Deng, HaoA new type of Run-by-Run controller based on the DHOBE (Dasgupta-Huang Optimal Bounding Ellipsoid) algorithm is designed and simulated for semiconductor manufacturing process. One approach is to use the algorithm to implement online model identification which leads to a model-reference controller. The other approach utilizes the worst case idea, to implement the set-valued controller. <p>Both kinds of controllers are applied to linear and quadratic models which are derived from experiments. The controllers are simulated for cases when processes satisfy slow drifting, abrupt shift, bad data and model errors. The controllers are tuned according to the requirements of the algorithm and process and the simulation data is analyzed according to the performance benchmark. <p>All the simulation results are compared to either the Exponentiallly Weighted Moving Average (EWMA) or Optimal Adaptive Quality Controller (OAQC) control method.en-USRun-to-Run ControlSemiconductor ManufacturingOptimal Bounded EllipsoidProcess Adaptive ControlWorst CaseIntelligent Control SystemsThesis