Tek bacaklı zıplayan bir robotun dinamik olarak gömülmüş yaylı ters sarkaç şablonu ile adaptif kontrolü

Practical realization of model-based dynamic legged behaviors is substantiallymore challenging than statically stable behaviors due to their heavy dependenceon second-order system dynamics. This problem is further aggravated by the difficultyof accurately measuring or estimating dynamic parameters such as springand damping constants for associated models and the fact that such parametersare prone to change in time due to heavy use and associated material fatigue.In the first part of this thesis, we present an on-line, model-based adaptive controlmethod for running with a planar spring-mass hopper based on a once-per-stepparameter correction scheme. Our method can be used both as a system identificationtool to determine possibly time-varying spring and damping constants of amiscalibrated system, or as an adaptive controller that can eliminate steady-statetracking errors through appropriate adjustments on dynamic system parameters.We use Spring-Loaded Inverted Pendulum (SLIP) model, which is the mostlyused, effective and accurate descriptive tool for running animals of different sizesand morphologies, to evaluate our algorithm. We present systematic simulationstudies to show that our method can successfully accomplish both accurate trackingand system identification tasks on this model. Additionally, we extend oursimulations to Torque-Actuated Dissipative Spring-Loaded Inverted Pendulum(TD-SLIP) model towards its implementation on an actual robot platform.In the second part of the thesis, we present the design and construction of a oneleggedhopping robot we built to test the practical applicability of our adaptivecontrol algorithm. We summarize the mechanical, electronics and software designof our robot as well as the performed system identification studies to calibrate theunknown system parameters. Finally, we investigate the robot?s motion achievedby a simple torque-actuated open loop controller.