Swarm Robot Consensus and Collision Avoidance

ISEF Category: Engineering Technology: Statics and Dynamics

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

The Hook

Five robots can fail as a team even when each one works fine alone. A small radio glitch can ripple through the whole swarm. That makes swarm control a great science fair problem, because you can measure how information loss changes group behavior. You get to test both coordination and safety at once.

What Is It?

This project studies how a small group of robots makes decisions together when each robot only has partial information. In a swarm, no single robot acts like the boss. Each bot follows local rules, like keep a certain distance, match heading, and move toward a target. That local rule set can create a group pattern, much like birds flocking or ants following a trail.

Your version uses 5 differential-drive Pi-Bots, which means each robot moves with two driven wheels. ESP-NOW is a radio link that lets the robots send short messages to each other without a full Wi-Fi network. Packet loss means some messages never arrive. That is a good stress test, because real robots often lose data, lag, or get noisy sensor readings. You can then ask whether the swarm still keeps formation, avoids collisions, or reaches an obstacle target when communication gets worse.

Why This Is a Good Topic

This is a strong science fair topic because you can change one thing, packet loss, and measure how the whole team responds. That gives you clear variables, clear outputs, and real graphs. The project connects to warehouse robots, search-and-rescue swarms, and autonomous vehicle coordination. You can also learn control theory, network effects, data logging, and performance metrics without needing a university lab.

Research Questions

• How does packet-loss rate affect formation error in a 5-robot differential-drive swarm? ?
• What is the effect of communication delay on time to reach a rendezvous target? ?
• Does adding a collision-avoidance rule reduce near-miss events under packet loss? ?
• To what extent does leader-follower control outperform distributed consensus when messages drop? ?
• Which formation pattern stays most stable when radio packets are lost? ?
• How does obstacle density change the swarm's ability to keep formation while moving? ?

Basic Materials

• 5 differential-drive Pi-Bots or similar small mobile robots.
• Raspberry Pi or onboard controller for each robot.
• ESP-NOW-capable radios or microcontrollers that support ESP-NOW.
• Laptop for code upload, logging, and analysis.
• Basic obstacle set, such as cones, boxes, or foam blocks.
• Measuring tape or floor grid for tracking distance and position.
• Large flat test area with visible boundary lines.
• Smartphone or overhead camera for video recording.
• Spare batteries and chargers for every robot.
• Masking tape for marking start zones and target zones.

Advanced Materials

• Motion capture system or overhead tracking camera.
• IMU sensors for heading and turn-rate logging.
• Wheel encoders with synchronized timestamp logging.
• Wireless packet logger for message drop analysis.
• External microcontroller test boards for comparing communication stacks.
• 3D-printed chassis parts for repeatable robot geometry.
• Lidar or depth camera for richer obstacle sensing.
• Data acquisition script for time-aligned robot state logs.
• Calibration jig for wheel radius and wheelbase checks.
• Safety-rated battery monitoring hardware.

Software & Tools

• Python: Cleans robot logs, computes formation error, and makes plots.
• ImageJ: Measures distances and angles from overhead video frames.
• OpenCV: Tracks robot position from camera footage when markers are visible.
• GNU Octave: Fits simple control models and compares conditions.
• Plotly: Builds clear interactive charts for packet loss and swarm performance.

Experiment Steps

1. Define the one control rule set you will test first, such as leader-follower or distributed consensus.
2. Choose one outcome metric for each task, such as formation error, collision count, or rendezvous time.
3. Plan a way to vary packet loss in controlled levels so each trial is comparable.
4. Design baseline and stress-test conditions that separate communication problems from motion problems.
5. Decide how you will record robot state, radio events, and video so you can align them later.
6. Build a comparison plan that tests whether obstacle load or packet loss hurts performance more.

Common Pitfalls

• Measuring success only by whether the swarm finished the task, which hides small but important formation errors.
• Letting each robot start with slightly different wheel calibration, which makes communication effects look worse than they are.
• Changing the obstacle layout between trials, which confounds packet loss with map difficulty.
• Logging radio drops without time stamps, which makes it hard to match packet loss to robot behavior.
• Using a crowded control rule with too many behaviors at once, which makes it impossible to tell why a failure happened.

What Makes This Competitive

A competitive version of this project would go past simple demo behavior and quantify how communication loss changes stability, safety, and task success. You would compare at least two coordination strategies, then test them across several packet-loss levels and obstacle layouts. Strong work also includes clean metrics, repeated trials, and a failure analysis that explains when the swarm breaks down and why. If you add a comparison between observed behavior and a simple control model, your project gets much stronger.

Project Variations

• Test how different formation shapes, such as line, V, or circle, change resilience to packet loss.
• Compare radio-based coordination with a vision-only fallback when messages drop.
• Study how adding obstacle density changes the balance between collision avoidance and formation keeping.

Learn More

• NASA Technical Reports Server: Search for swarm robotics, multi-agent coordination, and autonomous system testing papers.
• NIH PubMed: Search review articles on consensus control, multi-robot coordination, and distributed robotics.
• NOAA Digital Coast: Find articles and datasets on sensor networks, mapping, and field data collection workflows.
• MIT OpenCourseWare: Look for free control systems and robotics lecture notes that cover feedback, stability, and state estimation.
• IEEE Xplore abstracts: Search abstracts for swarm robotics and consensus control to learn current research language.