Mechanical Design Methodology for a Biarticularly Driven Biped Robot with Complex Joint Geometry

The biarticular actuators are widely used in the musculoskeletal bipedal robots. We derive the mathematical formulations to explain the static function of the leg muscles. The goal is to report analytical and biological observations to facilitate the design of the anthropomorphic robotic leg. Our analysis indicates that the ratio of the moment arms, in two spanning joints, has the profound influence on the function of the biarticular muscles. Detailed moment arm dataset of the human body is applied to verify the role of the biarticular muscles in various forms of locomotion. Satisfying the technical constraints, we provide a guideline on the structural design of the biarticular actuators based on the associated action directions. Our analyses are investigated using the compliant robotic leg Carl. A vertical jump experiment is performed to demonstrate the action of each actuator and the influence of the moment arm values.

Exoskeleton for motion assistance has obtained more and more attention due to its advantages in rehabilitation and assistance for daily life. This research designed an estimation method of human joint torque by the kinetic human–machine interaction between the operator’s elbow joint torque and the output of exoskeleton. The human elbow joint torque estimation was obtained by back propagation (BP) neural network with physiological and physical input elements including shoulder posture, elbow joint-related muscles activation, elbow joint position, and angular velocity. An elbow-powered exoskeleton was developed to verify the validity of the human elbow joint torque estimation. The average correlation coefficients of estimated and measured three shoulder joint angles are 97.9%, 96.2%, and 98.1%, which show that estimated joint angles are consistent with the measured joint angle. The average root-mean-square error between estimated elbow joint torque and measured values is about 0.143[Formula: see text]N[Formula: see text]m. The experiment results proved that the proposed strategy had good performance in human joint torque estimation.

Bipedal robotic leg kinematics and dynamics are a key component in designing biomimetic humanoid robots. This work describes the process of designing artificial muscle attachment locations of the legs of a bipedal robot utilizing pneumatic artificial muscles (PAMs). PAMs offer similar force and activation times to real muscles, while being lightweight and low power. However, not all properties are identical, and substituting artificial muscles in with the same attachment locations results in different torque profiles about each joint. This work analyzes muscle length, moment arm, and isometric force of muscles over a range of different joint configurations, based on muscle attachment and wrapping points of the Gait2392 model in OpenSim. These parameters are then used to find joint torques produced by each muscle. This process is repeated for a model of the robot actuated by PAMs. Joint torques for individual muscles and groups of muscles are then compared between the models. Results indicate that several muscles torque profiles in the robot model closely match that of the human model, however, many muscles fall far short of the human capabilities. Where appropriate, additional results demonstrate how muscle attachment locations on the robot have been modified to better match the torque capabilities demonstrated by the OpenSim model. Matlab tools created for this project will facilitate further design refinement before implementation on the robot.

This paper presents a novel type jumping robot using bi-articular spring and a mono-articular electric motor, JUMPBiE. The system design is based on the concept of the mono-bi configuration which can effectively reduce the number of actuators and thus cost, retaining the characteristics of bi-articular actuators such as high compliance, stabilization, accuracy, and homogeneity of motion. By showing extensive jumping control of JUMPBiE, it proves the feasibility of the proposed system configuration and provides in-depth insights on under-actuated bi-articular actuation. The background of the robot design and its modeling were elaborated, and discussed with relevant simulation results.

The interaction of muscle moment arm, knee laxity, and torque in a multi-scale musculoskeletal model of the lower limb

Human Motion Capture (MoCap) data can be used for animation of virtual human-like characters in distributed virtual reality applications and networked games. MoCap data compressed using the standard MPEG-4 encoding pipeline comprising of predictive encoding (and/or DCT decorrelation), quantization, and arithmetic/Huffman encoding, entails significant power consumption for the purpose of decompression. In this paper, we propose a novel algorithm for compression of MoCap data, which is based on smart indexing of the MoCap data by exploiting structural information derived from the skeletal virtual human model. The indexing algorithm can be fine-controlled using three predefined quality control parameters (QCPs). We demonstrate how an efficient combination of the three QCPs results in a lower network bandwidth requirement and reduced power consumption for data decompression at the client end when compared to standard MPEG-4 compression. Since the proposed algorithm exploits structural information derived from the skeletal virtual human model, it is observed to result in virtual human animation of visually acceptable quality upon decompression.

