THE JOURNAL OF VISUALIZATION AND COMPUTER ANIMATION J. Visual. Comput. Animat. 2003; 14: 73–92 (DOI: 10.1002/vis.306) ****************************************************************************************************** Human gait simulation with a neuromusculoskeletal model and evolutionary computation Kazunori Hase*, Kazuo Miyashita, Sooyol Ok and Yoshiki Arakawa ************************************************************************************ This paper describes a human gait animation system with a precise neuromusculoskeletal model and evolutionary computation. The neuromusculoskeletal model incorporates 14 rigid bodies, 19 degrees of freedom, 60 muscular models, 16 pairs of the neural oscillators, and other neuronal systems. By changing the search parameters and the evaluative criteria of the evolutionary search process, we demonstrate various locomotive patterns, such as normal gait, pathological gait, running and ape-like walking. The proposed simulation system takes not only kinematic data but also in vivo dynamic data such as energy consumption information into consideration, so that the resultant locomotion patterns are natural and valid from a biomechanical point of view. Hence the simulation system can also be used for finding a biologically appropriate physical model to realize a desired gait by simultaneously modifying the body dynamics parameters with the neuronal parameters. This capability creates a novel application of human gait simulation systems, such as rehabilitation tool design and consultation for physically handicapped people. Copyright # 2003 John Wiley & Sons, Ltd. Received: 10 April 2002; Revised: 28 November 2002 KEY WORDS: physics-based animation; human gait; evolutionary computation; genetic algorithms; neuromusculoskeletal model; neural oscillator Introduction The modelling of natural human locomotion is a widely researched subject in computer graphics because loco- motion is a fundamental and common activity. Recent studies of physics-based animation techniques and phy- sics-based human gait animation systems attend to the gravitational and inertial properties of the human body, 1,2 but they do not model the mechanical proper- ties of the muscles and the controlling properties of the nervous system. If an animation system employs a physics-based model with anatomical and neurophy- siological characteristics, the system may generate a more natural movement. The reasons are that the ana- tomical and neurophysiological properties of the body influence body movement, 3 and that human locomotion patterns are determined in particular by biomechanical factors such as energy consumption in the body. 4 There- fore, we believe that the construction of a precise body model of a human is important not only for biomecha- nical research, but also for developing novel graphics tools for computer animation. The purpose of the present study is to generate hu- man gait patterns based on both physical and biological properties of a human. We propose a human gait simulation/animation system that involves the anato- mical characteristics of the human body and the neuro- physiological characteristics of the nervous control system. The proposed body dynamics system has 14 rigid three-dimensional bodies, 19 degrees of freedom and 60 muscular models. The control system has 16 pairs of neural oscillators that model the neural rhythm generator, and embeds somatic sensory feedback, such as the joint angles and foot–ground contact signal. Embedded with such neuronal and musculoskeletal models, the system is expected to generate not only adequate gait patterns but also robust motion against mechanical perturbation. A problem with such a precise and complex locomo- tion model is that the details of the control properties of the neuronal system must be determined appropriately according to the motion to be generated. Though the ****************************************************************************************************** Copyright # 2003 John Wiley & Sons, Ltd. *Correspondence to: K. Hase, National Institute of Advanced Industrial Science and Technology, Central 6, 1-1-1 Higashi, Tsukuba 305-8566, Japan. E-mail: kazunori.hase@aist.go.jp