Painter Robot for Calligraphy-Style Images

ISEF Category: Technology Enhances the Arts

Subcategory: Engineering Effects  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A robot can turn a photo into something that looks hand-painted, not just printed. That matters because people read texture, stroke order, and spacing as part of the art itself. Your project asks a simple question with a hard answer, can a machine make marks that people judge as more artistic than a flat raster print?

What Is It?

This project blends robotics, computer vision, and art. The robot does not paint by copying pixels one by one. Instead, a planner chooses a path for brush strokes, kind of like a player picking moves in a strategy game. The goal is to make the final image feel more like calligraphy or brush art, where stroke shape, direction, and rhythm matter.

The 3-DoF gantry gives the robot three main movements, usually left-right, forward-back, and up-down or brush control. A servo on the brush changes pressure or contact. Reinforcement learning trains the planner by trial and error in simulation. In simple terms, the software gets feedback on whether a stroke sequence looks closer to the target style, then it keeps improving its choices. You can compare that system with a raster baseline, which is the plain printer-like method that fills pixels without thinking about brush behavior.

Why This Is a Good Topic

This is a strong science fair topic because you can test both engineering performance and human judgment. You can measure things like path accuracy, stroke smoothness, image similarity, and viewer ratings. That gives you real data, not just a cool demo. The project connects to robotics, digital art, assistive fabrication, and human-computer interaction. A student can learn simulation, control, evaluation design, and basic statistics while still working on one clear question.

Research Questions

• How does stroke-based planning affect viewer aesthetic ratings compared with raster printing?
• What is the effect of brush pressure control on perceived calligraphy quality?
• Does reinforcement learning in simulation improve path efficiency versus a fixed hand-coded planner?
• To what extent does stroke ordering change ratings of visual balance and flow?
• Which image types produce the largest gap between robot painting and raster output?
• What is the effect of adding stroke curvature constraints on final image similarity and perceived artistry?

Basic Materials

• 3-DoF gantry robot kit or custom Cartesian plotter with brush mount.
• Brush servo for contact control.
• Acrylic or watercolor brush set with consistent tip size.
• Plain paper, cardstock, or test canvas sheets.
• Smartphone or digital camera for image capture.
• Tripod or fixed camera mount for repeatable photos.
• Computer for simulation, planning, and analysis.
• Free drawing or image-editing software for making target images.
• Calibrated ruler or printed grid sheet for alignment checks.
• Digital kitchen scale for tracking ink or paint use if needed.

Advanced Materials

• Custom 3-DoF gantry with higher repeatability and encoder feedback.
• Force or pressure sensor at the brush tip.
• Motor driver board with logging support.
• Controlled lighting box for image capture and rating sessions.
• Stylus tablet or motion capture tool for collecting human stroke references.
• Simulation software for robot path planning and reward testing.
• Computer with Python and machine learning libraries.
• High-resolution camera for documenting stroke texture.
• Image analysis targets or color calibration card.
• Standardized art surfaces with known texture and absorbency.

Software & Tools

• Python: Runs the planner, training scripts, and data analysis for your robot system.
• ImageJ: Measures stroke width, contrast, and image similarity from photos of finished paintings.
• OpenCV: Processes images, finds edges, and helps compare robot output to target art.
• PyBullet: Lets you simulate robot motion and test stroke strategies before hardware trials.
• Google Colab: Gives you free cloud computing for lighter machine learning experiments and quick notebooks.

Experiment Steps

1. Define the art quality you will measure, such as viewer ratings, stroke smoothness, or image similarity.
2. Choose one target style and one baseline method so your comparison stays fair.
3. Build a simulation model that lets the planner test stroke paths before you run hardware trials.
4. Set controls for paper type, brush type, lighting, and camera angle so your measurements stay consistent.
5. Plan a scoring method that combines human ratings with objective image metrics.
6. Organize repeated trials so you can compare the planner, the baseline, and any ablation tests.

Common Pitfalls

• Using uncontrolled lighting when photographing paintings, which changes color and contrast scores between trials.
• Letting brush wear vary across runs, which makes stroke width and texture look inconsistent.
• Comparing the robot against a weak raster baseline, which makes the performance gap meaningless.
• Training only on a tiny set of target images, which causes the planner to fail on new styles.
• Relying on one viewer's taste, which turns aesthetic scoring into personal opinion instead of usable data.

What Makes This Competitive

A stronger version of this project goes beyond a pretty robot demo. You would need clean controls, a fair baseline, and metrics that separate style quality from simple image accuracy. You could also test whether human ratings match objective measures, or whether one kind of stroke planning works better for certain image classes. A project like that shows real engineering judgment, not just coding or building.

Project Variations

• Test the same painter robot on watercolor-style landscapes instead of Chinese-calligraphy-style images.
• Replace reinforcement learning with a rule-based stroke planner and compare both methods on aesthetic ratings.
• Analyze whether different viewer groups, such as artists and non-artists, rate the robot paintings differently.

Learn More

• MIT OpenCourseWare: Search for robotics, control systems, and machine learning courses that cover path planning and feedback control.
• PyBullet documentation: Learn how to simulate robot motion and test planning ideas before hardware work.
• OpenCV documentation: Find image processing tools for measuring stroke edges, contrast, and similarity.
• NIH PubMed: Search for review articles on human aesthetic judgment, visual perception, and art evaluation methods.
• NASA Open Data Portal: Explore examples of image analysis workflows and calibration thinking used in real data projects.
• International Journal of Arts and Technology: Search for peer-reviewed articles on creative robots, digital fabrication, and art-making systems.