A feasible walking gait planning method for humanoid robot based on human motion capture data is proposed in this paper. Firstly, we collect human walking data with a human motion capture system and extract 14 key actions that represent the characteristics of human walking gait in a gait cycle; Then, we take the minimum distance from the toe and the heel of the subject to the level ground as the pivotal parameter of robot kinematic model, and compute the motion data of the 14 key moments in the joint space using the 7-link biped model of humanoid robot with consideration of the difference between the kinematic models of human and of humanoid; Finally, Our approach is validated by simulation of NAO robot in Webots.

Learning to combine control at the level of joint torques with longer-term goal-directed behavior is a long-standing challenge for physically embodied artificial agents. Intelligent behavior in the physical world unfolds across multiple spatial and temporal scales: Although movements are ultimately executed at the level of instantaneous muscle tensions or joint torques, they must be selected to serve goals that are defined on much longer time scales and that often involve complex interactions with the environment and other agents. Recent research has demonstrated the potential of learning-based approaches applied to the respective problems of complex movement, long-term planning, and multiagent coordination. However, their integration traditionally required the design and optimization of independent subsystems and remains challenging. In this work, we tackled the integration of motor control and long-horizon decision-making in the context of simulated humanoid football, which requires agile motor control and multiagent coordination. We optimized teams of agents to play simulated football via reinforcement learning, constraining the solution space to that of plausible movements learned using human motion capture data. They were trained to maximize several environment rewards and to imitate pretrained football-specific skills if doing so led to improved performance. The result is a team of coordinated humanoid football players that exhibit complex behavior at different scales, quantified by a range of analysis and statistics, including those used in real-world sport analytics. Our work constitutes a complete demonstration of learned integrated decision-making at multiple scales in a multiagent setting.

Moles have modified thoracic limbs with hypertrophied pectoral girdle muscles that allow them to apply remarkably high lateral out-forces during the power stroke when burrowing. To further understand the high force capabilities of mole forelimbs, architectural properties of the thoracic limb muscles were quantified in the Eastern mole (Scalopus aquaticus). Architectural properties measured included muscle mass, moment arm, belly length, fascicle length, and pennation angle, and these were used to provide estimates of maximum isometric force, joint torque, and power. Measurements of muscle moment arms and limb lever lengths were additionally used to analyze the out-force contributions of the major pectoral girdle muscles. Most muscles have relatively long fascicles and little-to-no pennation. The humeral abductor/rotators as a functional group are massive and are capable of relatively high force, power, and joint torque. Of this group, the bipennate m. teres major is the most massive and has the capacity to produce the highest force and joint torque to abduct and axially rotate the humerus. In general, the distal limb muscles are relatively small, but have the capacity for high force and mechanical work by fascicle shortening. The muscle architectural properties of the elbow extensors (e.g., m. triceps brachii) and carpal flexors (e.g., m. palmaris longus) are consistent with the function of these muscles to augment lateral out-force application. The humeral abductor/rotators m. latissimus dorsi, m. teres major, m. pectoralis, and m. subscapularis are calculated to contribute 13.9 N to out-force during the power stroke, and this force is applied in a 'frontal' plane causing abduction of the humerus about the sternoclavicular joint. Moles have several specializations of their digging apparatus that greatly enhance the application of out-force, and these morphological features suggest convergence on limb form and burrowing function between New and Old World moles.

