Swarm Robot Consensus Under Packet Loss

ISEF Category: Robotics and Intelligent Machines

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

The Hook

A robot swarm can fail even when every bot works perfectly. One dropped message can throw off the whole group. That makes packet loss a real test of control, not just coding. If you can keep the swarm together anyway, you are solving the same kind of problem used in drones, warehouse robots, and sensor networks.

What Is It?

This project studies how a group of small robots can agree on where to go, even when messages do not always get through. In control theory, that shared agreement is called consensus. Think of it like a group of hikers trying to meet at one trail junction. If everyone follows the same rule, the group should still gather, even when a few calls are missed.

Your bots can use local rules instead of a single boss robot. Each robot listens to nearby peers, updates its motion, and tries to reduce the difference between its own state and the group state. Rendezvous means the robots try to meet at one point. Formation means they try to keep a shape, like a line, triangle, or ring, while moving together.

The communication link matters a lot. ESP-NOW is a wireless protocol that sends short messages between devices with low delay. Packet loss means some messages never arrive. Your job is to measure how much loss the controller can handle before the swarm gets slower, less stable, or fails to form the target shape.

Why This Is a Good Topic

This is a strong science fair topic because you can test a real control system and change one variable at a time. Packet loss gives you a clean independent variable, and convergence time gives you a clear measured outcome. The project connects to drones, rescue robots, and distributed sensor networks, so the real-world link is easy to explain. You can also learn motion control, wireless communication, data logging, and statistical analysis without needing a university lab.

Research Questions

• How does packet-drop probability affect rendezvous convergence time in a 5 to 10 robot swarm?
• What is the effect of swarm size on the controller's ability to reach a shared point under the same packet-loss rate?
• Does adding a fallback motion rule improve formation stability when ESP-NOW packets are missed?
• To what extent does initial spacing between robots change final formation error after intermittent communication loss?
• Which formation shape, line, triangle, or ring, stays most stable under the same dropout pattern?
• How does message update rate affect the number of failed consensus runs at a fixed packet-drop probability?

Basic Materials

• 5 to 10 differential-drive robot kits or DIY robot chassis with motors and motor drivers.
• ESP-NOW-capable microcontrollers, such as ESP32 boards.
• Battery packs matched to each robot.
• A laptop for code upload, data logging, and plotting.
• Measuring tape or marked floor grid for path and position checks.
• Painter's tape or floor markers to define start zones and targets.
• Spare wheels, wires, and connectors for quick repairs.
• Stopwatch or video camera for timing runs.
• Digital kitchen scale for checking robot mass balance.
• Colored stickers or markers to identify each robot.

Advanced Materials

• Motion-capture access or overhead camera tracking system.
• High-frame-rate camera for position extraction.
• Calibration grid for camera-based coordinate mapping.
• Wi-Fi packet logging tools or firmware-level telemetry logs.
• IMU sensors for heading and motion feedback.
• Laser-cut or 3D-printed chassis parts for matched robot geometry.
• Current and voltage sensors for power draw analysis.
• External antenna setups for range testing.
• MATLAB or Python-compatible data export workflow.
• Spare microcontrollers and motor drivers for failure testing.

Software & Tools

• Arduino IDE: Uploads firmware to ESP32 boards and lets you test swarm control code.
• Python: Processes run data, computes convergence time, and makes plots.
• ImageJ: Tracks robot positions from overhead video frame by frame.
• OpenCV: Extracts robot paths from video and supports custom tracking scripts.
• GeoGebra: Helps you sketch formations and compare expected shapes with results.

Experiment Steps

1. Define the swarm task you want to test first, such as rendezvous or a fixed formation.
2. Choose the one network condition you will change, such as packet-drop probability, while keeping the motion rule constant.
3. Build a position-tracking plan so you can turn robot motion into numbers, not just video clips.
4. Design controls that separate communication failure from motor drift, battery sag, and bad starting layouts.
5. Set up a scoring method for convergence time, final shape error, and run-to-run stability.
6. Plan a comparison between at least two controller versions so you can test whether your fallback rule really helps.

Common Pitfalls

• Using robots with mismatched wheel speed, which makes control errors look like communication failure.
• Measuring success by eye instead of extracting positions from video, which hides small formation errors.
• Changing start spacing between trials, which makes convergence time comparisons unfair.
• Ignoring battery voltage drop, which can slow some robots and distort packet-loss results.
• Testing only one dropout pattern, which makes the controller look better or worse than it really is.

What Makes This Competitive

A stronger version of this project goes past simple success or failure. You can map a full performance curve, then compare multiple controller designs under the same loss pattern. You can also test whether one formation shape is more resilient than another, or whether the swarm keeps working after the packet loss becomes bursty instead of random. Careful tracking, clean controls, and a clear stability metric can make the work feel much more like research than a class demo.

Project Variations

• Test the same consensus controller on a line formation, a triangle, and a ring to see which shape resists packet loss best.
• Replace random packet loss with bursty dropouts to model real wireless interference and compare convergence time.
• Add one robot with delayed replies and measure whether the swarm still reaches rendezvous without a central leader.

Learn More

• MIT OpenCourseWare: Search for introductory controls, feedback systems, and robotics courses that explain consensus and stability.
• NASA Technical Reports Server: Search for swarm robotics, distributed control, and multi-agent coordination reports.
• IEEE Xplore: Search for review articles on consensus control, formation control, and communication loss in robot swarms.
• PubMed: Search for papers on human-swarm interaction if you want a broader systems angle.
• OpenCV Documentation: Read the official tutorials for video tracking and coordinate extraction.
• Python Scientific Computing Documentation: Use NumPy, SciPy, and Matplotlib docs for data analysis and plotting.