Blogs

Consumer Electronics Show (CES)

January saw Las Vegas again host the glitzy CES. The more bizarre products this year were a voice-controlled luxury toilet seat, an iPhone case incorporating a physical keyboard (eight years after the demise of the Blackberry), and ‘Flappie’ – an AI-enabled cat flap that locks shut if it detects your pet is carrying a bird, mouse, or small animal in its mouth. Other proof-of-concept products included a wireless mouse and keyboard that harvests electrical energy from mechanical movement; the Mouthpad – a denture plate that uses your tongue to control a mouse or other cursor; and, Skwheel – motorised, 2-wheel skates that can transport the wearer at up to 50mph.

CES tends to demo post-laboratory experiments – things that have progressed far enough from the R&D bench to have a user interface and exterior that is slick and familiar enough to people. However, such proofs-of-concept are usually far from the consumer market, with products sometimes taking years to make it to shelves (if successful). Nevertheless, CES presents a window into near-term technology developments, some of which may have defence applications (such as Skwheel transporting soldiers).

Developing training data for household robots

A team at New York University has developed an approach to recording training data for household tasks. An iPhone is used to record activity, capturing data on movement, depth and rotation. With only 13 hours of recorded data, an AI model called Dobb-E was able to be trained in 20 minutes to perform tasks such as pouring from a cup and opening blinds, executed through a commercial robot called Stretch. This approach is similar to a Google Deepmind project – RT Trajectory. Stanford University’s Mobile ALOHA project developed training data by human telemanipulation of robot arms to cook a meal which included scrambling eggs and pan-frying chicken. Repeated execution of the tasks, coupled with a deep learning AI model, allowed the robot to carry out tasks autonomously.

Getting robots to carry out tasks humans consider simple is a holy grail of robotics (see the DARPA robotics challenge fails). These approaches to developing training data may get researchers a step closer to achieving this goal.

Reducing lithium and cobalt in batteries

Microsoft and the US Department of Energy analysed 32 million inorganic materials using an AI model and a supercomputer to identify potential solid electrolytes for batteries. This quest produced 18 potential candidates in less than a week, with one (dubbed N2116) synthesised. It has been developed into a working, solid-state battery prototype and has the potential to reduce lithium use by 70%. In other battery news, Lamborghini has licenced an organic battery technology developed by MIT called TAQ (bis-tetraaminobenzoquinone). TAQ is a potential replacement for cobalt or lithium-iron-phosphate based cathodes. It has a similar structure to graphite (which is almost universally used as a battery anode), allowing it to store lithium ions. The paper claims TAQ offers greater energy density, faster charging times, and eliminates the need for cobalt (often associated with a high social cost of extraction, such as child labour).

Solid-state batteries are so called as they use a solid electrolyte; the transfer of ions between the anode and cathode is traditionally through a liquid electrolyte. Solid-state batteries are hoped to perform better and more safely than traditional batteries, but implementing a solid electrolyte is challenging. Ideas such as N2116 and other announcements (including by Toyota) in recent months suggest solid-state batteries are edging closer to commercial reality. Battery chemistries that use common precursor materials (unlike cobalt and lithium for example) may create cheaper, and more socially and environmentally responsible batteries for mass manufacture.

Rotating detonation engines

Since the 1960s, rotating detonation engines (RDEs) have been experimented with. Conventional jet engines combust fuel and air – a sub-sonic process; whereas RDEs use a controlled detonation of fuel and air to create a rotating, super-sonic shockwave that produces thrust as it is ejected at the rear of the engine. This reaction is self-sustaining and fuel-efficient, but is not without drawbacks – including the need for assistance at take-off and the noise produced. Using a novel 3D-printed alloy, NASA continuously ran an experimental RDE for over 4 minutes, and General Electric claim to have developed a dual-mode ramjet with rotating detonation in a supersonic flow.

While RDEs are not a new concept, new materials and manufacturing techniques may make them a realistic option for next-generation hypersonic military craft.

AI-enabled microscope

A US Defense Innovation Unit collaboration with Google, Veterans Affairs, Jenoptik and others, has created an augmented reality microscope. Aimed at benefiting pathologists providing cancer diagnosis, the microscope augments the visual field of view with inferences from AI algorithms trained on various cancers from lymph node biopsies. This microscope is a working prototype from research that began in 2019.

While of immediate benefit to US veterans, this technology could quickly make its way to the broader healthcare sector. The use of AI to detect cancerous cells has been developed for several years, though this represents a first for augmenting the visual field of view of a microscope.

Manoeuvring a plane with air jets

Aurora Flight Sciences (a Boeing company) has revealed the design for the X-65 aircraft, sponsored by DARPA. The unmanned X-65 does not have traditional, moving control surfaces such as a rudder and ailerons. Instead, pressurised air is piped to multiple nozzles placed along wing edges and flying surfaces which controls the direction of the aircraft. Real-world testing of the X-65 (which is about the size of a BAE Hawk) is planned for summer 2025, including performance comparison against an identical aircraft with traditional control surfaces. In other X-plane related news, NASA has unveiled the X-59, designed to travel at supersonic speeds without the associated sonic boom.

Reducing the mechanical complexity of an aircraft brings obvious advantages. However, the efficacy of this novel approach is yet to be seen. Should it prove successful, this may have significant benefits to both military and civilian aviation.

Drone detection technology

The conflict between Ukraine and Russia is a proving ground for how cheap, mass-produced drones are influencing the conduct of war. More recently, Ukrainian start-ups have been developing drone detection technologies in response to the prohibitive cost of foreign made systems. Products such as Eter and Obriy use off the shelf components combined with proprietary software to assist Ukrainian troops detect and avoid enemy drones.

Offensive capabilities typically beget defensive capabilities, and this example is no exception. While innovative, these approaches may lack the rigorous testing and hardening typical of more traditional Western military technology procurement. However, being launched straight into battlefields, with a rapid iteration and exploitation cycle, this will quickly catch-up.

Rydberg antenna in field trials

The US Army Network Modernisation Experiment has recently demonstrated a Rydberg antenna in field trials. Rydberg atoms have a high energy state with one or more electrons far from the nucleus. This distance increases the sensitivity of Rydberg atoms to electric fields, meaning they can work as sensitive receivers for radio transmissions in very wide bands – theoretically from zero to THz ranges. This quantum receiver was demonstrated in the US Army Research Laboratory in 2018.

The practical, field demonstration of Rydberg effects to receive radio transmissions is a significant step forward in electromagnetic sensing. Detecting low-powered transmissions over wide frequency bands could provide much greater visibility to enemy equipment and operations.

Follow Futures Lab on LinkedIn to keep up to date with our latest news.