- vol.15, pp.187-198, 2000
Many computer simulation studies of human bipedal walking have been conducted in the field of biomechanics. The musculo-skeletal systems in previous models, however, have been simplified two-dimensionally, and theoretical methods in robotics have been applied for the motor control mechanisms. The purpose of this study was to develop a more precise simulation model of human walking in order to improve the practicability of the simulation method. As a result, a model was created in which the musculo-skeletal system of the entire body is modeled three-dimensionally, and a mechanism for motor control was constructed by a neuronal system model having a hierarchical structure. The inertial properties of the entire human body were represented by a three-dimensional , 14-rigid-link system. These links include the feet, calves, thighs, pelvis, lower lumbar region, upper lumbar region, thorax, upper arms, and forearms. The body's dynamic model is driven by 42 muscles for the entire body. The arrangement of each muscle was represented as a series of line segments, the direction of which changes according to joint angle. Energy consumption, including heat production, in the muscle was calculated from the generating tension. The hierarchical neuronal system includes three levels. First, at the highest level, there is a neuronal system corresponding to the higher center level. The function of adjusting changes in walking pattern is assumed to exist at this system level. The model was expressed by a computational multi-layered neural network. Second, at the middle level , there is a neuronal system corresponding to the spinal cord level. This neuronal system, representing a rhythm-generation mechanism, was modeled as a network system consisting of neural oscillators. They generate the neuronal stimulus combined for each degree of freedom by receiving nonspecific stimulus from the higher center and feedback signals from the somatic senses. Each neural oscillator is mathematically expressed by two differential equations. Third, at the lowest level, there is a neuronal system corresponding to the peripheral level. The neuronal system divides the combined neuronal stimulus from the neural oscillator into the neuronal stimulus to each muscle. The model is mathematically represented as an optimization problem. The simulated walking pattern was continuous and stable. The walking pattern closely agrees with actual human walking in terms not only of joint movement but also of muscle activities and energy consumption. In order to investigate the effects of higher center system functioning in adjusting walking patterns, we compared a walking pattern generated by a model incorporating a higher center system with the patterns obtained from a model without the higher center system, in terms of robustness of mechanical perturbation. Although the model without the higher center system could not stabilize its walking pattern and finally fell down, the model with the higher center system could perform continuous walking without falling down.