Smart Cane Tip Surface Detection

ISEF Category: Embedded Systems

Subcategory: Sensors  ·  Difficulty: Intermediate  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

A cane can tell you more than where the ground is. A tiny sensor pack at the tip can feel vibration, pressure, and tilt changes that hint at carpet, tile, gravel, or stairs. That means feedback can reach the user without a camera at all. Your project asks a simple question, can a low-cost tip sensor learn the ground under it?

What Is It?

This project turns a cane tip into a surface detector. You use an inertial measurement unit, or IMU, which tracks motion and tilt, a force-sensitive resistor, which measures how hard the tip presses, and a barometer, which can help detect changes in height or step-like transitions. The goal is to sort walking surfaces into classes such as carpet, tile, gravel, and stairs.

Think of it like giving the cane a sense of touch. Carpet feels soft and damped. Tile feels hard and smooth. Gravel gives messy, uneven signals. Stairs can create a sudden change in pitch, pressure, or height pattern. Your job is to collect those patterns, label them, and see whether a simple model can tell them apart.

Why This Is a Good Topic

This makes a strong science fair topic because you can test real sensor signals, compare classes, and measure accuracy. You do not need a camera or a huge budget. You can build a small system, collect data from different surfaces, and ask whether vibration, pressure, and motion data are enough to classify terrain. The project also connects to accessibility, which gives it a clear real-world purpose.

Research Questions

• How does adding force data to IMU data change surface classification accuracy?
• What is the effect of using barometer readings on stair detection accuracy?
• Does sensor placement near the cane tip improve classification compared with placement higher up the shaft?
• To what extent can a low-cost model distinguish carpet from tile using vibration and pressure patterns?
• Which features, such as peak force, signal variance, or tilt change, best predict gravel versus hard floors?
• How does the number of surface samples used for training affect model performance?

Basic Materials

• Microcontroller board such as Arduino or ESP32.
• IMU sensor module with accelerometer and gyroscope.
• Force-sensitive resistor and matching resistor for a voltage divider.
• Barometer sensor module compatible with your board.
• Small vibration motor or coin vibration motor.
• Rechargeable battery pack or USB power bank.
• Breadboard and jumper wires.
• Cane, walking stick, or cane-sized test rod.
• Notebook or spreadsheet for labeled data.
• Tape, zip ties, or a small enclosure for mounting sensors.
• Phone camera or stopwatch for syncing trials and notes.

Advanced Materials

• Microcontroller board with data logging support.
• IMU sensor module with high sample-rate output.
• Force-sensitive resistor or load cell with amplifier.
• Barometer or pressure sensor module.
• SD card logging module.
• Compact vibration motor or haptic driver.
• 3D-printed tip mount or machined enclosure.
• Bench power supply for bench testing.
• Reference scale or calibrated force rig.
• University lab access for repeatability testing across multiple floor samples.
• Surface test panels for carpet, tile, gravel, and stair tread materials.

Software & Tools

• Arduino IDE: Programs the microcontroller and records sensor output.
• Python: Cleans data, builds features, and trains simple classifiers.
• Jupyter Notebook: Lets you compare sensor signals and model results in one place.
• ImageJ: Measures cane tip wear or surface contact marks if you document the setup with photos.
• Excel: Organizes labeled trials and basic summary statistics.

Experiment Steps

1. Define the surface classes you will test and the exact motion pattern the cane will follow on each one.
2. Choose sensor signals that should change across surfaces, then plan how you will log them together.
3. Design a labeling system so every trial matches one surface, one walking condition, and one sensor file.
4. Build a feature set that turns raw sensor streams into numbers a model can compare.
5. Plan a baseline classifier first, then test whether extra sensors improve accuracy over a simpler setup.
6. Decide how you will validate results with separate test data so your model does not memorize the training set.

Common Pitfalls

• Holding the cane at different angles each trial, which changes IMU readings more than the surface does.
• Letting the sensor mount wobble, which adds fake vibration patterns that look like rough ground.
• Mixing stair data with flat-surface data, which makes the classifier learn height changes instead of true surface texture.
• Training on one hallway or one room, which causes the model to fail on a new floor with different acoustics and friction.
• Ignoring class imbalance, which makes the model look accurate while it barely recognizes the rarest surface type.

What Makes This Competitive

A stronger version of this project goes past a simple yes-or-no classifier. You can compare sensor combinations, test different placements, and report confusion matrices that show which surfaces get mixed up. You can also ask whether the system works across multiple users, multiple canes, or multiple buildings. If you explain why the model succeeds or fails, your project starts to look like real engineering research.

Project Variations

• Test whether the same sensor setup can distinguish wet tile from dry tile, which adds a real-world slip-risk angle.
• Compare a tip-mounted sensor package with a shaft-mounted package to see which location gives cleaner surface signals.
• Swap the hand-coded threshold logic for a simple machine learning model and compare accuracy, speed, and interpretability.

Learn More

• NIH PubMed: Search for review articles on gait, mobility aids, and haptic feedback for visual impairment.
• NASA Open Data Portal: Useful for learning how to handle sensor data streams and metadata labeling.
• MIT OpenCourseWare: Search for embedded systems and sensor data courses that explain signal conditioning and microcontroller design.
• USGS Publications Warehouse: Good for background on surface materials, terrain sensing, and field measurement methods.
• IEEE Xplore: Search for peer-reviewed papers on cane sensors, haptic feedback, and terrain classification.