In the development of video games, every scenario is expected. Every situation is accounted for and the system learns each response. Player 2 reaches Level 3: the system knows to open the Red Cavern of Secrets. This is similar to how self-driving cars learn today. These systems can be taught what to expect using decades’ worth of maps, GPS coordinates, street signs, and driving manuals.
In a warfighter scenario, this process of learning is inadequate as near-peer adversaries are quickly and often changing the rules of the road. But, there is much to be said for how autonomy and video games can be used to advance warfighting technology. From training that leverages virtual, augmented, and mixed reality simulations, to machine learning systems being tasked with playing and winning complex computer games, human and autonomous system teaming is entering a new era.
While “autonomy” is a system attribute that may use different techniques including: artificial intelligence, machine learning, advance learning, deep learning, reinforcement learning, and game theory, BAE Systems’ autonomy is rooted in control theory and deterministic algorithms that are proven and trusted by our Warfighters. No matter what techniques are used, BAE Systems’ overall goal remains the same — develop trusted technology that removes some cognitive tasks from warfighters and provides them with actionable information to be more effective and safe.
Our FAST LabsTM research and development organization has decades-long leadership in autonomy. In fact, we were the original developers of the autonomy software used in the Defense Advanced Research Projects Agency’s (DARPA) Joint Unmanned Combat Air System (J-UCAS) program dating back to 2002, and we have a connection to the early days of the video game industry.
Some of our most recent programs and technologies that have helped to influence current and future systems include: All-Source Track and Identity Fuser (ATIF), Multi-INT Analytics for Pattern Learning & Exploitation (MAPLE), — which has recently been leveraged for DARPA’s Geospatial Cloud Analytics (GCA) program, Distributed Battle Management (DBM), Multi-domain Adaptive Request Service (MARS), Causal MOdeling for kNowledge Transfer, Exploration, and Temporal Simulation (CONTEXTS), MindfuL™ technology and work on DARPA’s Squad X program.
“While this merging of autonomy and simulation is a new trend, the reality is that we’ve been working toward this goal for years,” said Dan Zwillinger, chief scientist at BAE Systems’ FAST Labs. “The work we’ve been doing for the last 20 years serves as the foundation for the programs of today and tomorrow.”
Leveraging these and other recent successes, we are focused on pushing the boundaries of machine learning to bring the power of simulation out of the video game and into the complex scenarios faced by our military planners and warfighters.
The most recent advances in autonomy technology are leading to trusted systems designed to reason about the present in addition to learn from the past. In other words, the advanced autonomous systems are now able to react to scenarios for which they were never explicitly trained to expect. As we pioneer these possibilities, we will continue to leverage our decades of autonomy experience in the ground, sea, air, and space domains – meaning, “Game Over” for our adversaries.