It’s a breathtaking sight when meteor showers light up the night sky. However, the threat of larger celestial objects colliding with Earth poses a real danger. To counteract this potential catastrophe, a team led by physics professor Philip Lubin and his undergraduates at the University of California, Santa Barbara (UCSB) is working on a groundbreaking PI-Terminal Planetary Defense initiative. Their goal is to detect and mitigate space threats more efficiently, and they have recently received phase II funding from NASA for their research. NVIDIA has provided the team with an NVIDIA RTX A6000 graphics card through their Applied Research Accelerator Program to aid them in their mission. Let’s dive into the details of this innovative AI project that aims to safeguard our planet from cosmic hazards.
Also Read: Alien-Inspired Spacecraft Design: NASA’s Bold Leap Into Space’s Future
Pulverizing Space Threats
The core objective of the PI-Terminal Planetary Defense initiative is to detect relevant threats sooner and take decisive action to minimize their impact. In the face of an impending collision, the UCSB team plans to utilize an array of hypervelocity kinetic penetrators. These specialized devices are designed to pulverize and disassemble an asteroid or small comet, effectively neutralizing the threat before it reaches Earth’s surface. By breaking down these celestial bodies, the potential damage and risk to life on Earth can be greatly minimized.
Detecting Impending Catastrophe
Recognizing threats is the first crucial step in protecting Earth. Lubin and his students have harnessed the power of artificial intelligence (AI) to analyze vast amounts of astrophysical data. While modern surveys collect massive amounts of data, processing and analyzing this information at the required speed is challenging. To overcome this hurdle, the UCSB team is designing a large-scale survey tailored to planetary defense. This survey will generate even more data, which needs to be rapidly processed and analyzed.
Training an AI Sentinel
Lubin’s group has trained a neural network called “You Only Look Once Darknet” using machine learning techniques. This near real-time object detection system operates in less than 25 milliseconds per image. By utilizing a large dataset of labeled images, the neural network has been trained to identify low-level geometric features such as lines, edges, circles, and threats like asteroids and comets. Early results indicate that the AI-powered source extraction process is up to 10 times faster and nearly 3 times more accurate than traditional methods.
Also Read: AI Discovers New Planet Outside the Solar System, Scientists Failed to Find
Supercharging Processing Speed
To accelerate their image analysis process, the UCSB team has incorporated the NVIDIA RTX A6000 GPU and the CUDA parallel computing platform. The team initially faced challenges in reducing the processing time and meeting GPU memory requirements. However, with the RTX A6000’s 48GB of memory, they can handle complex graphics and large datasets without sacrificing performance. By implementing new CuPy-based algorithms, the team significantly reduced their subtraction and identification time, allowing the entire pipeline to run in just six seconds.
Tackling Technical Challenges
As the project grows and accumulates more training data, the team faces the challenge of handling increasingly large file sizes. The RTX A6000’s generous memory capacity enables the team to handle datasets of images with resolutions of approximately 100 megapixels. This powerful GPU eliminates the data transfer bottleneck, ensuring smooth processing and analysis.
Realistic Simulations for Precise Solutions
The UCSB team conducts simulations to demonstrate various aspects of their project. These simulations include modeling the ground effects of shock waves & optical light pulses emitted by fragments burning in Earth’s atmosphere. The team develops custom codes in multithreaded, multiprocessor C++ and Python for local simulations. For more intensive visualizations, such as the hypervelocity intercept of threat fragments, the team relies on the NASA Advanced Supercomputing (NAS) facility at the NASA Ames Research Center. Equipped with Intel Xeon CPUs and NVIDIA RTX A6000 GPUs, the NAS supercomputers provide over 13 petaflops of computing performance.
Also Read: NVIDIA Builds AI SuperComputer DGX GH200
Our Say
NASA’s decision to invest in space exploring AI technology seems to be in the right direction. The PI-Terminal Planetary Defense initiative led by Professor Philip Lubin and his team at UCSB represents an innovative approach to safeguarding Earth from space threats. The model combines cutting-edge AI technology, such as the NVIDIA RTX A6000 GPU, and innovative data processing and analysis methods. This makes it capable of detecting and mitigating cosmic hazards faster and more efficiently than ever before. With their ongoing research and development, the team brings us one step closer to a safer future where we can confidently admire meteor showers without fearing the unknown.
Sabreena Basheer is an architect-turned-writer who's passionate about documenting anything that interests her. She's currently exploring the world of AI and Data Science as a Content Manager at Analytics Vidhya.
