Genetic algorithm introduction and advice – Adrian Boeing Advice, encouragement, coffee conversation and many ideas –

As a contribution toward the objective of developing useful walking machines, this dissertation considers solutions to some of the problems involved with bipedal robot development. The main area of focus involves control system design and implementation for dynamically balanced walking robots. A new algorithm “Torso Driven Walking ” is presented, which reduces the complexity of the control problem to that of balancing the robot’s torso. All other aspects of the system are indirectly controlled by the changing robot state resulting from direct control of the robot’s torso. The result is literally a “top-down ” approach to control, where the control system actively balances the top-most component of the robot’s body, leaving the control of the lower limbs to a passive “state-driven ” system designed to ensure the robot always keeps at least one leg between the torso and the ground. A series of low-cost robots and simulation systems have been constructed as experimental platforms for testing the proposed new control system. The robots have been designed to balance on “point ” feet, and so the control system must be able to dynamically maintain balance, while moving at a variable velocity. The Torso Driven Walking control system achieves a fully dynamic, variable speed walking behaviour that does not rely on maintaining a stable supporting polygon for balance. In addition, the system exhibits a high degree of tolerance for low frequency “bias ” or “drift ” errors. These measurement errors are commonly encountered when using sensors for detecting torso inclination. ii Acknowledgements In completion of this Thesis I would like to thank several people for their support and valuable contributions