Ashish D. Deshpande

Mechanisms of the Anatomically Correct Testbed Hand

We have built an anatomically correct testbed (ACT) hand with the purpose of understanding the intrinsic biomechanical and control features in human hands that are critical for achieving robust, versatile, and dexterous movements, as well as rich object and world exploration. By mimicking the underlying mechanics and controls of the human hand in a hardware platform, our goal is to achieve previously unmatched grasping and manipulation skills. In this paper, the novel constituting mechanisms, unique muscle to joint relationships, and movement demonstrations of the thumb, index finger, middle finger, and wrist of the ACT Hand are presented. The grasping and manipulation abilities of the ACT Hand are also illustrated. The fully functional ACT Hand platform allows for the possibility to design and experiment with novel control algorithms leading to a deeper understanding of human dexterity.

An index finger exoskeleton with series elastic actuation for rehabilitation

Rehabilitation of the hands is critical for the restoration of independence in activities of daily living for individuals exhibiting disabilities of the upper extremities. There is initial evidence that robotic devices with force-control-based strategies can help in effective rehabilitation of human limbs. However, to the best of our knowledge, none of the existing hand exoskeletons allow for accurate force or torque control. In this work, we present a novel index finger exoskeleton with Bowden-cable-based series elastic actuation allowing for bidirectional torque control of the device with high backdrivability and low reflected inertia. We present exoskeleton and finger joint torque controllers along with an optimization-based offline parameter estimator. Finally, we carry out tests with the developed prototype to characterize its kinematics, dynamics, and controller performance. Results show that the device preserves the characteristics of natural motion of finger and can be controlled to achieve both exoskeleton and finger joint torque control. Finally, dynamic transparency tests show that the device can be controlled to offer minimal resistance to finger motion. Beyond the present application of the device as a hand rehabilitation exoskeleton, it has the potential to be used as a haptic device for teleoperation.

Understanding variable moment arms for the index finger MCP joints through the ACT hand

Human levels of dexterity has not been duplicated in a robotic form to date. Dexterity is achieved in part due to the biomechanical structure, and in part due to the neural control of movement. An anatomically correct test-bed (ACT) hand has been constructed to investigate the importance and behavioral consequences of anatomical features and neural control strategies of the human hand. This paper focused on the role of the human handpsilas variable moment arm. System identification was conducted on the ACT index fingerpsilas two degrees of freedom at the metacarpal-phalange (MCP) joint to provide an understanding of, for the first time, how the moment arms vary with multiple joints moving simultaneously. The specific combination of nonlinear moment arms results in an increased ability to produce force at the fingertip for the same neural input when the fingerpsilas flexion and adduction angles increase (that is toward the middle of the hand). This preliminary work will lead to answering what biomechanical and neural functions are required to construct fully dexterous robotic and prosthetic hands in the future.

Statistical method for prediction of gait kinematics with Gaussian process regression

We propose a novel methodology for predicting human gait pattern kinematics based on a statistical and stochastic approach using a method called Gaussian process regression (GPR). We selected 14 body parameters that significantly affect the gait pattern and 14 joint motions that represent gait kinematics. The body parameter and gait kinematics data were recorded from 113 subjects by anthropometric measurements and a motion capture system. We generated a regression model with GPR for gait pattern prediction and built a stochastic function mapping from body parameters to gait kinematics based on the database and GPR, and validated the model with a cross validation method. The function can not only produce trajectories for the joint motions associated with gait kinematics, but can also estimate the associated uncertainties. Our approach results in a novel, low-cost and subject-specific method for predicting gait kinematics with only the subjects body parameters as the necessary input, and also enables a comprehensive understanding of the correlation and uncertainty between body parameters and gait kinematics.

Anatomically correct testbed hand control: Muscle and joint control strategies

Human hands are capable of many dexterous grasping and manipulation tasks. To understand human levels of dexterity and to achieve it with robotic hands, we constructed an anatomically correct testbed (ACT) hand which allows for the investigation of the biomechanical features and neural control strategies of the human hand. This paper focuses on developing control strategies for the index finger motion of the ACT Hand. A direct muscle position control and a force-optimized joint control are implemented as building blocks and tools for comparisons with future biological control approaches. We show how Gaussian process regression techniques can be used to determine the relationships between the muscle and joint motions in both controllers. Our experiments demonstrate that the direct muscle position controller allows for accurate and fast position tracking, while the force-optimized joint controller allows for exploitation of actuation redundancy in the finger critical for this redundant system. Furthermore, a comparison between Gaussian processes and least squares regression method shows that Gaussian processes provide better parameter estimation and tracking performance. This first control investigation on the ACT hand opens doors to implement biological strategies observed in humans and achieve the ultimate human-level dexterity.

