M.Sc Thesis

M.Sc StudentHartston Ron
SubjectImplementation of a Natural Dynamic Controller on an
Underactuated Compass-Biped Robot
DepartmentDepartment of Mechanical Engineering
Supervisors PROF. Miriam Zacksenhouse
Full Thesis textFull thesis text - English Version


Bipedal walking is a fascinating field of research that is critical for humanoid robots. Key issues for walking robots include robustness, energy efficiency, the required sensory measurements, and the complexity of the control algorithm.

Natural dynamic controllers aim to perform the desired, cyclic, task by exploiting the natural dynamics of the system. This can be accomplished by generating torque patterns to actuate the system rather than accurately following predefined trajectories.

Natural dynamic control of a compass-biped using this method has been previously demonstrated in our lab, in simulation. In this thesis we demonstrate successful implementation of this dynamic controller on an under-actuated, dynamically walking, compass-biped robot.

A prototype of a fully functional robot has been designed, manufactured, and constructed. The robot is built of an internal (central) leg and two outer (lateral) legs that are connected and thus function as a single leg. The robot includes motors, actuators and an electronics board that performs all acquisition and control functions. At this stage it can walk only along straight lines. The robot is autonomous except for an external power source.

The parameters of the controller, in particular the magnitude and timing of torque primitives, were optimized offline using multi-objective genetic algorithm, accounting for speed and energy efficiency. The torque pulses prescribed by the optimization algorithm were activated on the robot using a low-level position controller. This low-level controller was designed and optimized based on a detailed model whose parameters were identified using a series of experiments.

The robot demonstrates orbitally stable walking for two operation modes - (1) Open loop - where the required torque is sent directly to the motor and (2) closed loop - where a low-level controller is used to apply the required torque.

The significance of this thesis is in demonstrating the simple, yet effective, nature of the control paradigm used to generate a dynamically stable bipedal walking robot. This proof-of concept provides the basis for future extensions to more complex robots.