This is are the individual components for the Eco Data Collector
The Camera Module 3 NoIR is used for capturing 720p 120 frames per second video, allowing us to take video during the dark and during the day.
A real-time clock module is included to keep accurate time and date when we turn off the wifi. This is crucial for timestamping the collected data, ensuring that all information is accurately logged and synchronized.
To prevent overheating and ensure the longevity of the overclocked device, a heat sink is incorporated. This component helps dissipate heat generated by the Raspberry Pi, maintaining optimal operating temperatures for high quality video capturing.
This is an optional component we will be offering that allows the ability to do inference on the device. Other options are available.
These are the batteries we are currently using, our current model contains three Lithium Ion batteries that supply 10-18 Amp hours
Infrared lights are included to enable night-time video capture. The infrared lights contain a light sensor that only has the infrared lights on during the night as a way to save on power and energy.
A Samsung 512GB microSD card is included for extensive data storage. This high-capacity memory card ensures that the device can store large amounts of data over extended periods.
The Raspberry Pi 2 W is a compact and efficient microcontroller, serving as the central processing unit for the Eco Data Collector. Everything is going to be essentially connected to this device.
These are the solar panels we are going to be using, we are currently using a total of 9 solar panels
Using the solar panels, the 5 volt solar power manager, charges the device and batteries during the day. During the night time the batteries discharge and are charging the device.
This is an optional multi-function environmental module that can be used to detect temperature, humidity, barometric pressure, altitude, TVOC and eCO2.
This is an optional Temperature & Humidity sensor housed in a weather-proof enclosure that we are offering.
A terrarium is a miniature, sealed ecosystem contained within a glass container. These systems rely solely on solar energy to sustain the plants, fungi, and small animals within them. The natural water and life cycles are balanced in such a way that they can thrive for extended periods sometimes even decades without external intervention.
Now, let's contrast this with traditional data centers. They're typically large and complex needing continuous electricity, water cooling, and bandwidth for data transmission to operate properly. This consumption of resources leads to substantial environmental impacts, making sustainability a significant concern for the industry. Instead, we are providing a net zero energy with a closed-loop cooling system, that's terrarium-inspired.
Our data terrarium prototype balances performance with energy efficiency, to perform various data processing tasks sustainably. Starting with the central processing unit, our prototype utilizes the Intel N100 CPU. Equipped with four cores it can multitask while being energy efficient. We are also looking into other energy-efficient solutions such as ARM-based CPUs. Our prototype has 8GB of RAM and a 64GB boot drive. Additionally, our prototype is equipped with up to 16TB of SSD storage. This substantial storage capacity allows the data terrarium to handle datasets, store extensive amounts of data, and support data-intensive applications. With SSDs rather than traditional hard drives it can access data more quickly and reliably even (especially feels weird) in extreme temperatures. In terms of power consumption, at only 6 to 35 watts our prototype is highly efficient. This is key to the sustainability of the data terrarium effectively operating with solar power and lithium phosphate batteries. The wide range ensures the flexibility to increase the number of solar panels to provide up to a KW, accommodating various workloads while optimizing energy usage.
Now, let's focus on the GPU integrated into our data terrarium prototype. This component is critical for handling the more intensive computational tasks that our system will perform. We have chosen the NVIDIA L4 GPU for our prototype. Known for it's exceptional yet power-efficient performance. The 24 GB of GPU RAM at 72 watts provides substantial computational power and range, performing demanding applications, such as AI acceleration, energy-efficient execution of large-scale computation, genomics, and real-time processing, at low power. Certain cases where the higher computational task is prioritized, we can substitute the Nvidia L4 with Nvidia’s A16 which has higher RAM per watt at 64 GB of GPU RAM, allowing it to excel in deep learning, running sophisticated models for user interaction with chatbots, and generative AI applications. By integrating the NVIDIA L4 GPU or the NVIDIA A16, our data terrarium can perform high-performance inference tasks while adhering to our commitment to sustainability.
Equipped with a lithium-phosphate battery as backup, the system can continue to operate even during times of low sunlight or at night. The solar panels used, increase efficiency from 500 to 1000 watts of energy meeting and exceeding the power requirements. For safety we plan to place the batteries underground equipped with a fuse to prevent short circuits. As opposed to traditional data centers which use significant amounts of water for cooling, the closed-loop cooling system we implemented uses an advanced heat exchange mechanism to dissipate heads without any water, ensuring efficient and sustainable cooling. This conserves water and maintains optimal operating temperatures for the hardware, enhancing reliability and longevity. It also allows our data terrarium to be an enclosed system, only requiring solar energy as external input, and data transmitted over WiFi, as the primary output.
To maximize the efficiency of our data terrariums, much like how you'd put a traditional terrarium in a sunnier spot in the winter or cooler spot in the summer for optimal solar energy absorption. We propose using a sophisticated scheduling program that leverages GIS data to find those places that will provide our terrariums with the most solar energy throughout the year, allowing them to operate with a net-zero external energy input maximizing sustainability. To further enhance this system, we envision creating a network of data terrariums for distributed computing, These data terrariums can also be used in conjunction with the eco data collectors, offloading computing and inference from the EDCs to our data terrarium, effectively making it a central point of communication for the end user. Internet access can be achieved through satellite-based internet such as StarLink. Our goal is to demonstrate the feasibility of this concept at a small scale initially. By starting small, we can carefully monitor and refine the system, ensuring it works seamlessly before scaling up. Once the system is perfected, we plan to apply the lessons learned to larger data centers. The insights gained from these initial deployments will allow us to develop advanced software capable of predicting solar capacity needs and improving sustainability for larger installations.
