Maglev Control: Sliding Mode vs LQG

ISEF Category: Engineering Technology: Statics and Dynamics

Subcategory: Control Theory  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A magnet can hold a metal ball in midair, but keeping it stable is the hard part. If the controller reacts too slowly, the ball drops. If it reacts too aggressively, the system can start to chatter and waste power. That makes this project a great test of how smart control can beat raw force.

What Is It?

Magnetic levitation, or maglev, uses an electromagnet to hold an object in place without touching it. In a single-axis ball-floater setup, your job is to keep a small ball hovering at one height. The control system watches the ball’s position, then changes coil current to push or pull the ball back toward the target.

Sliding-mode control is one way to do that. Think of it like a coach that keeps snapping the system back onto a chosen path. LQG, or linear quadratic Gaussian control, is a different strategy. It tries to find a smooth, balanced response using a model of the system and noisy sensor data. Your project compares those two control styles on the same maglev platform and asks which one holds position better, uses less energy, and handles disturbance more cleanly.

Why This Is a Good Topic

This topic works well because you can measure it in clear numbers. You can track position error, current draw, overshoot, settling time, and response to bumps or added disturbances. That gives you a real engineering story, not just a demo. It also connects to real systems like train levitation, precision actuators, and robotics, where stable control matters more than raw power.

Research Questions

• How does sliding-mode control compare with LQG in steady-state position error for a single-axis maglev ball-floater?
• What is the effect of controller choice on coil current consumption during stable levitation?
• Does sliding-mode control reduce disturbance recovery time after a small external push?
• To what extent does chattering change the current waveform under sliding-mode control?
• Which controller produces less position oscillation when the sensor signal contains noise?
• How does the control method affect overshoot when the target levitation height changes?
• To what extent does hand-wound coil variation change controller performance across repeated trials?

Basic Materials

• Single-axis magnetic-levitation ball-floater kit or custom frame with coil and position sensor.
• Hand-wound electromagnet coil with known wire gauge and core material.
• INA219 current sensor module.
• Microcontroller such as Arduino or ESP32.
• Stable DC power supply with current readout.
• Small metal ball or levitated object matched to the platform.
• Breadboard, jumper wires, and screw terminals.
• Digital multimeter.
• Ruler or caliper for checking levitation height.
• Laptop for logging data and tuning code.

Advanced Materials

• Magnetic-levitation test rig with optical or Hall-effect position sensing.
• Function generator for disturbance injection testing.
• Oscilloscope for current ripple and chattering analysis.
• Precision power supply with low-noise output.
• High-bandwidth current probe or inline shunt setup.
• Calibrated reference weights or test masses.
• Data acquisition interface for synchronized logging.
• Temperature sensor for coil heating checks.
• Extra coil builds with different turns, resistance, or wire gauge.
• Vibration isolation base or optical table if available.

Software & Tools

• Arduino IDE: Lets you program the microcontroller and log sensor values from the maglev setup.
• Python: Helps you plot response curves, compare controllers, and calculate error metrics.
• Jupyter Notebook: Keeps your analysis, code, and graphs in one place for easy revision.
• ImageJ: Measures levitation height from video frames if you use camera-based position tracking.
• LibreOffice Calc: Organizes trial data and does quick summaries before deeper analysis.

Experiment Steps

1. Define the exact performance metrics you will compare, such as stability, energy use, disturbance recovery, and chattering.
2. Build a simple plant model of the maglev system, then decide which sensor signals you trust for feedback.
3. Design both controllers around the same hardware limits, so you compare control logic instead of mismatched setups.
4. Plan a calibration method that turns raw sensor output into levitation height and current into energy estimates.
5. Set up disturbance tests that are repeatable, so you can judge recovery time and oscillation fairly.
6. Choose a statistics plan that compares trial-to-trial variation, not just one clean demo run.

Common Pitfalls

• Tuning sliding-mode gain too high, which creates severe chattering and can overheat the coil.
• Comparing controllers with different sensor filters, which makes one controller look better for the wrong reason.
• Ignoring coil resistance changes from heating, which shifts current and alters the baseline over time.
• Measuring disturbance response with inconsistent pushes, which makes recovery data hard to trust.
• Using position data without calibration, which turns height error into a guess instead of a measurement.

What Makes This Competitive

A strong version of this project goes past a simple side-by-side demo. You should compare controllers under matched hardware limits, matched sensing, and matched disturbance tests. Better analysis would separate stability, energy cost, and ripple in the current waveform instead of blending them into one score. A very strong project also explains why one controller wins in some conditions and loses in others, using both data and control theory.

Project Variations

• Compare sliding-mode control with PID instead of LQG to see how a simpler baseline handles the same maglev system.
• Test how different coil winding patterns change stability, current use, and heat buildup under the same controller.
• Add camera-based position tracking and compare optical measurements with the INA219 current data to study sensing tradeoffs.

Learn More

• MIT OpenCourseWare, control systems courses: Search MIT OpenCourseWare for lecture notes on feedback control, state-space models, and LQG.
• NASA Glenn Research Center, control systems resources: Search NASA for educational pages on feedback loops, sensors, and actuator control.
• NIST Digital Library of Mathematical Functions, signal and noise references: Use NIST background material when you need careful definitions for measurement error and filtering.
• PubMed: Search for review articles on magnetic levitation control, actuator dynamics, and sensor noise in engineering systems.
• IEEE Xplore abstracts and papers: Search for magnetic levitation control, sliding-mode control, and LQG comparisons to find methods and metrics.