Abstract. To emulate the actual neuro-control mecha- nism of human bipedal locomotion, an anatomically and physiologically based neuro-musculo-skeletal model is developed. The human musculo-skeletal system is con- structed as seven rigid links in a sagittal plane, with a total of nine principal muscles. The nervous system consists of an alpha motoneuron and proprioceptors such as a muscle spindle and a Golgi tendon organ for each muscle. At the motoneurons, feedback signals from the proprioceptors are integrated with the signal induced by foot±ground contact and input from the rhythm pattern generator; a muscle activation signal is produced accordingly. Weights of connection in the neural network are optimized using a genetic algorithm, thus maximizing walking distance and minimizing energy consumption. The generated walking pattern is in remarkably good agreement with that of actual human walking, indicating that the locomotory pattern could be generated automatically, according to the musculo- skeletal structures and the connections of the peripheral nervous system, particularly due to the reciprocal innervation in the muscle spindles. Using the proposed model, the ¯ow of sensory-motor information during locomotion is estimated and a possible neuro-control mechanism is discussed. 1 Introduction Humans are able to walk without tumbling by skillfully coordinating the activation of redundant numbers of muscles in the lower extremities. Previous research on decerebrate cats has revealed that such dexterous locomotory pattern is spontaneously generated by alternating the activities of the extensor and ¯exor muscles under the control of a group of rhythm- generating neural circuits in the spinal cord known as the central pattern generator CPG); the details of locomotion are not controlled by the higher centers Grillner 1975; Shik and Orlovsky 1976). However, the simple rhythmic signal generated by the CPG does not independently generate a successful locomotory pattern. For instance, Gray 1950) reported that deaerented frogs did not exhibit normal locomotor patterns, indicating that proprioceptive information is required for normal locomotion. A similar conclusion is drawn for higher vertebrates as well Pearson and Ramirez 1997). Therefore, the alternating signals generated by the CPG and the aerent proprioceptive information are both essential in constructing locomotion and must be mutually coordinated in the lower nervous system. In contrast to these neurophysiological approaches, Taga 1995) attempted to clarify the neuro-control mechanism of human locomotion using a more com- putational approach. The study demonstrates that walking movement is autonomously generated as a sta- ble limit cycle that emerges from the dynamic interaction between neural oscillation originating in the CPG and pendulum oscillation of the body linkage. Its neural network, however, does not correspond to the actual nervous system. In the present study, a more physiologically and an- atomically based model of the human lower extremity is constructed. The proposed model corresponds closely to the actual human structure and synthesizes bipedal lo- comotion autonomously. Such a model seems to be ca- pable of generating a walking pattern that is more similar to that in humans and enables comprehensive analysis of the dynamic behavior of the neuro-musculo- skeletal system during locomotion. In addition to the Taga model, several models have been proposed for neuronal synthesis of human loco- motion Bay and Hemami 1987; Srinivasan et al. 1992). However, almost no attempts have been made to syn- thesize locomotion by precisely imitating the human neuro-musculo-skeletal system. Yamaguchi and Zajac 1990) and Gerritsen et al. 1998) have proposed human Correspondence to: N. Ogihara Department of Zoology, Kyoto University, Sakyo-ku, Kyoto 606-8502, Japan e-mail: ogihara@anthro.zool.kyoto-u.ac.jp; Fax: +81-75-7534083) Biol. Cybern. 84, 1±11 2001) Generation of human bipedal locomotion by a bio-mimetic neuro-musculo-skeletal model Naomichi Ogihara, Nobutoshi Yamazaki Department of Biomedical Engineering, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan Received: 03 December 1998 / Accepted in revised form: 09 June 2000