Fusion of Real-Time Data with Generative AI for Military
How do Vantiq Applications enhance Military Situational Awareness?
From experience gained through working closely with technical, academic, defense, and security colleagues, Vantiq has integrated enabling AI technologies (such as LLMs and GenAI) and has developed a roadmap to exploit these, to transform friendly force situational awareness within a contested battlespace.
The context for this discussion is the concerted effort in the world of Multi-Domain Integration (MDI)/C-JADC2 to harness a dramatically expanding set of real-time information sourced from military IoT deployed across multiple domains. This data explosion presents both new challenges and opportunities to defense consumers, and success will be realized via the deployment of a new layer of processing that integrates with and sits across existing and future systems. Of course, such processing is not an end in itself but, rather, it provides a foundation to better inform decision-makers in time-critical situations. There are at least two interesting opportunities: (a) to provide an accurate picture of the dynamics of the battlespace in real time; and, (b) to use this situational awareness to extract relevant information from a body of organizational knowledge to better inform the decision-maker of what they are seeing and what to do about it, communicating with them using natural language delivered via text or speech.
Vantiq’s approach is driven by practical experience in solving these critical pain points built upon work undertaken with partners in the safety, security, subsurface drones, and energy markets. Here are some of the main drivers behind our customers’ successful adoption of Vantiq:
- Multi-Domain Integration (MDI) Automation: At the heart of a Vantiq application sits the ability to significantly increase the speed of the OODA loop – capturing and processing events from multiple sources (defense IoT, AI, ML, etc.) and boosting their signal-to-noise ratio (using logic deployed on the edge to fuse, filter, correlate and further understand events) to isolate only ‘situations of interest’. Edge deployment is critical to dispose of the vast quantities of irrelevant data which renders conventional centralized architectures both too slow, vulnerable to disruption, and uneconomical. Effective processing of MDI events requires a native ‘edge-to-cloud’ distributed application model (which is what Vantiq delivers). This MDI component of a military Vantiq application delivers real-time ‘situations of interest’ to any interested subscribers across the network and provides the informational foundation through which to exploit a new generation of AI to further enhance decision-making as described below;
- Augmented Situational Awareness: Vantiq applications are designed to optimize the capabilities of the human in, on, or out of the loop across the whole defense and national security IoT. The main idea is to leverage the ‘situations of interest’ that emerge from processing ISR in real time to greatly enhance the situational awareness of a decision-maker when seconds count. The key enabling technology here is Vantiq’s automated integration with information stores (such as Vector/Graph databases) and GenAI to enable flexible communication. This means that a Vantiq application can leverage the intelligence of discovered hostile assets and, for each target, automate prompt engineering to extract, in real-time, relevant information from the collective body of organizational knowledge. This is used to drive better decision-making (e.g. tell me about each asset’s capabilities, tell me from Lessons Learned what happened the last time we encountered this asset and how we countered it, etc.). Vantiq’s integration with GenAI enables bi-directional communication with a user in text or speech format. A Vantiq application can further automate actions based on such extracts (e.g. give me a safe route to navigate through the known range of all hostile assets);
- Adaptable to Degraded Operations: A Vantiq application assumes that friendly force networks will be subject to EW, may be denied, and that elements of the electromagnetic spectrum will be incomplete, or at best intermittent. With that in mind Vantiq: (a) allows autonomous operations of all application nodes in a distributed model; (b) nodes may communicate using a fully meshed networking model which has no single points of failure. Users can dynamically detach and reattach on a publish and subscribe basis and ‘events of interest’ can be reliably delivered to all permitted subscribers when the network is reestablished, and the right opportunities emerge.
- Open, Event-Driven Integration: Vantiq is technology-agnostic and is easily dropped into existing and future systems through an open architecture. Vantiq’s edge approach can integrate numerous systems of systems, including across various allies and agencies with differing content and security requirements. There are numerous defense and security use cases, particularly in the C-JADC2 and NATO / Sovereign MDI domains. Furthermore, a Vantiq application can implement a ‘data mesh’ architecture, that shares MDI metadata across a distributed application so that data is moved only on demand;
- Agile Delivery: Vantiq is designed for use by professional developers, but its low-code approach means the system can be programmed and adapted quickly. Our support of agile methods allows for rapid customer and technological upgrades – there are no sealed black boxes.
Intelligent C2 Operations
The Joint Operations Staff had received a request for the rapid deployment of specialist staff and troops to a location 300 miles away. The initial request had been made verbally by troops in contact with the enemy. Their verbal contact report into their edge devices, had been automatically recognized by the monitoring Vantiq application, which then initiated three automated prompts into GenAI to interrogate and extract from a vector database (a) what hostile assets do we know about in this area; (b) what friendly assets are immediately available; and, (c) what is my immediate action? Specific words and phrases made to GenAI, coupled with feedback from the troops’ wearables, had triggered an automated response from the Vantiq application. This might include re-tasking local ISR assets, bringing medivac helicopters to readiness, and warning combat air assets of a potential task. Having saved vital minutes, it was now up to air operations staff – either in, on, or out of the loop – to authorize further action.
