Australian firm Dexata has a salve for ISR operators struggling to spot actionable intel in mountains of data: they call it “intuition in a box”.
A conversation with Dexata Corporation sales engineer Boris Novak can rove across myriad horizons, from database interrogation to air defence strategy, human factors and the design of video game consoles. But it all circles back to solving what he has identified as a growing problem in our “fifth-generation” age of exponential growth in networked sensor platforms: how to sift the resulting avalanche of data to find the nuggets and patterns that form the bedrock of “intuition” so critical to decision making.
Novak’s team believes the sheer volume of data available on the battlefield today can potentially overwhelm human ability to track and identify the patterns of movement that have traditionally signalled a military adversary’s capabilities and intentions.
And operational data collection and analysis is something the company understands. Dexata is known for its ability to provide ISR-related training systems and analysis tools. The ADF used the company’s Mobile Rocket Impact Scoring System during commissioning of its Tiger Armed Reconnaissance Helicopters. Firing the Tiger’s unguided rockets on South Australia’s Woomera Test Range Dexata’s mobile system recorded impact locations to within one metre, using cameras mounted up to 5.5 kilometres away. Its accuracy and reliability saw the test program completed in five weeks instead of the scheduled 12.
Horus
Dexata has built on its experience to develop something new; its “Horus” data engine, named for the mythological Egyptian deity known for its “all seeing eye”, combines big data and machine learning to help operators of battle management and ISR systems spot critical patterns and events in the “data-rich” environment of military conflict.
Horus identifies and records patterns of behaviour from both historical and real time data feeds so the operator doesn’t have to. And it keeps on doing so, storing an ever growing database to inform surveillance, analysis, planning, coordination and decision support at any level from foot soldier through armoured vehicle, warship and Airborne Warning and Control aircraft, to senior command.
“Usually when you have a combat system or a BMS or a surveillance picture, you have the server or the mission system itself that’s collating all the information and a user interface to get information out of the system,” Novak said. “Horus would attach to the mission system: it ingests the data from the mission system, but then it also would allow the operator’s user interface to bi-directionally interrogate Horus, with Horus sending messages to it as well. The Horus knowledge node could also be made available for consumption by other users and systems, not necessarily dependent upon the source mission system itself.”
“Horus helps answer the questions: what in this picture needs my attention, why does it need my attention and do I need to do anything about it? It works by improving the ability of the most critical component of all of these systems, the people charged with monitoring, interpreting, managing and acting upon the information they present.
“Unfortunately, intuition is an inconstant, unquantifiable attribute and trained, experienced people are an expensive commodity. What our system provides is basically established intuition in a box.”
Pattern analysis is nothing new but Dexata has added a twist. Rather than sticking with a traditional pre-loaded “ruleset” against which observed tracks and patterns are compared to be judged as normal or unusual the Horus system also learns as it goes, identifying and storing patterns, then noting diversions and anomalies. This effectively shows patterns and change against fresh, timely data, not pre-loaded legacy data that may no longer reflect reality.
Previous systems have often been based on a pre-programmed “pattern of life” ruleset, where behaviour such as an object crossing a geographic boundary at a certain time or in a certain way is flagged if it diverts from what is considered “normal”.
“The problem with this approach is that codified patterns of life are based upon people’s prior assumptions of what a pattern of life ought to be,” Novak said. “Often, assumptions are not an accurate reflection of reality. I would argue that previous paradigms of patterns of life are really models of life, compared to our system, which is statistically-derived by the data, on the go.”
Machine learning
Machine learning also means new observations automatically become part of the database, where older systems needed manual intervention to update the underlying ruleset with new, real world data. The result is better data with a higher level of confidence, in real time. And today’s networking means the knowledge can be broadcast. A ship, vehicle or aircraft entering an operational area for the first time will already hold real time situational awareness of current events and patterns in that space. It will know what’s happening right now.
“With Horus the quality of operational decisions is different, in the sense that a warfighter should have a higher confidence in terms of what they’re targeting. And they should establish that confidence with far greater speed and accuracy with our system and with fewer degrees of separation in how they obtain that knowledge. On the intelligence side, it can also help them find insights far more quickly than they otherwise would be able to.”
Dexata is still at the start of the curve. The company has faced challenges in integrating Horus with myriad systems, and in building a system that will consistently digest and present many different forms of data. Unlocking that ability is the key to the system, and its success in doing so is the point of difference Dexata believes has put the team at the head of global development in the field.
“We have a visually rich environment, where we’re trying to add context and provide users access to information that exists behind a system that a human just wouldn’t be able to access normally,” Novak said.
“The issue with that kind of data is it’s not stored in what I’d call a data-mining-friendly format. It’s not like a spreadsheet or a relational database where you can just slice and dice and find interesting insights simply by adding things in certain ways. It’s difficult to manipulate and we’ve managed to find a technique to do that.”
The ADF has conducted an initial trial of Horus and interest from global defence forces is growing, but the system is yet to be deployed operationally. Novak believes it has taken some time for defence forces to understand the realities of a Big Data approach to analysis. But it’s coming.
“We are having conversations around the globe,” he said. “Markets are catching up and saying we need to find tools like this. In order to be able to ingest and digest the data to comprehensible levels, it needs something like this.”
This article first appeared in the November edition of ADM.