Stewart Platform Pose Accuracy Study

ISEF Category: Robotics and Intelligent Machines

Subcategory: Robot Kinematics  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A tiny stage can miss its target by less than a millimeter and still ruin a wafer scan. That is why precision motion systems live or die by pose accuracy, not just motor movement. In this project, you test whether a homebuilt Stewart platform really lands where math says it should. You also find out what kinds of errors grow as the platform moves.

What Is It?

A Stewart platform is a six-legged robot platform that can move in six directions, three for position and three for rotation. Think of it like a table on six adjustable legs. If you change each leg length in the right way, the top plate can tilt, slide, and rotate with high precision.

This project asks a simple question with hard math behind it, how close does the real platform come to the pose predicted by forward kinematics? Forward kinematics is the calculation that turns actuator lengths into platform position and angle. You can compare the predicted pose to the measured pose from a webcam and ArUco markers, which are square visual markers the computer can detect and track. The gap between prediction and reality tells you how much backlash, flex, calibration error, and mounting geometry affect the system.

Why This Is a Good Topic

This makes a strong science fair topic because you can measure real error, change one design factor at a time, and turn the results into clear graphs. It connects to semiconductor tools, camera calibration stages, surgical robots, and motion systems that need exact placement. You can learn kinematics, calibration, image analysis, and error propagation without needing a giant lab. The topic also gives you room to improve the platform, then prove whether your fix actually works.

Research Questions

• How does platform height affect pose error on a six-actuator Stewart platform? ?
• What is the effect of actuator calibration error on forward-kinematics accuracy? ?
• Does adding a camera-based correction step reduce measured pose error? ?
• To what extent does payload mass change repeatability across the motion range? ?
• Which platform region produces the largest difference between predicted and measured pose? ?
• How does motion path shape affect accumulated error during scanning patterns? ?

Basic Materials

• Six hobby linear actuators or screw-driven linear actuators.
• Six matching motor drivers or controller channels.
• Rigid top and base plates with ball joints or rod-end connectors.
• Microcontroller such as Arduino or Raspberry Pi Pico.
• Webcam with stable mounting.
• Printed ArUco markers.
• Digital calipers or ruler for geometry checks.
• Laptop for logging and analysis.
• Power supply sized for all actuators.
• Basic fasteners, brackets, and wiring supplies.

Advanced Materials

• Precision linear actuators with encoder feedback.
• Machined or 3D-printed spherical joints with low backlash.
• Optical breadboard or rigid metrology table.
• Industrial webcam or machine-vision camera with known lens parameters.
• Calibration target for camera intrinsic calibration.
• Digital inclinometer or laser distance sensor for validation.
• Load cells for payload testing.
• Data acquisition hardware for synchronized logging.
• CAD files for exact joint geometry measurement.
• Vibration isolation pads or stage mount.

Software & Tools

• OpenCV: Detects ArUco markers and estimates platform pose from webcam images.
• Python: Logs actuator commands, processes images, and computes error statistics.
• ImageJ: Helps inspect marker sharpness, perspective distortion, and image quality.
• ArUco tools in OpenCV: Generates and tracks markers for pose measurement.
• MIT OpenCourseWare: Offers free robotics and kinematics lectures for the math behind the platform.

Experiment Steps

1. Define the exact pose variables you will measure, then decide which one will be your main accuracy metric.
2. Model the platform geometry in CAD or a spreadsheet, then list every length, angle, and joint location you must know.
3. Build a measurement plan that pairs commanded actuator lengths with webcam-based pose estimates.
4. Set calibration rules for the camera and the actuator geometry before you collect any trial data.
5. Design a test matrix that changes one factor at a time, such as payload, pose location, or motion path.
6. Plan your error analysis so you can compare predicted pose, measured pose, and repeatability on the same graph.

Common Pitfalls

• Skipping camera calibration, which makes pose estimates drift because lens distortion stays in the image.
• Treating actuator length as exact, which hides backlash and screw play in the kinematics model.
• Mounting the ArUco marker on a flexy surface, which makes the marker move even when the platform does not.
• Comparing only one pose target, which misses how error changes across the full workspace.
• Ignoring joint slop or plate flex, which can make the forward model look wrong when the hardware is the real problem.

What Makes This Competitive

A stronger project goes past simple before-and-after measurements. You can map error across the whole workspace, test more than one correction method, and separate systematic error from random noise. A good entry will also explain why the error happens, not just report that it exists. If you build a clean calibration model and validate it on new poses, your project starts to look like real robotics research.

Project Variations

• Test the same platform with different payload masses to see how deflection changes pose accuracy.
• Compare ArUco-based pose tracking with a second method, such as laser range checks or an external motion sensor.
• Change the control path from straight scans to circular or raster motions, then measure whether path shape changes cumulative error.

Learn More

• OpenCV Documentation: Search the official OpenCV docs for ArUco marker detection and camera calibration guides.
• NASA Technical Reports Server: Search for papers on precision positioning stages, calibration, and motion control.
• MIT OpenCourseWare, Robotics: Search MIT OpenCourseWare for robot kinematics and manipulator motion lectures.
• The International Journal of Robotics Research: Search for review and research articles on parallel robots and Stewart platforms through your library or PubMed-style journal access tools.
• IEEE Xplore: Search for papers on parallel manipulators, pose estimation, and stage accuracy if your school has access.