Muscle-tendon units provide limited contributions to the passive stiffness of the index finger metacarpophalangeal joint

The passive stiffness at the MCP joint is a result of the elasticity of muscle-tendon units (MTUs) and capsule ligament complex (CLC), however, the relative contributions of these two components are unknown. We hypothesize that the MTUs provide the majority of the contributions to the joint stiffness by generating resistive forces when the MCP joint is flexed or extended. We used the work done by passive moments as a measure for the determination of the contributions to the joint stiffness. We conducted experiments with ten human subjects and collected joint angle and finger tip force data. The total passive moment and joint angle data were fitted with a double exponential model, and the passive moments due to the MTUs were determined by developing subject-specific models of the passive force-length change relationships. Our results show that for all the subjects, the work done by the passive moments from the MTUs is less than 50% of the total work done, and the CLC provides dominant contributions to the joint stiffness throughout the flexion-extension range of the joint angle. Therefore, the hypothesis that the MTUs provide the majority of the contributions to the MCP joint stiffness is not supported. We also determined that the majority of the MTUs passive moment was generated by the extrinsic MTUs and the contributions of the intrinsic MTUs was negligible.

Acquiring Variable Moment Arms for Index Finger Using a Robotic Testbed

Human level of dexterity has not been duplicated in a robotic form to date. Dexterity is achieved in part due to the biomechanical structure of the human body and in part due to the neural control of movement. We have developed an anatomically correct testbed (ACT) hand to investigate the importance and behavioral consequences of anatomical features and neural control strategies of the human hand. One of the critical aspects of understanding dexterity is the analysis of the relationships between the hand muscle movements and joint movements, defined by the moment arms of the muscles. It is known that the moment arms for the hand muscles are configuration-dependent and vary substantially with change in posture. This paper presents a methodology for determining continuous variations in the moment arms with respect to multiple joints moving simultaneously. To determine variations in the moment arms of the ACT hand index finger muscles, we employed a nonparametric regression method called Gaussian processes (GPs). GPs give a functional mapping between the joint angles and muscle excursions, and the gradients of these mappings are the muscle moment arms. We compared the moment arm relationships of the ACT hand with those determined from the available cadaver data. We present the implications of the determination of variable moment arms toward understanding of the biomechanical properties of the human hand and for the neuromuscular control for the ACT hand index finger movements.

Contributions of Intrinsic Visco-Elastic Torques During Planar Index Finger and Wrist Movements

Human hand movements have been studied for many decades, yet the role of hand biomechanics in achieving dexterity has not been fully understood. In this paper, we investigate the contributions of the intrinsic passive viscoelastic component in the hand during the coordinated wrist and hand movements. We compare the contributions of stiffness, damping, and dynamics torques under two types of joint phase movements at two speeds. The analysis of the data collected from subject studies demonstrated that the passive visco-elastic component is dominant over dynamic coupling terms. Although the exact contributions of the three torques vary under different speeds and phasic movements, the stiffness torque was the highest (at least 47%) followed by the damping torque, while the dynamics torque was the lowest (less than 11%) in all movement scenarios. Comparisons with studies involving coordinated arm movements illustrate that dominant torques in arm and hand movements are different suggesting that neural control strategies might be distinct as well.

Contribution of passive properties of muscle-tendon units to the metacarpophalangeal joint torque of the index finger

The passive joint behavior in the human hand is the result of passive properties of the muscle-tendon units (MTUs) and the elasticity of the soft tissues across the joint. The purpose of the study was to investigate the relative contribution of the MTUs to the net passive torque of the index finger metacarpophalangeal (MCP) joint in flexion and extension. We developed mathematical models to explicitly determine the passive contributions of the seven MTUs to the MCP joint torque. We then compared the computed MTU passive torque to the net passive torque derived from data collected from human subjects. The results show that the MTU properties did not produce the greatest contribution to the the total passive joint torque, especially, at the extremities of the range of motion are dues to factors such as the soft joint tissues. Also, the extrinsic MTUs produced much higher passive joint torques compared to the intrinsic MTUs, and the intrinsic MTUs presented small counterbalance to the net MTU torque in extension. The revelation that most of the net passive joint torque is due to the joint tissue and not due to the MTU elasticity is important for understanding the human hand controls, and also for designing the next generation of robotic hands.

Control strategies for the index finger of a tendon-driven hand

To understand how versatile dexterity is achieved in the human hand and to achieve it in a robotic form, we have constructed an anatomically correct testbed (ACT) hand. This paper focuses on the development of control strategies for the index finger motion and implementation of joint passive behavior in the ACT hand. A direct muscle position control and a force-optimized joint control are implemented for position tracking through muscle force control. The relationships between the muscle and joint motions play a critical role in both of the controllers and we implemented a Gaussian process regression technique to determine these relationships. Our experiments demonstrate that the direct muscle position controller allows for fast position tracking, while the force-optimized joint controller allows for the exploitation of actuation redundancy in the finger critical for this redundant system. We demonstrate that by implementing a passive force–length relationship at each muscle we are able to precisely match joint stiffness of the metacarpophalangeal (MCP) joint of the ACT to that of a human MCP joint. We also show the results from improved position tracking when implemented in the presence of passive muscle control schemes. The control schemes for position tracking and passive behavior are inspired by human neuromuscular control, and form the building blocks for developing future human-like control approaches.