Chen, XuJawale, Neel Anand2025-01-232025-01-232025-01-232024Jawale_washington_0250O_27674.pdfhttps://hdl.handle.net/1773/52829Thesis (Master's)--University of Washington, 2024Ensuring safe and reliable object handling without slippage remains a critical challenge in robotic manipulation, especially as robots are increasingly deployed in industrial applications. Traditional methods often treat slip as a binary event (slip/no-slip); however, accurately quantifying slip as a continuous variable is essential for precise and adaptive control. This continuous measurement allows slip to be integrated as a control variable, enabling strategies such as adjusting gripper force or position and optimizing trajectories to minimize slippage. In this thesis, we leverage tactile sensing to achieve real-time slip detection and quantification. Utilizing machine learning models, we accurately measure slip in real-time and incorporate this feedback into sophisticated control algorithms to effectively mitigate slippage. Additionally, we demonstrate a proof-of-concept showing how sampling-based Model Predictive Control can optimize robot motions to identify and execute trajectories with minimal slip.application/pdfen-USnoneAdaptive ControlMachine LearningModel Predictive ControlRobotic ManipulationTactile SensingTrajectory OptimizationRoboticsComputer scienceMechanical engineeringMechanical engineeringThesis