Design and Evaluation of an Autonomous Robotic System for Multi-Layer Lincoln Log Assembly

Abstract

This thesis presents the design and experimental evaluation of a deterministic autonomous robotic system for the multi-layer assembly of interlocking Lincoln Log structures using a UR3 manipulator. Investigating a paradigm where reliability is achieved through structured environments and pre-computed motion rather than real-time perception, the system integrates a parametric toolpath generator, a gravity-fed dispenser for repeatable part presentation, and a custom-calibrated base for structural alignment. Through an iterative design process (V1 → V3), the system evolved from a three-segment prototype to a four-segment architecture that resolved critical synchronization errors between the end-effector and the pneumatic gripper. This refined pipeline executes a seven-waypoint motion sequence for each component. Experimental results demonstrate 100% pick success and 100% placement reliability in single-layer builds. However, performance degrades in multi-layer configurations—declining to approximately 75% in two-layer and 83% in three-layer builds—as uncorrected positional deviations from mechanical tolerances and gripper variability propagate through sequential operations. These findings define the operational limits of deterministic robotic assembly, demonstrating that constraint-based strategies can achieve high reliability within bounded task regimes but remain limited by cumulative error propagation. This work provides both an integrated robotic assembly pipeline and a rigorous empirical characterization of perception-free execution, establishing a foundation for hybrid systems that incorporate selective sensing to augment performance.