The Joint Operations Staff had minimum time to make a plan and deploy vital assets in a hostile environment. The head of CJ3 – Major Smith, sat in the JOC secure communications facility and outlined the situation to her specialist staff using automated prompts to GenAI to summarise and share the current situation. As with all ops personnel, her use of keywords and phrases (that accord with extant military nomenclature), ensures a close mission partnership with the Air Ops GenAI and its accompanying databases. Importantly, the Air Ops GenAI will be able to offer planning options based on the entirety of known doctrine, more focussed tactics, techniques and procedures, local history and intelligence, lessons identified, and real-time reports reflecting what is happening at the edge – across the entire battlespace – only seconds beforehand. Also, the Air Ops GenAI’s ability to give insight into likely future events and outcomes, based on Maj Smith’s decisions, was invaluable. The interactions between situational information and the Air Ops GenAI were entirely automated by a Vantiq application.
In real-time, the intelligence picture covering hostile force threats, on route to and around the target area, was constantly updated by a host of ISR assets, integrated by the Vantiq application which updates the intelligence Vector database. Reports varied from highly nuanced long-term intelligence to the latest situation, and real-time reporting from sensors whose data was captured at the edge and forwarded to Major Smith only seconds later. Similarly, the latest status of friendly forces was constantly updated in real-time. Availability, serviceability, fuel, and weapon load, EW fit, the status of sensors, training levels of personnel, nationality, rules of engagement, command status, etc. are all captured, managed, and shared with relevant subscribers by the Vantiq application. Through the use of Vantiq’s open standards and open architecture, the interoperability and integration with mission partners existing and future systems are ensured to be a success without the typical lengthy lifecycles and subsequent system updates.
Major Smith and her staff verbally inputted essential mission information to GenAI and employed input devices to enter data such as range to the target, characteristics of the equipment needed, etc. More nuanced information, such as the effects required, was conveyed to GenAI by the mission verbs spoken by Maj Smith. These mission verbs were identified by the Graph Data Base and used to shape the nature of the planning options provided back by the Air Ops GenAI. This allowed, for example, the estimation of optimal routing and force packages needed for success.
On completion of Major Smith’s 10-minute planning session, GenAI offered 5 options. All were designed to optimize available friendly forces’ efficacy in the hostile environment. Each option offered different levels of risk and time on target and included contingency options for friendly forces to mitigate the effects of hostile action.
After further iterations of questioning via GenAI, Major Smith selected the optimal Course of Action and briefed her team. Through high levels of automation, orchestration, and integration, provided via Vantiq, the units likely to be involved were immediately given deployment orders.
The force package was deployed within a significantly shortened timescale. This freed Maj Smith and her staff to execute the next mission and deliver the capabilities of the troops deployed 500 km away.
Throughout the mission, air, and ground personnel queried and verbally updated the Air Ops GenAI on threats and potential mitigations to reduce force exposure to hostile action and optimize chances of success.
During the mission, the Air Ops GenAI assumed a hostile EW environment with limited and intermittent access to the electromagnetic spectrum. Vantiq allowed key information to be cached and ‘squirted’ or received when possible and the processing of events of interest to front-line operators at the edge.
Saving Lives and Regaining the Initiative
A team combat medic is injured by an IED while on patrol – several other members of the patrol are fragged or suffer shock. Other, non-medically trained, patrol members can input the casualty’s data into an ‘automated combat partner’ edge / AI system by voice, or input device (see Optac-X for an example of a Vantiq partner that does precisely this). The patrol members’ edge wearables will also detect and react to the event and communicate the medical situation. The automated combat partner will identify a wormhole through the enemy’s EW measures and cache, or transmit whatever data is required.
In the seconds following the event – when the patrol is in shock and vulnerable, the automated combat partner will be able to provide life-saving medical advice but also suggest casualty evacuation procedures, identify the best local landing sites for helicopter casualty extraction, suggest procedures (options) the patrol commander should immediately consider to counter any further IEDs / risks of ambush, write a draft contact / after action report and transmit, or wait for approval.
Depending on the levels of automation authorized, the automated combat partner might scramble a casevac helicopter and its escort, call tactical air support into range – should it be required – provide a brief, and deploy the unit’s quick reaction forces to a holding position, focus ISR – satellites / UAV / aircraft – and edge compute assets onto the area to look for enemy targets / suggest helpful effects (eg EW to jam other IEDs / track enemy comms / counter hostile AI systems), pull together previous patrol reports and intelligence and present options and insights to inform the local senior commander’s next move. Graph databases would be particularly powerful in understanding the potential enemy target sets.
Based on the senior commander’s AI-advised decisions/options, offer advice on likely futures – ie what effect does the commander want to create next if option A, B, or C is followed? A Lesson Identified / After Action report will be automatically produced in seconds – for approval, constantly updated, and integrated with higher command planning, future training, and equipment design. Logistically, the consumption of fuel and wear on the rotor blades of the casevac and escort helicopters will be built into the maintenance and re-supply system – improving serviceability and saving money.
All the above would be initiated in real-time, fully integrated across the IoT involved, and orchestrated throughout the multiple layers of command – from Private to President/Prime Minister and NATO allies (depending on pub/sub and classification rules)
In a significant crisis, events like this – whether kinetic, logistic, C5ISR, etc – would be occurring many times per day. The integration of numerous systems, real-time decision-making at the edge, orchestration of numerous AI systems, and harnessing of LLMs / Vector and Graph databases would greatly increase the efficacy of the friendly force, significantly shorten the OODA loops, and amplify C-JADC2/MDI advantage. This advantage would be critical to overmatch hostile AI systems. That aside, the savings in terms of lives, time and money would be significant.