Towards a sensorised robotic hand-arm system for humanoids

• Main Author: Sinha, Anoop Kumar
• Other Authors: Cai Yiyu
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Mechanical engineering::Robots
• Online Access: https://hdl.handle.net/10356/167047
• Institution: Nanyang Technological University

Anthropomorphism is considered crucial for physical Human-Robot Interaction (pHRI). It is essential for improved efficiency and meaningful pHRI. Anthropomorphism can be imbued in robotic artifacts on three bases: (1) the perceptional bases, which refer to similarities between the robot and humans in form and appearance; (2) the functional bases, which refer to similarities between the robot and humans in motion and sensing capabilities; and (3) the cognitive and emotional bases, which refer to similarities in cognition and emotion between the robot and humans. The goal of this thesis is to achieve anthropomorphism in humanoid robot hand-arm systems on perceptional and functional bases. To this end, the research was divided into the following five specific tasks: (1) design and develop an anatomically correct humanoid robot hand-arm system with perceptional similarity to the human hand and arm in form and whole configuration for our social humanoid Nadine. In addition to being anatomically and proportionally similar to the human hand and arm, the hand-arm system must also have low inertia for safe pHRI and high power-to-mass and high power-to-volume ratios for adequate gripping strength; (2) devise a method to measure kinematic anthropomorphism of humanoid robot hand-arm systems; (3) design and develop artificial fingertip for operating touch sensitive screens; (4) design, develop, and characterize tactile sensors capable of providing object information during manipulation tasks; (5) design and develop multi-modal tactile sensors. To address the problems of specific task (1), Anthropomorphic proportions and features of the human hand and arm were studied and incorporated into the hand-arm system from very early in the design stage to achieve perceptional similarity. The hand-arm system developed in this research consists of two subsystems: seven degrees of freedom arm and five degrees of freedom polyarticulated robotic hand with fourteen movable joints. The hand-arm system has a modular design. The power-to-mass and power-to-volume ratios of the proposed hand-arm system is 3.83 W/kg and 0.02 W/cm3 respectively. It has low inertia of 0.54 kg-m2. Majority of methods proposed for measuring anthropomorphism focus on quantifying the structural anthropomorphic form and gestural anthropomorphic form of robotic artifacts. These methods are based on visual inspection of the grasp types performed by the robot and comparing it with the standard grasp taxonomies. Such methods can be subjectively biased and complex processes. To address the problems of specific task (2), an analytical method for quantifying the anthropomorphism of robot hand-arm systems is proposed in this thesis. This method compares the workspace coverage and manipulability of the robot arm and fingers with the human arm and fingers. The analytical method proposed in this thesis is used to measure the kinematic anthropomorphism of robotic hand-arm systems. It was found that the kinematic anthropomorphism of the humanoid robot hand-arm system developed in this thesis is 23.305 %. In order to operate touch-sensitive screens as mentioned in specific task (3), self-powered, soft, and conductive artificial fingertip pads have been developed using a composite mixture of silicone and microfine graphite powder. The concept of providing sufficient ground is used to make the fingertip pads self-powered. Aerosol jet 3D printed fingertip tactile pressure sensors were fabricated and characterized to address the problems of specific task (4). Ultra-low-cost (< $1.5), ultra-thin, wide range, and crosstalk-free skin-inspired piezoresistive type tactile sensors were developed. The response time of individual sensing nodes is 4 ms which is faster than the response time of the human skin (30 - 50 ms). The sensors exhibit high sensitivity (1.35 kPa-1), low hysteresis (9.22 %), and a wide pressure sensing range (5 - 600 kPa). The sensor arrays are assembled on the fingertips of a commercial glove to make a smart glove. By combining the sensor information and deep learning, the smart glove is used to identify sharp and blunt objects with a classification accuracy of 95.9 % and the direction of applied pressure when touched by an object with a classification accuracy of 97.8 %. Furthermore, the smart glove is used to generate pressure maps in real time while grabbing six different objects handled by humans in daily life. Finally, to make the tactile sensors multi-modal as mentioned in specific task (5), a multi-modal sensing palmtop is developed. The palmtop features 390 tactile pressure sensing nodes and 5 hydration sensing nodes. The palmtop is then used to generate pressure maps of three objects of different shapes in real time. It is further used to measure hydration levels during evaporation of 1.3 µl of water in real time. The first main contribution of this thesis is the new humanoid robot hand-arm system for our social humanoid Nadine. This system features low inertia, high power-to-mass, and high power-to-weight ratios, enabling Nadine to perform more complex tasks and interact with humans more convincingly. Additionally, a self-powered, soft, and conductive touchscreen add-on enhances the hand-arm system’s capabilities to interact with touch-sensitive devices without requiring a bio-electrical connection. With the ability to interact with touch-sensitive devices, social humanoids can also control and manage various touch sensitive devices in smart home systems and public spaces such as malls, airport, and even hospitals, thereby enhancing convenience and comfort for users. The second main contribution is the anthropomorphism measurement tool proposed in this thesis. This method provides the designers with a tool to quantify the kinematic anthropomorphism of robot hand-arm systems. It also provides designers a tool to recognize possible improvements in the design of the hand-arm system analyzed. A higher anthropomorphic score can enhance the naturalness and intuitiveness of interactions, potentially making the robot with the higher score more approachable and user-friendly, ultimately leading to a more positive HRI experience. Finally, wide range, fast responding, and cross-talk free tactile sensors and the palmtop with distributed network of vast number of these sensors proposed and developed in this thesis can help capture meaningful object related crucial information during grasping and manipulation. This innovation has the potential to enhance the performance and precision of robotic systems, ultimately leading to more efficient and effective robotic systems.