Ko Ayusawa

Inertial parameters should be estimated accurately for precise robot control and simulation. Exciting motions, motions that sufficiently excite all robot dynamics, must be generated to obtain these parameters accurately. However, this process requires a certain level of accuracy in the inertial parameters themselves, leading to a circular dependency, especially in cases with a significant model-reality discrepancy. To address this challenge, we propose a constrained method for generating exciting motions within an iterative data acquisition process (active learning). It optimizes the condition number of the regressor matrix and achieves sufficient excitation by combining several motions. We evaluate our method on a manipulator robot not fixed to the ground, where dynamic constraints should be considered to maintain balance during experiments. Despite a substantial gap between the initial model and the actual robot (simulated by adding a 0.575 kg payload at the end effector), our method continuously improved the condition number without motion execution failures–whereas conventional methods resulted in robot tipping. Cross-validation confirmed that the proposed method achieves the lowest root mean square error among all approaches, demonstrating superior performance for inertial parameter identification. These results collectively demonstrate that our method is particularly practical in applications where dynamic constraints are critical for motion planning.

Flexible tools such as golf clubs and sports prostheses used for exercise are generally constructed from materials that are both strong and lightweight enough to withstand the weight and speed of human movement. The authors have previously proposed a hybrid-link system that integrates a rigid-link system with a flexible structure, achieving forward dynamics simulations of the hybrid-link system with floating base-link, such as a human and a humanoid. However, numerical simulations of models with stiff and light flexible structures diverge and are difficult to realize. Using implicit integration as a forward dynamics calculation method is expected to improve the stability of the calculation. However, it requires information on the gradient of the equations of motion. Therefore, this study extends the comprehensive dynamic gradient calculation method proposed for rigid-link systems to hybrid-link systems with floating base-link; moreover, a forward dynamics simulation with contact force calculation is realized using implicit integration in a simple pendulum model, a robot arm and a humanoid with flexible structures.

To evaluate the effectiveness of robotic rehabilitation alone, we developed a robotic rehabilitation model for rats and investigated the effects of inducing forced response-like movements on learning a choice reaction time task. The rats, while in a standing position, initially press two levers with their forepaws. An air-puff stimulation (go-signal: GS) is then applied to either their left or right forepaw. The rats respond by releasing the lever on the correct-response side (correct-RS). If the rats respond incorrectly, a lever-release motion is forcibly induced in the next repeated trial. Combining the lever activation side (correct-RS or incorrect-RS) and its timing (200 ms after GS or just before reaction time (RT)), we formed four experimental groups. The rats with a forepaw sensorimotor cerebral lesion were trained in serial reversal learning of the task, and their performance was compared using the error rate (ER) and RT. The group with incorrect-RS lever activation after GS (iRSaftGS) and the group with correct-RS lever activation before RT (cRSbfrRT) had lower ERs and RTs than the other groups. In addition, both groups showed large improvement in ER at high ER levels but only the iRSaftGS group’s improvement continued at lower ER levels. We conclude that the cRSbfrRT group facilitates only the extinction process and that the iRSaftGS group, which stretches agonistic muscles required for correct motion, is effective in both extinction and acquisition processes. Our findings provide an opportunity to reconsider the strategies for robotic rehabilitation and training.

Understanding motion intentionality is important to augment motor skills. This can be explained by the anatomical structure and innate immune mechanisms of humans. This study aimed to analyze human motion using musculoskeletal information and estimate motion intentionality, motion direction, and muscle properties. The calculation method for the easy-to-move musculoskeletal direction in the task space was reversed to calculate the optimal muscle properties and matching rate for the motion direction. The analysis results of each walking and feint motion showed that the muscles were trained toward reduced muscular effort and improved skills for each motion. In addition, the results of the optimization of the matching rate and muscle properties indicate that the optimized muscle properties can express the motion better than the reference values, and the average matching rate increases by 2.41×10^-1, particularly for a feint motion. Therefore, skill acquisition and augmentation were achieved.

Dynamic simulation of soft and flexible bodies is important for the motion planning and control of soft robot systems. This study presents a fast real-time simulation of the piecewise constant strain (PCS) model proposed in soft robotics research to calculate the dynamics of a beam or rod structure. We extend the theory of comprehensive motion transformation matrix (CMTM) to the PCS model, which allows us to systematically calculate the gradients in the equations of motion. Using the dynamic gradient, we perform a dynamic simulation of the PCS model via implicit integration. Compared to explicit integration, implicit integration allows us to use a larger time-step width for the integration without a flexible deformation divergence, which decreases the computational time. We present several examples of dynamic simulations, including a relatively rigid material such as carbon-fiber-reinforced plastics used in a leaf-spring-type sports prosthesis.

This study introduces a novel approach to Humanoid Robot Motion Generation using Functional Principal Component Analysis (FPCA) within the framework of Direct Optimal Control (DOC). FPCA efficiently compresses high-dimensional motion data, including ground reaction forces, into a low-dimensional space known as the FPC space. These low-dimensional elements preserve the essential characteristics of the original motions and allow for the synthesis of specific elements to create entirely new representations. By using these low-dimensional elements as optimization variables, we anticipated a significant reduction in computational time for motion generation. Experiments on the humanoid robot HRP-4 J, considering various objective functions, demonstrated the method's applicability for generating whole-body motions. The Root Mean Square Error (RMSE) for angle data between the proposed method and conventional methods yielded an average RMSE of 0.007 rd, indicating precise motion generation without compromising data characteristics. Moreover, the proposed method's computation time was 35.8 times faster than the conventional approach.

Estimation of muscle activity is very important as it can be a cue to assess a person’s movements and intentions. If muscle activity states can be obtained through non-contact measurement, through visual measurement systems, for example, muscle activity will provide data support and help for various study fields. In the present paper, we propose a method to predict human muscle activity from skin surface strain. This requires us to obtain a 3D reconstruction model with a high relative accuracy. The problem is that reconstruction errors due to noise on raw data generated in a visual measurement system are inevitable. In particular, the independent noise between each frame on the time series makes it difficult to accurately track the motion. In order to obtain more precise information about the human skin surface, we propose a method that introduces a temporal constraint in the non-rigid registration process. We can achieve more accurate tracking of shape and motion by constraining the point cloud motion over the time series. Using surface strain as input, we build a multilayer perceptron artificial neural network for inferring muscle activity. In the present paper, we investigate simple lower limb movements to train the network. As a result, we successfully achieve the estimation of muscle activity via surface strain.

Excessive muscle tension is implicitly caused by inactivity or tension in daily activities, and it results in increased joint stiffness and vibration, and thus, poor performance, failure, and injury in sports. Therefore, the routine measurement of muscle tension is important. However, a co-contraction observed in excessive muscle tension cannot be easily detected because it does not appear in motion owing to the counteracting muscle tension, and it cannot be measured by conventional motion capture systems. Therefore, we focused on the physiological characteristics of muscle, that is, the increase in muscle belly cross-sectional area during activity and softening during relaxation. Furthermore, we measured muscle tension, especially co-contraction and relaxation, using a DATSURYOKU sensor, which measures the circumference of the applied part. The experiments showed high interclass correlation between muscle activities and circumference across maximal voluntary co-contractions of the thigh muscles and squats. Moreover, the circumference sensor can measure passive muscle deformation that does not appear in muscle activities. Therefore, the DATSURYOKU sensor showed the potential to routinely measure muscle tension and relaxation, thus avoiding the risk of failure and injury owing to excessive muscle tension and can contribute to the realization of preemptive medicine by measuring daily changes.