📋 Case Study

Vision-Guided Palletizing Robot for Mixed-SKU E-Commerce Fulfillment

Unstructured tote input with random part orientation, variable box dimensions, and reflective surfaces causing glare

🏗️ Project Overview

Automated distribution center serving Amazon Prime logistics hub

🎯 Challenge

Unstructured tote input with random part orientation, variable box dimensions, and reflective surfaces causing glare

🔧 Design Approach

Dual-camera stereo setup with polarized dome lighting; custom CNN-based pose estimator trained on synthetic data; dynamic pallet layer planning with real-time collision-free path regeneration

📐 Design Diagram

Vision-Guided Palletizing RobotMixed-SKU E-Commerce FulfillmentLRFOV: 1240×980 mmFOV: 1240×980 mmPolarized DomeUnstructured Tote(random orientation)Dynamic PalletLayer PlanningCNN Pose Estimator(Synthetic Data Trained)Stereo Depth MapReal-time Path RegenGlare (Reflective Surfaces)Contrast ≥ 0.72ChallengeSystemProcessLighting

AI-generated project design illustration

📐 Key Calculations

FOV Coverage per Camera

2 × WD × tan(θ/2)
Result: 1240 mm × 980 mm
Ensures full tote coverage at 1.2 m working distance

Minimum Contrast Ratio

(L_max − L_min)/(L_max + L_min)
Result: 0.72
Guarantees reliable edge detection on glossy cartons

📊 Results

99.8% placement accuracy across 212 SKU types; 22% increase in palletizing throughput; zero manual intervention for 14+ hours per shift

💡 Lessons Learned

  • Synthetic data augmentation is essential for rare SKU variants
  • Polarized lighting eliminated >90% specular artifacts
  • Dynamic layer planning reduced pallet instability by 40%

Key Takeaways

  • 1Synthetic data augmentation is essential for rare SKU variants
  • 2Polarized lighting eliminated >90% specular artifacts
  • 3Dynamic layer planning reduced pallet instability by 40%