Gait Transition from Quadrupedal to Bipedal Locomotion of an Oscillator-driven Biped Robot

Studies on biped robots have attracted interest due to such problems as inherent poor stability and the cooperation of a large degree of freedom. Furthermore, recent advanced technology, including hardware and software, allows these problems to be tackled, accelerating the interest. Actually, many sophisticated biped robots have already been developed that have successfully achieved such various motions as straight walking, turning, climbing slopes, rising motion, and running (Aoi & Tsuchiya, 2005; Aoi et al., 2004; Hirai et al., 1998; Kuniyoshi et al., 2004; Kuroki et al. 2003; Löffler et al., 2003; Nagasaki et al., 2004). Steady gait for a biped robot implies a stable limit cycle in its state space. Therefore, different steady gait patterns have different limit cycles, and gait transition indicates that the state of the robot moves from one limit cycle to another. Even if the robot obtains steady gait patterns, their transition is not necessarily confirmed as completed. Thus, smooth transition between gait patterns remains difficult. To overcome such difficulty, many studies have concentrated on model-based approaches using inverse kinematics and kinetics. These approaches basically generate robot motions based on such criteria as zero moment point (Vukobratovi et al., 1990) and manipulate robot joints using motors. However, they require accurate modeling of both the robot and the environment as well as complicated computations. The difficulty of achieving adaptability to various environments in the real world is often pointed out, which means that in these approaches the robot is too rigid to react appropriately to environmental changes. Therefore, the key issue in the control is to establish a soft robot by adequately changing the structure and response based on environmental changes. In contrast to robots, millions of animal species adapt themselves to various environments by cooperatively manipulating their complicated and redundant musculoskeletal systems. Many studies have elucidated the mechanisms in their motion generation and control. In particular, neurophysiological studies have revealed that muscle tone control plays important roles in generating adaptive motions (Mori, 1987; Rossignol, 1996; Takakusaki et al., 2003), suggesting the importance of compliance in walking. Actually, many studies on robotics have demonstrated the essential roles of compliance. Specifically, by appropriately employing the mechanical compliance of robots, simple control systems have attained highly adaptive and robust motions, especially in hexapod (Altendorfer et al., 2001; Cham et al., 2004; Quinn et al., 2003; Saranli et al., 2001), quadruped (Fukuoka et al., 2003; Poulakakis