Design and Development of Tactile Robot Hands for Enhanced Dexterous Manipulation

Abstract

Tactile robotic hands play a crucial role in enabling physical interaction between robots and their environment. These hands, equipped with tactile systems, deliver rich contact information, such as forces and object properties, to the robot's control system. This enables the precise and stable grasping of delicate objects and supports complex in-hand manipulation tasks.

Despite significant progress, the diversity in sensing principles, materials, multi-modalities, and interpretation methods among tactile sensors presents a major challenge: the absence of a unified classification framework for optical tactile sensors. To address this gap, a new classification method specifically for optical tactile sensors is proposed in Chapter 2. The proposed framework categorizes optical tactile sensors based on their sensing mechanisms and features. By establishing clear classifications, this framework enhances the understanding of optical tactile sensors, facilitate their evaluation, and inspire future innovations in the field.

Underactuated tendon-driven robotic hands are widely used in grasping tasks due to their excellent passive compliance, allowing them to adapt to various object surfaces. However, achieving finger synchronization using a single motor and differential mechanism has long been a challenging research problem. In Chapter 3, a novel underactuated tendon-driven robotic hand, named BRL/Pisa/IIT SoftHand, is introduced. This robotic hand integrates both soft and adaptive synergies, enabling a single motor to drive all 15 degrees of freedom in the hand. The innovative differential mechanism enhances coordination among the fingers, allowing the hand to adapt effectively to more complex object surfaces.

Many existing optical tactile sensors rely on silicone casting methods to fabricate soft contact modules. However, this approach is not only complex and time-consuming but also limits the design and distribution of internal markers. In Chapter 4, a novel multi-material, multi-component printing method for optical tactile sensor fabrication is proposed. This method enables the customization of markers in arbitrary shapes, overcoming the constraints imposed by traditional silicone casting on marker design. The entire manufacturing process requires no post-processing steps. Additionally, a fully 3D-printed tactile fingertip was designed and fabricated, achieving seamless integration of the optical tactile sensor with robotic phalanges. By leveraging an interpretable tactile model, the fingertip is capable of contact area localization and slip detection. The proposed method enhances the functionality and customization capabilities of optical tactile sensors and paves the way for more efficient and versatile manufacturing processes in robotic tactile sensing.

Improving the dexterous manipulation capabilities of tendon-driven robotic hands while preserving their adaptive capabilities has long been a challenging problem, Chapter 5 introduces the Tactile SoftHand-A to solve this problem. This design builds upon the BRL/Pisa/IIT SoftHand prototype (Chapter 3) and integrates fully 3D-printed tactile fingertips (Chapter 4). Equipped with an innovative antagonistic tendon mechanism, the Tactile SoftHand-A enables active closing, opening, and precise grasp posture control for the entire hand using just two motors. Furthermore, a human-guided tactile feedback grasping control system was developed, allowing the Tactile SoftHand-A to mirror human gestures, monitor object status in real-time during grasping, and maintain stable grips. In the event of slippage, the system can automatically adjust the grasp posture to prevent the object from sliding.

To enhance the capabilities of marker-based optical tactile sensors, enabling them to utilize both marker displacement and light intensity variations for predicting contact location and force, Chapter 6 introduces a novel optical tactile sensor named BioTacTip. This sensor features a sensing unit composed of two distinct types of tips, which can accurately predict tactile information such as contact location, depth, normal force, and shear force. Building on its sensing mechanism, an analyzable and interpretable tactile model was developed. Leveraging this model, the sensor eliminates the need for extensive dataset collection and can be utilized immediately after a simple calibration process.

This thesis focuses on the electromechanical system design of tactile robotic hands. It proposes a new classification method and a novel 3D printing approach for optical tactile sensors. Additionally, it introduces an underactuated robotic hand with an active antagonistic tendon mechanism, along with a human-guided tactile feedback grasping control system.

Date of Award	4 Feb 2025
Original language	English
Awarding Institution	University of Bristol
Supervisor	Nathan F Lepora (Supervisor) & Efi Psomopoulou (Supervisor)