General Atomics Aeronautical Systems, Inc. (GA-ASI) has further advanced its Cooperative Combat Aircraft ( CCA) ecosystem. The company’s Avenger® Unmanned Aircraft System (UAS) is paired with a “digital twin” aircraft to autonomously execute real-time, virtual and constructive (LVC) multi-objective cooperative combat missions. The flights departed from GA-ASI’s Desert Horizons flight operations facility in El Mirage, California, on Dec. 14, 2022, demonstrating the company’s commitment to the use of artificial intelligence (artificial intelligence) and machine learning (ML). This provides a new and innovative tool for next-generation military platforms to make decisions under dynamic and uncertain real-world conditions.
The flight used GA-ASI’s novel reinforcement learning (RL) architecture, built using agile software development methodologies and industry standard tools such as Docker and Kubernetes, to develop and validate three deep learning RL algorithms in an operationally relevant environment. RL agents exhibit single, multiple, and hierarchical agent behavior. A single-agent RL model successfully navigates a real-time aircraft while dynamically avoiding threats to complete its mission. A multi-agent RL model pilots real-time and virtual Avengers to collaboratively pursue targets while avoiding threats. A hierarchical RL agent uses sensor information to choose a course of action based on its understanding of the state of the world. This demonstrates the AI pilot’s ability to successfully process and act on real-time information independently of human operators to make mission-critical decisions at relevant speeds.
For the mission, the flight path is updated in real-time based on fused sensor trajectories provided by the Virtual Advanced Simulation, Integration and Modeling Framework (AFSIM) model, and RL agent tasks are dynamically selected by the operator as the aircraft lifts off, demonstrating real-time, effective human Machines cooperate to achieve autonomy. This real-time operational data describing the AI pilot’s performance will be fed into GA-ASI’s rapid retraining process, analyzed and used to improve future agent performance.
“These flight demonstrated concepts set the standard for operationally relevant mission system functionality on the CCA platform,” said Michael Atwood, senior director of advanced programs at GA-ASI. “The combination of onboard high-performance computing, sensor fusion, human-machine collaboration, and AI pilots making decisions at relevant speeds demonstrates the rapidly maturing capabilities of GA-ASI as we achieve autonomy for CCA.”
The team used the government-provided Cooperative Operations in Denied Environment (CODE) autonomy engine and the government-standard OMS messaging protocol to enable communication between the RL agent and the LVC system. Leveraging government standards such as OMS will enable rapid integration of CCA’s autonomy.
In addition, GA-ASI uses EMC2 from General Dynamics Mission Systems to run the autonomous architecture. The EMC2, an open-architecture multifunction processor with a multi-level security infrastructure for hosting an autonomous architecture, demonstrated the ability to bring high-performance computing resources into the CCA to quickly execute a customizable set of tasks based on the operating environment.
This is another in a series of ongoing autonomous flights using internal R&D funding to demonstrate important AI/ML concepts for UASs.
About GA-ASI
General Atomics Aeronautical Systems, Inc. (GA-ASI), a subsidiary of General Atomics, is a leading designer and manufacturer of proven, reliable remotely piloted aircraft (RPA) systems, radar, electro-optical and related mission systems, including the Predator® RPA Series and Lynx® Multimode Radar. With more than 7 million flight hours, GA-ASI provides integrated sensor and data link systems for long-endurance, mission-capable aircraft to provide sustained flight for situational awareness and rapid strike. The company also produces various ground control stations and sensor control/image analysis software, provides pilot training and support services, and develops metamaterial antennas.
For more information, please visit www.ga-asi.com
