FarmFlock

What it does

FarmFlock employs a swarm of mobile robots equipped with various sensors for obtaining data about flock health. The data is then published for the different parts of the farm to a web application where a farm operator can review and take appropriate actions.

Your inspiration

The US slaughtered 5.3 million chickens in one facility during the April 2022 avian flu pandemic, following the loss of 22 million chickens across various states due to delayed disease detection. These losses are not only inhumane, but also financially disastrous, resulting in the loss of nearly 5 million jobs annually.  Constant medication, particularly antibiotics, has risks as a control approach. They leave residues in the birds that, when eaten, can lead to the development of antibiotic resistance in humans, reducing protection against illnesses. The goal of this project was to help with food security.

How it works

FarmFlock is a swarm of mobile robots which roam autonomously within the poultry farm, collecting critical data related to flock health, such as ammonia levels, temperature, moisture, behavior, and carbon dioxide levels. Utilizing the Firefly Algorithm, the robots are programmed to move intelligently around the farm. When one robot detects an environmental parameter that crosses a threshold, it signals others to converge on that area for a comprehensive data assessment. This ensures rapid response to any potential health threats. Additionally, the robots analyze the rate of deaths of chicken within the farm. Data gathered is wirelessly transmitted in real-time to a centralized web application. Farm operators can easily access this platform, allowing them to review the data, pinpoint areas of concern, and initiate timely interventions, be it isolation or vaccination. The overarching aim is to quickly identify potential outbreaks and minimize losses.

Design process

The design process for the FarmFlock project involved several stages and iterations to develop an efficient and cost-effective solution. Initially, a prototype utilizing the ROSbot 2 PRO was employed, integrating ROS and a LiDAR for spatial analysis and navigation. To test the required sensors, a circuit incorporating MQ-3, MQ-135, MQ-137, DHT-11, and PIR sensors was constructed. This circuit facilitated the collection of environmental data, which was stored on a web server. Limitations arose due to the high cost and large size of the ROSbot 2 PRO, which would disrupt movement inside the farm. Following this, multiple low-cost robot prototypes were created using a Raspberry Pi 4 for the computational ability of the robot. They equipped with ROS and established communication through a common ROS topic, enabling coordination for the implementation of the Firefly Algorithm. Initial analysis of using a single robot proved low efficiency, hence the idea of a swarm was implemented. For understanding the navigation around a mobile flock, there was the use of a custom laser pointer module that was 3D printed. The incorporation of swarm robotics allowed for enhanced efficiency and effectiveness in farm environments allowing for improved data collection around large farms.

How it is different

When comparing FarmFlock to similar products in the market, the project offers distinct advantages over ceiling-mounted sensor systems commonly used in poultry farms. FarmFlock's robot design positions the sensors closer to the ground, resulting in improved measurement accuracy and precision. This proximity allows for more reliable data collection, benefiting the overall monitoring and management of the poultry farm environment. Furthermore, FarmFlock introduces innovative algorithms for detecting deceased chickens, a feature not found in existing systems. This capability enhances animal welfare and facilitates prompt intervention in case of unexpected events, contributing to the overall health and productivity of the poultry flock. Additionally, the use of swarm behavior within poultry farms is novel and shows an increased efficiency in comparison to existing solutions.

Future plans

One future idea is the use of sound analysis to identify respiratory diseases in the poultry farm, leveraging bioacoustics and vocalizations to detect early signs of illness or stress. Another enhancement involves employing Multilayer Perceptrons (MLPs) to predict disease based on deviations from baseline parameters. MLPs trained on historical farm data can reduce manual monitoring, optimizing resource allocation and lowering operational costs. Finally, expanding the sensor network to gather more data points, including camera-vision for behavioral analysis and additional sensors, can enhance the system's robustness.

Awards

This project won the best final year project award within the cohort. It was awarded The Phoenix Contact Middle East- Automation Award. There was a conference paper submitted to a IEEE journal on this project and it was awarded the Best Paper award. This project is also pending the publication of one more conference paper.