|The University of Adelaide||Home | Departments | Search|
Stumpy: A Pneumatic Muscle Actuated Bi-pedal Robot
Keywords: biped, mobile robot, pneumatic muscle, gait synthesis
(Commenced: 01-Jan-2005,Concluded: 01-Dec-2005)
Bipedal locomotion is an extremely sophisticated process, involving highly non-linear and intrinsically unstable dynamics. A new approach to bipedal robot design has simplified the task of human gait replication, and made the creation of fully functioning human-like robots a tangible prospect. By simulating the natural kinematics of human locomotion, rather than blindly controlling joint motions and torques, a robot can be created that exhibits a natural-looking and energy efficient gait. These are the two main factors that have stunted the development of bipedal robots based on established control methods. Consequently, this is a vital step towards creating advanced prosthetics for the disabled, and more efficient and effective humanoid robots for the purposes of entertainment and human assistance.
This honours project has utilised this new approach to create a bipedal robot, named Stumpy, which can walk on flat ground under its own power and control. To ensure maximal efficiency and human-like behaviour, the design process began with two prototype walkers that could walk under purely gravitational power down a slight decline, one with knees and one without. All iterations of the design have no upper body, and have four legs on a single hip joint, with the inner and outer pair each behaving as one, thus constraining motion to the sagittal plane. The ankles are rigid, and all other joints are constrained to rotate in the pitch axis. Such design inclusions not only greatly simplify the task, but allow focus to be maintained on the essential aspects of the walking process. To allow Stumpy to walk on flat ground, the joints have been actuated by means of pneumatic muscles, which are currently under development at the university. The use of pneumatic muscles has aided in maintaining anthropomorphic realism, both in terms of structure and the resulting dynamics, as well as minimising energy expenditure.
Future design stages will aim to extend the actuated 2D walker to 3D. Suitable control algorithms have to be developed and implemented on the microcontroller platform of the robot. It may be necessary to replace this platform by a more powerful system, such as a DSP or FPGA based design. A regulating control algorithm would ensure that the robot remains in a stable upright position at all times (standing, similar to the inverted pendulum problem). Further extensions should look at more advanced maneuvers such as the onset of walking motion from a standstill position and the coming to hold from walking motion.
Further details about the current state of the project can be found here: