Battery State-of-Charge Estimation for a Rover Pack

ISEF Category: Robotics and Intelligent Machines

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

The Hook

Your rover can look fully powered and still be one hill away from dying. Battery percent is not a simple meter reading, it is a prediction problem. If you can estimate state of charge well, your robot can plan better paths, avoid brownouts, and finish missions. That makes this a real control systems project, not just a battery test.

What Is It?

State of charge, or SoC, means how much usable energy is left in a battery. Think of it like a gas gauge that you cannot read directly. You have to estimate it from clues. For a rover, those clues usually include voltage, current, and temperature.

A simple method is coulomb counting. That means you add up how much current leaves the battery over time. It sounds easy, but small sensor errors build up fast. An extended Kalman filter is a smarter estimator. It combines a battery model with measurements and keeps correcting itself. A nonlinear observer does a similar job, but it uses control theory to track the hidden battery state more directly.

Why This Is a Good Topic

This makes a strong science fair topic because the problem is real, measurable, and full of room for comparison. You can test whether a simple estimator, like coulomb counting, falls apart under stop-and-go rover use, then see if a model-based observer does better. That connects directly to robotics, power systems, and autonomous vehicles. You can also learn model fitting, sensor error, and performance metrics without needing to invent a brand-new battery chemistry.

Research Questions

• How does discharge profile shape affect state-of-charge error for coulomb counting?
• What is the effect of temperature changes on observer accuracy for an 18650 pack?
• Does a nonlinear observer track state of charge more accurately than an extended Kalman filter under pulsed rover loads?
• To what extent does sensor noise in voltage and current measurements change estimation drift over time?
• Which battery model structure gives the lowest state-of-charge error across repeated discharge cycles?
• How does starting from the wrong initial state of charge affect each estimator?

Basic Materials

• 18650 cell pack with holder or protected pack.
• Small rover chassis or electronic load that can create changing discharge profiles.
• INA219 current and voltage sensor.
• Microcontroller such as Arduino or Raspberry Pi Pico.
• Digital thermometer or thermistor for pack temperature.
• Multimeter for verification measurements.
• Resistors, wiring, and breadboard or prototyping board.
• Laptop for logging and analysis.
• Safety gear, including eye protection and a fire-safe charging area.

Advanced Materials

• Matched 18650 cells with battery management support.
• Programmable electronic load or motor test stand.
• Precision shunt resistor and calibrated current measurement setup.
• Thermocouples or temperature sensors placed at the cell surface and pack surface.
• Data acquisition system with synchronized sampling.
• Environmental chamber or controlled-temperature space.
• Reference battery cycler for repeatable discharge testing.
• Lab power supply with current limit.
• Fire-resistant battery enclosure and charging equipment.

Software & Tools

• Python: Fits battery models, filters sensor data, and compares estimator error across discharge profiles.
• Jupyter Notebook: Lets you document each analysis step and keep plots next to your code.
• MATLAB or GNU Octave: Supports state-space modeling, observer design, and filter testing.
• ImageJ: Not needed for the core analysis, so skip it unless you make thermal image maps.
• Excel or Google Sheets: Helps you inspect raw data, label trials, and check for logging mistakes.

Experiment Steps

1. Define the battery state you want to estimate, then choose one clear accuracy metric for comparison.
2. Build a simple battery model that connects voltage, current, temperature, and state of charge.
3. Plan three estimators, one simple baseline, one observer, and one filter, so you can compare them fairly.
4. Design discharge profiles that mimic rover motion, such as smooth driving, stop-and-go motion, and heavy load bursts.
5. Decide how you will validate each estimate against a reference capacity measurement or a known end-of-discharge point.
6. Set up analysis plots that show error over time, not just one final number.

Common Pitfalls

• Using the battery's terminal voltage as state of charge without correcting for load, which makes the estimate look better or worse depending on the current draw.
• Letting sensor offsets accumulate in coulomb counting, which causes the estimate to drift farther away with every trial.
• Testing only one discharge pattern, which hides how badly the estimator fails on rover-like pulses.
• Ignoring temperature, which can shift voltage behavior enough to break model-based predictions.
• Comparing methods with different starting state-of-charge guesses, which makes the results unfair.

What Makes This Competitive

A stronger project does more than report which estimator had the smallest error. You can test multiple load profiles, include temperature effects, and show where each method fails, not just where it works. Better projects use clear validation, uncertainty analysis, and fair tuning rules for every estimator. If you compare performance across realistic rover conditions, your study starts to look like control research instead of a classroom demo.

Project Variations

• Test the same estimators on a single cell instead of a pack to isolate model error from balancing effects.
• Compare rover-style pulsed loads with steady loads to see which profile breaks each estimator first.
• Add temperature as a separate input feature and test whether it improves state-of-charge tracking across warm and cool runs.

Learn More

• NASA Battery 101 materials: Search NASA for battery basics, safety guidance, and space battery testing concepts.
• NIH PubMed: Search review articles on lithium-ion state-of-charge estimation and observer design.
• MIT OpenCourseWare: Look for control systems courses with state-space modeling and observer lectures.
• USGS battery safety resources: Search for public guidance on lithium-ion handling and fire risk.
• Journal of Power Sources: Search for review papers on battery state-of-charge estimation and model-based filters.