Multi-Robot Farm Swarm

A multi-agent system where several ROS 2 rovers cooperate over a DDS-based discovery layer and an MQTT coordination bus. Robots share occupancy maps, crop-health observations, and task claims so the swarm partitions a field, avoids redundant coverage, and merges sensor data into a shared world model. LiDAR and Jetson-based perception run on each node.

A single rover is a bottleneck on large fields, and naive multi-robot setups duplicate work or collide. Coordinating several autonomous robots so they share maps, divide tasks, and avoid redundant coverage is a hard distributed-systems problem.

Each rover runs a ROS 2 stack with LiDAR and Jetson perception. DDS handles peer discovery and map sharing, while an MQTT coordination bus carries task claims and crop-health observations. A distributed task-allocation scheme partitions coverage and merges per-robot occupancy grids into a shared world model.

Planned — to be measured in simulation and field trials: coverage speedup versus a single rover, map-synchronization latency, and task-allocation efficiency as the swarm scales.

1. Phase 1

   Single-agent base

   Stand up a ROS 2 rover with LiDAR mapping and Jetson perception.

2. Phase 2

   Shared world model

   Use DDS to share and merge occupancy maps across robots.

3. Phase 3

   Coordination

   Add an MQTT task-allocation layer for coverage partitioning and conflict avoidance.

4. Phase 4

   Swarm trials

   Evaluate speedup, sync latency, and scaling across multiple robots.