Data terrariums present several compelling advantages. Firstly, they are environmentally friendly, utilizing solar energy and reducing reliance on external resources. This approach minimizes the carbon footprint of data processing. Secondly, they are resilient; being self-sustaining means they can continue operating independently of the grid, providing a reliable solution even in remote or off-grid locations. Thirdly, they are cost-effective. With low operational costs and an affordable initial setup, they offer an economical solution for various data processing needs. The applications for data terrariums are vast. They can power deep learning models, enabling user interactions with chatbots and handling complex image and video processing tasks. By harnessing the power of the Nvidia L4 GPU, efficient solar panels, and intelligent power management, our data terrariums provide a reliable and eco-friendly platform for various data processing tasks.
This is our budget breakdown for the entire ecosystem that we are planning. The primary expenditure for the Data Terrarium project is the hardware. For our hardware costs, we are planning to build 5 data terrariums and 20 Eco Data Collectors which comes down to $35,000.
For our testing and implementation category, we will be conducting field testing, data collection, and analysis which will come down to $10,000.
We also have a miscellaneous costs category which includes transportation, shipping, documentation and 3D printing hardware. Our total project costs will be $50,000.
Tailored Services & Interested Parties
Davis: Services for Detecting Pollinators and Soil Species Diversity
In UC Davis, our services for detecting pollinators and soil species diversity are designed to provide comprehensive insights into the health and diversity of these critical ecological components.
For pollinator detection, we offer automated pollinator monitoring, utilizing advanced cameras paired with machine learning algorithms to identify and track pollinators in real-time. This technology allows us to gather continuous data on pollinator activity and behavior.
For soil species diversity, our services include soil microbiome analysis, where we monitor the soil. This analysis provides a detailed understanding of the soil's microbial diversity and health.
Riverside: Services for Flowering and Disease Monitoring in Citrus
In UC Riverside, our focus shifts to monitoring flowering and disease in citrus crops, crucial for optimizing yield and maintaining plant health.
Our flowering monitoring services include time-lapse photography, where we install cameras to capture the flowering process. This method allows us to determine the precise timing of flower opening and closing, providing valuable insights into the flowering cycle. We also employ machine learning for phenology, using advanced models to predict flowering periods based on environmental data. This predictive capability helps in planning and managing citrus production effectively.
For disease monitoring, we deploy disease detection sensors to identify early signs of disease in citrus plants. These sensors are crucial for early intervention and management.
East Coast: Services for Species Diversity and Population Movements
In the University of Georgia, our services are tailored to monitor species diversity and track population movements, essential for conservation and biodiversity management.
Our species diversity services include monitoring biodiversity in an area, where we document the species present in various ecosystems. This data is vital for understanding and preserving biodiversity. Providing a non-invasive and efficient way to monitor biodiversity. Additionally, we create detailed habitat maps, identifying key areas that support high species diversity, which is critical for conservation planning.
For population movement and monitoring, we employ video capturing to monitor the movements of individual animals. This tracking provides valuable data on animal behavior and habitat use. We set up camera traps to capture wildlife activity and movement patterns, offering insights into species interactions and population dynamics.
By offering these tailored services, we are dedicated to supporting the specific ecological data collection needs of each party. Through our efforts, we aim to contribute significantly to environmental research and conservation.
Here is the cost breakdown for the Eco Data Collector, notice that this is including the optional google coral usb for inference to make it more of an edge ai device. It does push the device closer to 300 but without it, it’s close to 200 dollars. These prices reflect the cost of an individual Eco Data Collector, once we ramp up production, we believe the price will fall over time.
This is the price breakdown for all the electronic components for the Data Terrarium, which include the GPU, CPU, board, Aluminum housing, solar panels, controllers, power supplies and batteries. This comes down to $5,443.45
In terms of the specs of the portable Eco Data collector, with the sustainable plastic casing, it is waterproof, allowing for deployments in damp and humid climates.
The 512GB storage allows for large amounts of data collection to be stored directly on the data collector.
The solar panels and rechargeable batteries allow for net-zero energy input. The video capture will be in 1-5 minute increments of HD video clips at 120 frames per second.
It supports both wide and normal angles and has infrared sensors that allow for night capture.
(PLAY during video) Here is a video clip of the normal angle camera taking 1080 by 720 pixels at 120 frames per second outside of the CSE building during the weekend.
The frame on the left was taken out of a 120 fps 1536 by 684 pixel video capture. The color is adjusted to capture infrared light for the ability to capture videos in the dark.
The frame on the right, was taken indoors also taken out of a 120 fps video capture. Using a ruler and the flower measuring 2 feet away from the camera, it shows high resolution and color. The cameras can be interchanged to provide higher fps and resolution.