Vision-Guided Precision Assembly

Robot System Integration
Adaptive Pick & Place for High-Variance Cylindrical Parts
Project type
Robot System Integration
Specific skills
Vision-Guided Assembly, Random Bin Picking, Real-Time Visual Servoing, Auto-Calibration Logic, Kinematic Optimization, High-Speed Pick & Place, Fanuc iRVision 2D, Adaptive Process Control, Collision Avoidance Strategy, Precision Part Alignment

Intro

The Industrial Challenge: High-Mix / Low-Volume production requires robots to adapt to parts that are never in the same position. In this project, the goal was to automate the assembly of multi-layered cylindrical components (Matrioshkas) with high dimensional variance and random orientation.

My Solution: I engineered a Vision-Centric Cell where the robot acts as the master. Instead of relying on rigid fixtures, I used Real-Time Visual Servoing to detect, align, and assemble parts on the fly.

Key Results:

  • Fixtureless Assembly: Eliminated the need for expensive mechanical centering devices.
  • Auto-Calibration: Developed a self-calibrating vision routine, reducing setup time from hours to minutes.
  • Process Stability: Achieved homogeneous deburring quality by inverting the motion logic (Part-on-Tool).

THE DE-RISKING PROTOCOL

Self-Healing Machinery

The Problem: Calibrating a vision system usually requires manual grid placement, introducing human error and downtime.

The Innovation: I automated the calibration process. The robot mounts the target on its tool and autonomously maps the workspace.

Business Value: This "Self-Healing" capability means the machine can recalibrate itself after a collision or maintenance without needing a vision specialist on-site.

DCS Feature Images

Dynamic Bin-Picking Logic

The Strategy: Parts arrive randomly in trays. Using standard "blob detection" isn't enough when parts overlap or touch.

The Logic: I implemented a Space-Awareness Algorithm. Before picking, the vision system checks not just the part's location, but the clearance around it for the gripper fingers. This prevents the "domino effect" of knocking over adjacent parts, a common cause of jams in bin-picking.

HMI Feature Images

Real-Time Visual Servoing

The Challenge: Aligning two rotating parts with +/- 22° misalignment without stopping the cycle.

The Solution: I implemented a parallel task: while the robot rotates the part, the camera triggers high-speed capture (Visual Servoing) to detect the alignment markers in real-time.

The Result: Perfect alignment achieved during the motion phase, effectively hiding the inspection time within the handling time (Zero-Cycle-Time Impact).

Sealing Feature Images

Perfect Quality Consistency

The Optimization: Traditional deburring moves the tool around the part (slow, complex path).

My Approach: I used One-Point Rotation. The robot holds the part and rotates it against a static tool. This simplifies the trajectory calculation to a single axis, ensuring 100% constant contact pressure and uniform finish quality.

Home Check Feature Images

Other projects

Let's work together!

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.