SynchroSwarm explores decentralized coordination among homogeneous micro-mobile ground robots. Moving away from vulnerable centralized master-node architectures, each swarm agent independently runs distributed consensus protocols, dynamic Voronoi partitioning, and Reciprocal Velocity Obstacle (RVO) kinematic planning. The swarm collaboratively transports oversized payloads and navigates congested dynamic corridors without single-point failure risks.
Centralized multi-robot orchestration systems in automated warehouses (such as automated guided vehicle fleets) fail catastrophically when wireless access points experience latency spikes or packet drops. Furthermore, computational complexity scales exponentially with fleet size, resulting in routing bottlenecks, deadlocks, and costly operational shutdowns.
Build a fleet of 10 autonomous micro-differential drive ground robots equipped with Ultra-Wideband (UWB) relative localization.
Implement an ad-hoc peer-to-peer wireless mesh network running custom collision-free consensus protocols with sub-5ms message broadcast.
Deploy 2D Reciprocal Velocity Obstacles (RVO2) to prevent kinematic deadlocks in dense intersection crossings.
Demonstrate cooperative multi-agent payload transport where multiple robots dock beneath and jointly maneuver an unactuated pallet.
Agents broadcast state vectors (position, velocity, target intent) at 50 Hz via low-latency mesh protocols. Each agent projects neighboring trajectories into velocity-space cones (velocity obstacles), selecting the optimal collision-free acceleration vector that minimizes goal divergence while ensuring bilateral clearance.
- Agent Hardware: Custom dual-motor micro-chassis with precision N20 metal gearmotors, optical encoders (720 pulses/rev), and omnidirectional caster.
- Localization & Mesh: Decawave DWM1000 Ultra-Wideband transceiver delivering ±5cm inter-agent ranging; ESP32-S3 dual-core microcontroller (240 MHz).
- Proximity Sensing: 360-degree ring of 6 time-of-flight (ToF) laser ranging sensors (VL53L1X) detecting uncommunicative obstacles up to 4 meters.
- Power Subsystem: 7.4V 2S Li-Po battery with integrated gas-gauge IC and automated inductive self-docking charging coils.
- Simulation Environment: 100-agent swarm simulated in PyBullet and Gazebo prior to hardware deployment.
- Physical Swarm Testing: 10 physical hardware units deployed across a 6m x 6m structured logistics testbed with dynamic static obstacles.
- Deadlock Analysis: Evaluated across symmetrical head-on collision tests, circular anti-phase swapping, and bottleneck doorway transits.
Robots autonomously negotiate transport priorities via leaderless consensus voting. When a transport task is published to the local mesh, agents form dynamic coalitions based on proximity and battery reserves, converge on the pallet, lock into formation, and synchronize drive torques.
Metrics tracked total mission execution time, energy consumption per transported kilogram, inter-agent minimum clearance margins, and packet-loss resilience during forced RF jamming.
- Inter-Agent Collision Rate: 0.00% across 500 autonomous crossing trials.
- Mesh Synchronization Latency: 3.8 ± 0.4 ms across 10 hops.
- Maximum Cooperative Payload: 8.5 kg transported stably by a 4-agent coalition.
- Reconfiguration Time: < 1.4 seconds when an agent is simulated as abruptly failing.
- Hardware: ESP32-S3, DWM1000 UWB Modules, VL53L1X ToF Sensors, Magnetic Quadrature Encoders.
- Software: FreeRTOS, ESP-IDF, C++17, Python (Swarm Monitor Tool), WebSockets, Plotly Dash.
Hardware open-source PCB design files and 3D printable chassis components.
Embedded C++ Reciprocal Velocity Obstacle (RVO2) library optimized for microcontrollers.
Decentralized consensus and dynamic task allocation firmware protocol suite.
Extensive experimental video repository and quantitative performance documentation.
Next-generation distributed fulfillment centers, automated container port shunting, disaster rubble search, and collective planetary surface regolith hauling.
Distributed autonomous systems must guarantee fail-safe state transitions during communication degradation. All agents implement passive dynamic braking upon consecutive packet timeout thresholds.
A fully operational, completely decentralized micro-robot swarm demonstrating robust obstacle avoidance and collaborative multi-agent logistics without requiring a master server or central Wi-Fi router.
Extending algorithms to 3D aerial drone swarms, incorporating visual relative pose estimation via onboard micro-cameras, and exploring flocking algorithms in GPS-denied turbulent outdoor environments.
Masterless multi-agent robotics orchestrating resilient warehouse logistics through decentralized consensus.
A decentralized swarm robotics platform combining peer-to-peer mesh communications with Reciprocal Velocity Obstacle algorithms to achieve autonomous, collision-free collaborative logistics without centralized servers.
