Pacific Northwest National Laboratory (PNNL) is dedicated to pushing the boundaries of computing, striving to deliver innovative capabilities. Adaptive autonomous systems are at the forefront, leveraging diverse computing environments to enhance scientific and security advancements.

What are adaptive autonomous systems

Adaptive autonomous systems are technologies that use AI to perceive their environments, learn from what they experience, and adjust their behavior accordingly. For example, autonomous vehicles rely on adaptive autonomous systems so they can respond to changes in their surroundings like road conditions, adverse weather, and other traffic.

Pacific Northwest National Laboratory (PNNL) uses adaptive autonomous systems to push the boundaries of advanced computing, deliver innovative capabilities, and advance scientific and national security, specifically through Cloud Computing, High-Performance Computing, and Edge Computing.

Each of these computing environments lends itself to particular situations and applications. PNNL’s vision is to create tools and automation capabilities that allow adaptive autonomous systems to use AI to automatically shift between these three environments based on situational needs. This dynamic ability can help create virtual and distributed adaptive autonomous systems.

Such systems retain the benefits of each individual computing environment while also allowing information and data to automatically move between them, taking advantage of additional computing or other services.

How PNNL is advancing AI for more resilient national security operations

Drones—also known as unmanned aerial vehicles, or UAVs—have surged in popularity among the public and are increasingly used in national security operations. However, UAVs often encounter challenges related to size, weight, and power constraints because of their small size or limited computing capabilities.

PNNL is developing and testing strategies to integrate advanced AI models into UAVs and infrastructure that supports their operation, like ground control stations and sensors—referred to as unmanned aircraft systems (UASs)—enhancing their effectiveness, performance, and efficiency.

Research and technology development in adaptive autonomous systems improves data transfer between the cloud and edge computing, such as edge devices on UAS/UAVs like lidar scanners, thermal cameras, and navigation equipment. These AI-driven systems adapt to swiftly analyze visual and electromagnetic data to make rapid decisions.

This cutting-edge technology not only optimizes UAS/UAV operations but also lays a secure foundation for cloud-to-edge communications in critical national security missions.

Autonomous research in the open ocean

At PNNL-Sequim, our team is harnessing the potential of autonomous surface vehicles—also known as “drone boats”—and UAS/UAVs, fostering seamless communication between them. The ultimate objective is autonomous research where surface, air, and subsurface vehicles collaborate, collecting and relaying important data gathered in the environment such as signals, video footage, water characteristics, and humidity levels.

Understanding these capabilities and constraints is crucial for many national security missions, including protecting critical infrastructure and completing risky missions without endangering human lives.

This project is also helping tackle the unresolved challenges of coupling adaptive autonomy with cloud computing. Pairing adaptive autonomous systems with the cloud allows vessels like drone boats to continue operating safely and efficiently when navigating remote ocean waters. In such a vessel, instant safety decisions like avoiding a sudden obstacle on the water’s surface can be handled with local edge computing, while cloud computing allows the vessel to continue navigating and staying clear of larger adverse weather patterns, such as brewing storms.

Automation, scientific computing, and new materials

New materials have the potential to improve solar generation, advance battery technology, develop new health care treatments, and enable improved techniques in computing. Both theorists and experimentalists are tackling the problem, but each from different angles.

While theory benefits from experimental measurement and validation, experiments benefit from theory’s guidance and explanations.

PNNL is developing computing and coordination techniques that can be used to automate cycles between theory and experimental work, accelerating scientific discovery. New methods are needed to represent and execute the computational workflows formed between instruments and computational devices, ranging from near-instrument devices, high-performance computing systems, and specialized demand-driven cloud resources. Our research in this area is exploring a wide range of techniques for productive and portable representations, scheduling, and resource assignment. Diagnosing and removing performance bottlenecks, advancing AI/machine learning methods, and customizing data flow is also being explored.