Seventeen cases with heightening of the greater trochanter in congenital dislocation of the hip were treated by transplantation of the greater trochanter. Of those cases, ten cases with acetabular dysplasia combined with Chiari's medial displacement osteotomy. The age at operation ranged from 9 to 33 years.Observation periods averaged 3.4 years. These cases were evaluated clinically and roentgenologically. The following roentgenological measurements were performed; 1) the CE angle 2) articulo-trochanteric-distance (ATD) 3) lever arm ratio.The clinical assessemt was made using Colton's point-soring system. The clinical assessment correlated with the postoperative CE angle and lever arm ratio except two cases but did not with ATD. Fourteen cases had good results but two cases poor because of weakness of the abductor muscle. In this report, trying to assess the quantitative fatigue of the abductors (m. gluteus medius) by means of dynamic spectrum analysis (frequency analysis) of EMG was introduced.

Although legged locomotion is better at tackling complicated terrains compared with wheeled locomotion, legged robots are rare, in part, because of the lack of simple design tools. The dynamics governing legged locomotion are generally nonlinear and hybrid (piecewise-continuous) and so require numerical simulation for analysis and are not easily applied to robot designs. During the past decade, a few approximated analytical solutions of Spring-Loaded Inverted Pendulum (SLIP), a canonical model in legged locomotion, have been developed. However, SLIP is energy conserving and cannot predict the dynamical stability of real-world legged locomotion. To develop new analytical tools for legged robot designs, we first analytically solved SLIP in a new way. Then based on SLIP solution, we developed an analytical solution of a hip-actuated Spring-Loaded Inverted Pendulum (hip-actuated-SLIP) model, which is more biologically relevant and stable than the canonical energy conserving SLIP model. The analytical approximations offered here for SLIP and the hip actuated-SLIP solutions compare well with the numerical simulations of each. The analytical solutions presented here are simpler in form than those resulting from existing analytical approximations. The analytical solutions of SLIP and the hip actuated-SLIP can be used as tools for robot design or for generating biological hypotheses.

Many legged robots have been developed, and some of them have already achieved dynamically stable bipedal and quadruped locomotion. Common characteristics of these conventional legged robots are that their motions are precisely manipulated by their control (i.e., gait and balance). In the case of ASIMO (Hirose et al., 2001) and HRP2 (Kaneko et al., 2004), their dynamically stable bipedal locomotion is based primarily on the Zero Moment Point (ZMP) concept, which was originally developed by Vukobratovic (Vukobratovic, 1969): the ZMP is a fictitious point on the ground plane, where the torques around the (horizontal) x and y axes (generated by ground reaction forces, inertial forces, and torques) are equal to zero. These robots dynamically stabilize their balance by manipulating the ZMP to remain within the support polygon area defined by their square feet, and their gait control is applied on top of the balance control. As a result, these conventional legged robots require rapid information processing and high-power drives to achieve locomotion. Therefore, they are characterized as having more complex control systems and larger energy consumption than those of biological locomotion. On the other hand, these design requirements (i.e., fast information processing and highpower drives) are regarded as disadvantageous in biologically inspired robotics. In the field, the specific behaviors and/or structures of biological organisms are imitated (Vogel, 1998) (Alexander, 2002) with robotic technology. As a result, the robot systems have simple design requirements compared to conventional robot systems, and achieve complex tasks. Thus, the design effort tends to focus on control complexity and energy requirements. A representative instance is the Passive Dynamic Walker (PDW), which was originally developed by McGeer (McGeer, 1990). The PDW has no controller (i.e., no sensor and no motor), and the structure is based on the physical characteristics of human walking: passive hip joints, latch knee joints, and curved feet. The PDW walks down a slope, and its structure exploits gravity as the driving force. This design principle has been applied to the Cornell Biped (Collins & Ruina, 2005) and Denise at TU Delft (Wisse, 2003), and the powered passive dynamic walkers achieved passive dynamic walking on a flat plane. In a summary of these two approaches, conventional robots require expensive design components to achieve dynamically stable locomotion. On the other hand, biologically inspired robots require physical characteristics that exploit their own dynamics to achieve

Studies on decerebrate walking cats have shown that phase transition is strongly related to muscular sensory signals at limbs. To further investigate the role of such signals terminating the stance phase, we developed a biomimetic feline platform. Adopting link lengths and moment arms from an Acinonyx jubatus, we built a pair of hindlimbs connected to a hindquarter and attached it to a sliding strut, simulating solid forelimbs. Artificial pneumatic muscles simulate biological muscles through a control method based on EMG signals from walking cats (Felis catus). Using the bio-inspired muscular unloading rule, where a decreasing ground reaction force triggers phase transition, stable walking on a treadmill was achieved. Finally, an alternating gait is possible using the unloading rule, withstanding disturbances and systematic muscular changes, not only contributing to our understanding on how cats may walk, but also helping develop better legged robots.

