with the simple Jai controller by setting the state of the controller and selecting the object as a minimizing the data AB from the tech feedback.
Browse more from
Robotics

Stanford Webinar - Human-Robot Interaction

Stanford Seminar - Robots that aren't afraid of contact, an embodied approach

Stanford Seminar - Flying Robots: Exploring Hybrid Locomotion and Physical Interaction

Marc Raibert: Boston Dynamics and the Future of Robotics | Lex Fridman Podcast #412

Stanford Seminar - Embodied Intelligence for Extreme Environments

Stanford Seminar - Robot Skill Acquisition: Policy Representation and Data Generation
Ready to get started?
Save, summarize & chat with your content.
GET STARTED
IT'S FREE
No credit card required · 30 Day Refund on Premium · 24 Hour Support
