DARLA-02

Demonstration of artificial reasoning, learning, and analysis

2022-present

About

The central focus of the DARLA-02 mission is to create a network of distributed sensor platforms that detect and react to events in their surrounding environments. This network will consist of one physical CubeSat developed by the SSRL team and eight additional virtual satellites. Each entity within the network will autonomously run ARES, machine learning software that can detect, learn from, and react to events and abnormalities in its environment.

The CubeSat is 3U, measuring approximately 10x10x30 cm, and is designed to operate in Low Earth orbit. The ARES AI software running on-board the satellites was initially flight tested on the Argus-02 CubeSat and is developed by Bennett Research Technologies.

Mission

The scientific data generated by the DARLA-02 mission is different from SSRL missions of the past. Rather than running a quantifiable experiment contained on a CubeSat, the DARLA-02 mission’s goal is to validate the functionality of a network based around event detection and machine learning. Full mission success requires multiple satellites working in conjunction with multiple ground-based platforms. The satellites and ground platforms are made so the ARES software can sense events. In the case of the DARLA-02 CubeSat, this data event data would come in the form of Radio Frequency (RF) readings from a radio or pictures from a camera. This data can then be sent out to any entity within the network (other satellites or ground platforms). Each instance of the ARES software can learn from event data using machine learning techniques and over time begin to predict or react to such events. The ARES software itself runs autonomously on each platform and determines what data constitutes an event. This means that the limited capabilities of the communications link within the network are used only to share unusual or “interesting” data, freeing up bandwidth for other mission-critical data.

Ultimately, the aim of the DARLA-02 mission is to help answer a fundamental design question: Can a network of low-cost, lower-performance, disposable spacecraft match the capabilities of one or a few high-cost, high-performance, exquisite spacecraft for their ability to detect and respond to events? A similar question hovers around other CubeSat and small satellite missions: whether meaningful scientific data can be generated while reducing costs to time, budget, resources, and personnel.

Engineering Process 

The DARLA-02 mission was initially developed as a part of the University Nanosatellite Program, UNP NS-11 Phase A, along with missions from other schools like Auburn University, Purdue University, and the University of Maryland. After concluding Phase A, the mission was accepted into NASA’s CubeSat Launch Initiative.

Payload

The payload on each of the DARLA-02 CubeSats is fairly simple. Each requires a computer that can run the ARES software, in this case, a Raspberry Pi, and sensors to collect event data. The payload sensors flown on the satellites include four Software Defined Radios (SDR) and two cameras. These components are required to gather meaningful data to share with the network.

Flight Hardware

The DARLA-02 team has selected NanoAvionics and EnduroSat to supply the majority of our satellite subsystems. The remaining components are being developed in-house by our undergraduate student team or procured from a third party.

DARLA-02 Spacecraft Leadership Team