Path Planning and Control Optimization for Collaborative Palletizing

End-of-line palletizing appears straightforward: a box arrives on a conveyor, the robot picks it up and places it on a pallet, repeatedly. The underlying software, however, presents significant engineering challenges — generating a robot program that is collision-free, kinematically valid, and directly executable, automatically, for any box size, pallet layout, or stack height, without requiring an operator to write code. This thesis presents PalGea, a framework developed during an industrial internship at Robogea S.r.l. When the internship began, trajectory generation relied on manually programmed waypoints, with no collision checking, singularity detection, or cycle-time estimation. PalGea replaces all of this with a fully automated, operator-facing software framework validated on the real robot. PalGea takes a palletizing recipe and produces a ready-to-run URScript program for a Universal Robots UR5e via five modules: a deterministic nine-waypoint Cartesian planner; a closed-form analytical IK solver evaluating up to eight configurations per pose; an AABB collision checker; a calibrated cycle-time estimator; and a gripper-agnostic URScript generator supporting digital-I/O, OnRobot VGC10, and Robotiq Hand-E grippers. A Vue.js/FastAPI web HMI provides a five-step wizard, live pallet visualisation, telemetry, and mid-stack job resume — a capability absent from commercial alternatives such as Pally URCap. Validated on the physical UR5e in March 2026, the system averaged 53.5 s/box with zero faults. A May 2026 campaign across 57 runs and 35 configurations achieved ±0.72% cycle-time accuracy for H ≤ 10 cm, extended to ≤0.43% up to H = 25 cm via an empirical residual model. The planning core also ports to the UR15 platform unchanged. Future work covers intermediate-speed characterisation, conveyor integration, and coordinated motion with a seventh-axis lift or linear rail.