Muscle force (MF) is linearly related to physiological cross-sectional area (CSA), which is obtained from muscle volume (MV) divided by fibre length. Taking into account the fact that joint torque (TQ) is determined by MF multiplied by the moment arm, the maximal TQ would be a function of MV. This proposition was tested in the present study by investigating the relationship between MV and TQ for elbow flexor (EF) and extensor (EE) muscles of 26 males. The MVs of EF and EE were determined from a series of muscle CSA by magnetic resonance imaging (MRI), and pennation angle (theta) and FL by ultrasonography (US). Maximal isometric TQ was measured at right angle of elbow joint for EF and EE. There was a highly significant correlation between MV and TQ both for EF and EE (r=0.95 and 0.96 respectively) compared with that between muscle CSA and TQ, suggesting the dependence of TQ on MV. Furthermore, prediction equations for MV (MVULT) from muscle thickness (MT) measured by US was developed with reference to MVMRI by the MRI on 26 subjects, and the equations were applied to estimate MV of healthy university students (CON; 160 males) and sports athletes (ATH; 99 males). There were significant linear relationships between MVULT and TQ both for EF (r=0.783) and EE (r=0.695) for all subjects (n=259). The MVULT was significantly higher in ATH (by 32% for EF and 33% for EE, respectively) than in CON. Similarly, significantly greater TQ was observed in ATH (by 35% for EF, 37% for EE, respectively). The theta for EE showed no difference between both groups (17.8 degrees for CON and 17.5 degrees for ATH). On the other hand, the TQ to MV ratio were identical for CON and ATH. The results reveal that the muscle volume of the upper arm is a major determinant of joint torque (TQ), regardless of athletic training.

In this study, we aimed to determine whether the semitendinosus (ST) and biceps femoris long head (BF) exhibit systematically distinct preferred directions (PDs) and to evaluate the biomechanical mechanisms underlying the PDs. A total of 21 healthy young adults performed isometric torque tasks in 16 sagittal-plane directions while surface electromyography (EMG) signals were recorded from six lower limb muscles. Joint torque was estimated with inverse dynamics, and muscle PDs were determined through linear regression of EMG against torque data. Optimal PDs were predicted with static optimization modeling that incorporated both moment arm and physiological cross-sectional areas. Mechanical-only PDs were also estimated based solely on moment arm geometry. Experimentally derived PDs differed significantly between ST and BF (P < 0.001). The ST exhibited significantly greater alignment with optimization-based predictions (mean cosine similarity = 0.96 ± 0.08) than with mechanical-only predictions (0.82 ± 0.15, P = 0.0007), whereas the BF showed comparable levels of agreement with both models (optimization = 0.70 ± 0.35, mechanical = 0.76 ± 0.20, P = 0.3096). These findings indicate that muscle-specific neural modulation contributes to directional tuning, particularly in the ST, and highlight the necessity of incorporating both biomechanical and neural factors to understand spatial organization of muscle activity during complex multijoint motor tasks.NEW & NOTEWORTHY This study demonstrates that the semitendinosus and biceps femoris long head exhibit distinct preferred directions during multijoint isometric torque tasks. Whereas biceps femoris shows high interindividual variability despite anatomical alignment, semitendinosus consistently deviates from mechanical predictions, suggesting stable neural modulation. These findings reveal how the central nervous system selectively tunes muscle activity in a muscle-specific manner, balancing anatomical structure and task-dependent control across multijoint muscles.