Master’s thesis — Peg-in-hole assembly with four control modes
Thesis defence presentation (PDF, 2022)
Context
Master’s thesis, Innopolis University (Dec 2021 – Jul 2022). Builds on bilateral teleoperation, VR interfaces, and peg-in-hole modules on the same robot.
Problem
Inserting a cylindrical peg into a hole with clearance at or below the robot’s positioning accuracy — a standard benchmark where pure position control fails without contact sensing or human guidance.
Approach
Implemented and compared four control strategies on a KUKA LBR iiwa (ROS):
- Full autonomy — vision-guided pick and hybrid position/force assembly (IEEE NIR 2021 paper)
- Bilateral teleoperation — Touch haptic device with endeffector force feedback
- Shared autonomy — operator guidance with automated compliance
- Learning from demonstration — imitation-based insertion
Also integrated computer vision for part localisation, a unified hardware software stack, and user studies (design of experiments + survey).
Outcome
Reliable assembly under tight clearance; bilateral and shared modes improved operator performance vs. position-only control. Presented at thesis defence.
Stack
ROS · Python · C++ · Computer vision · Haptics · Control systems · KUKA Sunrise
Documents & links
- Thesis defence presentation (PDF) — Innopolis University, Jul 2022
- IEEE NIR 2021 publication — related pick-and-assembly work
- Demo (YouTube)
- MasterThesisWork (GitHub) — code and supplementary materials