📋 Case Study

Vision-Guided Bin Picking Trajectory Generation for Aerospace Fasteners

Cluttered bin geometry and reflective titanium fasteners causing pose estimation drift → failed grasps and dropped parts

🏗️ Project Overview

Lockheed Martin F-35 wing assembly line, Fort Worth TX

🎯 Challenge

Cluttered bin geometry and reflective titanium fasteners causing pose estimation drift → failed grasps and dropped parts

🔧 Design Approach

Deep learning-based 6D pose estimator fused with IMU data; motion planning via constrained sampling in SE(3) space with grasp feasibility scoring and dynamic re-planning on detection failure

📐 Design Diagram

AI-generated project design illustration

📐 Key Calculations

Grasp Stability Index (GSI)

GSI = det(J^T J) / ||J||_F

Result: ≥ 0.72

Guarantees wrench closure under ±15N disturbance

Replan Latency Budget

t_replan ≤ 1/(2×f_camera)

Result: ≤ 33 ms

Enables sub-60fps recovery without motion interruption

📊 Results

98.4% successful grasp rate, 2.7s avg. pick time, zero fastener damage incidents over 14 months

💡 Lessons Learned

• •SE(3) sampling avoids gimbal lock in high-DOF gripper approaches
• •GSI threshold must be calibrated per material friction coefficient

✅ Key Takeaways

• 1SE(3) sampling avoids gimbal lock in high-DOF gripper approaches
• 2GSI threshold must be calibrated per material friction coefficient