Intelligent Systems in EOD

Explosive Ordnance (EO) and Improvised Threats (collectively Explosive Hazards) caused 7,239 civilian casualties in 2017, and 15 of the ADF casualties on operations in Afghanistan. The threat of Explosive Hazards is high risk and continually evolving. We only have to look at the examples of Mosul and Marawi to see that, as the pace of war increases, this threat will continue to proliferate.

While my pitch focuses on Explosive Ordnance Disposal (EOD), my proposed solution has application across the entire Explosive Hazards spectrum, from route clearance all the way to post-conflict Explosive Remnants of War (ERW) clearance.

The problem is that our current EOD robots are dumb; they rely entirely on manual inputs from a human for operation, increasing cognitive and physical load, and they don't talk to anything else. As a result, our information relating to Explosive Hazards is siloed and fast becomes outdated. This prevents fusion and increases risk.

What if we could use smart systems and sensors to more accurately find EO and improvised threats and deal with them, all while providing decision support to technicians, tracking improvised threats as they arise globally, and feeding back into wider intelligence pictures?

Defence is investing heavily in autonomous systems across a number of projects. Currently, the Australian Army is conducting a trial of an autonomous enabled M113. Meanwhile, machine learning and AI (ML/AI) are perfect for parsing large amounts of information to find historical and in-service EO, as well as improvised threats as they evolve using open source intelligence.

Finally, processors and multispectral sensors are increasingly miniaturising, enabling them to be mounted on unmanned aircraft systems (UAS), and providing greater situational awareness while not flooding operators with too much information.

The solution

As we progress autonomous systems, we can scale their level of autonomy. Initially, we could automate simple functions, such as moving a robot forward to a safe radius where it can provide cognitive space for an EOD Technician to plan his or her course of action. Moving to full autonomy, systems could—without input—clear a marked area of ERW post-conflict.

Miniaturised multispectral sensors can build a picture—layering optic, thermal, ground-penetrating radar—and run chemical detection to sense and find the threat. Small form processors—such as Arduino, Raspberry Pi or Google’s Cloud Edge—will be able to use AI/ML to match the threat with databases, and provide techs with solutions for dealing with Explosive Hazards.

Finally, after executing an EOD procedure, the sensors and processors will be able to capture the results, modelling blast patterns in order to verify and validate the solutions offered, and improving on them in the future.

From a higher headquarters with greater processing capability, AI/ML will be able to seek out improvised threats on Open Source feeds as they evolve.