Autonomous forklifts in the fleet were experiencing intermittent navigation drift — deviating from mapped paths in specific warehouse zones. The issue was inconsistent and didn't reproduce reliably, making it difficult to diagnose through standard troubleshooting procedures.
Rather than treating symptoms, I instrumented the navigation stack to capture high-resolution pose and scan-match data across affected zones. By correlating drift events with environmental conditions — reflective surfaces, dynamic obstacles, and LiDAR occlusion patterns — I isolated the root cause to scan-matching failures in specific geometric configurations.
Developed a custom diagnostic workflow that combined automated data capture with targeted LiDAR recalibration procedures for problem zones. The workflow was adopted as standard operating procedure across the fleet and reduced navigation drift incidents significantly.
The diagnostic process I built became the go-to procedure for the team when similar issues arose at other sites. It shifted the troubleshooting culture from reactive part-swapping to systematic root-cause analysis.