FedRobo: Federated Learning Driven Autonomous Inter Robots Communication For Optimal Chemical Sprays

TL;DR

The paper tackles the data-privacy and communication-efficiency challenges of coordinating autonomous spraying robots in agriculture by deploying a cluster-based federated learning framework. It leverages $FedAvg$ aggregation with $FedSRC$ checkpointing to minimize unnecessary client participation while enabling inter-robot knowledge sharing. The study evaluates weed-detection capabilities using YOLOv7, SSD-300, and Faster R-CNN within a federated setting across 150 client nodes and 38,089 images, incorporating web-crawled disease data to enhance updates. Results indicate feasible, privacy-preserving collaboration that can reduce chemical usage and adapt to environmental changes, with discussions on security, reliability, and future improvements in group formation and attack resilience.

Abstract

Federated Learning enables robots to learn from each other's experiences without relying on centralized data collection. Each robot independently maintains a model of crop conditions and chemical spray effectiveness, which is periodically shared with other robots in the fleet. A communication protocol is designed to optimize chemical spray applications by facilitating the exchange of information about crop conditions, weather, and other critical factors. The federated learning algorithm leverages this shared data to continuously refine the chemical spray strategy, reducing waste and improving crop yields. This approach has the potential to revolutionize the agriculture industry by offering a scalable and efficient solution for crop protection. However, significant challenges remain, including the development of a secure and robust communication protocol, the design of a federated learning algorithm that effectively integrates data from multiple sources, and ensuring the safety and reliability of autonomous robots. The proposed cluster-based federated learning approach also effectively reduces the computational load on the global server and minimizes communication overhead among clients.

Figures (4)

• Figure 1: The architecture of autonomous robots communicating with the global server via local server station.
• Figure 2: An overview of federated learning architecture, where different colors represent different farm fields across the country.
• Figure 3: Architecture showing how checkpointing is performed based on two model updates for two different nodes or autonomous robots.
• Figure 4: List of all different classes used to train the plant weed detection model and later deploy the updates in autonomous robots through the global server