This thesis provides a brief intro to the peg-and-hole problem and goes through two active and passive compliance strategies as well as the cases that the compliance center position is not ideal. A specific scenario of peg being gripped at an error angle due to large vision uncertainties is raised and studied. Such setup can lead to an off compliance center position relative to the peg and can happen in real life but has not been solved by previous approaches. A potential solution to it by combining active control and passive compliance is provided and analyzed. By using the force and torque feedback and the robot joint angle information, the compliance center for the above scenario could be estimated without vision feedback to by-pass the potential accuracy limitation of vision sensors. And a position control with reference determined in real-time by these sensors would be able to cancel out the majority of the effect caused by an off compliance center. Additional recommendations to future work on integration of active and passive compliance strategies and utilization of arbitrary compliance center positions are provided as well.

Both position and angular errors during the insertion process, which cannot be easily predicted because of indeterminate collision situations, may cause failure of the assembly. One of the frequently applied strategies is to use a passive remote center compliance. We break the insertion problem down in to two phases: chamfer-crossing, and inserting (after chamfer-crossing)phase. In this article, the relationship between the position and angular errors during chamfer-crossing with different chamfer size and locality are thoroughly analysis. We also try to design a technological processes of minimizing the angular errors during chamfer-crossing. Besides single round peg insertion, two dimensional dual peg-in-hole insertion problems are briefly analysis.