How Discovery Machine, Inc. and Improbable partnered with the British Army to deliver responsive, adaptable and interchangeable AI experimentation
The Army Warfighting Experiment (AWE) is an annual event that sees industry, academia and allies participate to help ensure the UK’s resilience in the face of an increasingly complex, fluid and fast moving threat landscape. It’s an opportunity for the Army to explore how best to prepare personnel at every level for future operations.
Working in partnership with both the British Army and our partners Discovery Machine, Inc., Bohemia Interactive Simulations, SimCentric, Microsoft, Epic Games and Improbable’s games technology division, a small team from Improbable Defence addressed a series of predefined challenges. In this three-part series, we take a look at each challenge in turn.
Daniel Harris, Principal Systems Engineer
Even the most minute detail can affect A3ES response in a real-world battlespace. It might be the way you hold your rifle, whether you wear a beret or a helmet, or if you decided to don sunglasses that day. Combined, these elicit extremely nuanced responses from Audiences, Actors, Adversaries, Enemies and Specialists that ripple across the operating environment.
Creating A3ES that reflect these real-world responses is therefore crucial in realising effective immersive training environments. Better training can be conducted if realistic behaviours can be simulated at scale through the power of AI. This goal underpinned experiment #3 of AWE21, which tasked participants with achieving accurate A3ES representation through responsive, adaptable and interchangeable AI.
Alongside partners DiscoveryMachine®, Inc. (DMI), and with support from Epic Games’ Unreal Engine (UE) 4 and UE Marketplace, Improbable set out exploring how this could be solved to improve A3ES training for the British Army.
Through rapid and effective model integrations with Improbable’s synthetic environment development platform, DMI deployed two AI population models that provided adaptable behaviours in response to trainee actions, and were interchangeable to allow trainees to experience two distinct sets of civilian behaviours according to the region on which the model was based. What resulted was nuanced A3ES modelling, evidenced by a realistic and culturally relevant civilian population.
“It was a delight to see our highly complex and specialist AI come to life in a simulated environment at such huge scale, as demonstrated in AWE21 experiment #3. Working closely with The Multiplayer Group, who developed the connection between our application and the IO UnrealGDK, we provided the application and technical advice needed to enable the integration of rich, detailed models into immersive training environments. What resulted was focussed A3ES interactions with realistic, high-fidelity and geographically relevant representation, empowering the British Army to train in scalable, nuanced cultural engagements.”
– DiscoveryMachine®, Inc.
A subsidiary of Improbable, The Multiplayer Group helps developers and game studios deliver high-quality multiplayer experiences. Video games and military simulation may seem poles apart, but both address a similar problem to different ends: creating complex, rich, immersive environments. But where video games entertain, for the British Army the stakes are considerably higher.
By deploying industry-leading games technology, AI and models, experiment #3 demonstrated how military simulation innovation can be accelerated and the resulting solutions put into the hands of the British Army quickly, efficiently and economically to provide a crucial competitive edge.
Experiment #3 allowed users to choose to deploy in the simulation as a male or female soldier, walk through a busy marketplace and discover how gender plays a role in interactions between soldiers and civilians. Other gestures, such as drawing a weapon, would also yield different civilian reactions depending on gender and population set.
Training supervisors could choose the most relevant models for different training scenarios, giving the user greater control over how simulated A3ES responded to friendly action.
Users also had the ability to switch between entities seeded from a Central Asian population and entities seeded from South Pacific population, before gauging the different reaction from each in response to the same user action.
A high-fidelity virtual world powered by Unreal Engine 4 supported this simulated battlespace. It featured assets sourced from UE Marketplace and terrain reused from AWE21 Experiment #1, reducing development time and cost.
“Content is key for immersive virtual worlds. Unreal Marketplace proved an essential resource in providing participants of AWE21 experiment #3 with content, and continues to help tailor realistic engagement environments for true-to-life training.”
– Epic Games
Experiment #3’s exploratory work yielded actions and results beyond simply responding to the task at hand. It showed how a robust partner ecosystem can work at pace, flexibly and with optionality, to deliver integrated solutions via Improbable’s platform.
We’re living in an era of constant competition between states. Hybrid warfare and grey-zone operations are par for the course in what is an ever-changing threat landscape. Major decisions, involving wide-ranging, long-term consequences, can fall to personnel at any level. That’s why it’s more important than ever that the latest technology gets into the hands of the organisations we serve so that they can train together more often, more effectively and more economically.
As things stand, this need has yet to be fulfilled. Many of my ex military colleagues have described their collective training experience as rarely a fulfilling one. Trainees often have to rely on the strength of imagination to ‘pretend’ that a training environment is something other than what it clearly is, which means guessing at the cascading consequences of their actions.
As the character of warfare evolves, this approach will no longer fly. Personnel need ways of exploring, understanding and ultimately overcoming fluid and fast-moving threats as quickly as they appear. This is why Improbable Defence is aiming, above all else, to enable a new generation of multi-domain, multi-role synthetic capabilities like those we explored alongside our partners as part of AWE21.
Through a platform approach, involving crucial partner integrations that unite the latest, most reliable and most relevant content and technologies, organisations and personnel can harness the power of military simulations and synthetic environments to navigate today’s threat landscape. These richly detailed, highly realistic virtual worlds give soldiers the chance to freely make mistakes, so they’re safer when they head out on operations in the real one – whatever their headwear.
For an in-depth demonstration of the technology behind AWE experiment #3 and to discuss how Improbable Defence’s partner network can open up new opportunities for industry, get in touch with defence.enquiries[at]improbable.io.