For over a century, driverless vehicles have been seen as the next great frontier in the automotive industry – a frontier that remains tantalisingly out of reach for mass-market vehicle manufacturers.
Standardisation body SAE International has six clear tiers of automation, ranging from Level 0 – no automation – to Level 5, which is described as ‘steering wheel optional’. In other words, a Level 5 vehicle can operate without any human intervention whatsoever in any environment.
This opens up many exciting possibilities for the future. Courier companies can make deliveries in driverless vans that are always on time. Cars can sync with traffic control systems and GPS data to plot the optimal route to their destination, making commutes through rush hour traffic jams a thing of the past. Vehicles of all types can detect black ice long before a human driver could, adjusting steering, suspension, and tyre inflation accordingly to keep roads safe all year round. In this future, human error is entirely eliminated from the road. But how can we achieve it?
Next-generation cost-effective sensing and imaging technology is the key to unlocking this future for the automotive industry. That means quantum dots (QDs) will be at the heart of our fully autonomous roadways.
Driverless vehicles through the ages
In the 1920s, car manufacturers experimented with what they called ‘phantom autos’ – cars that were driven by remote control, operated by someone following in another car or a low-flying aircraft.
In the 1970s, Japanese scientists created a vehicle that used cameras and an onboard analogue computer to navigate around specially marked roadways. And, through the 80s and 90s, Carnegie Mellon University’s Navlab project produced several semi-autonomous vehicles to refine the concept further, using cutting-edge video hardware and the latest supercomputers of the time.
This gave innovators the confidence to push driverless vehicles even further, incorporating light detection and ranging (LiDAR) technology into cars. LiDAR uses a range of narrow lasers and an array of photoreceptors to map terrain at high resolutions, which enabled automated detection of obstacles and allowed self-driving cars to venture off-road.
In 2005, Stanley, a LiDAR-equipped Volkswagen Touraeg, became the first vehicle to complete the DARPA Grand Challenge – an exacting 132-mile off-road course in the Mojave Desert. It was one of five vehicles to finish the course, all of which featured some kind of LiDAR system.
Making the future into reality
Today, most modern cars incorporate some level of automation, such as cruise control, which is enough to achieve SAE Level 1. But, while the concepts behind autonomous vehicles have been proven by the Navlab project – which saw the Navlab 5 Pontiac Trans Sport minivan complete a journey from Pittsburgh to San Diego almost entirely without human navigation in 1995 – scaling this technology has proven difficult, and mass-produced autonomous cars are yet to arrive on the market.
Classified from the perspective of sensors used, there are two technical routes to achieve autonomous driving: machine vision, and LiDAR.
Machine vision systems offer more promising immediate results. Using visible light based cameras and mmWave radar can help to build models of surrounding objects. This adds data to neural networks for pure visual calculations – effectively enabling vehicles to automatically ‘recognise’ things in their environment and respond.
However, this machine learning approach will only be as good as the data it has to work with. Simply relying on driver behaviour and visible light misses so many crucial data points. To create a truly autonomous driving system, that system must be able to see through thick fog, detect black ice at night, and be unaffected by glare from bright sunlight. This means the sensing systems of the future will need to look beyond what humanity can see – and it is here that the invisible infrared spectrum of light shows us the way to our driverless future.
The QD solution
Near infrared (NIR) and short-wave infrared (SWIR) sensors can empower the cameras relied on by autonomous vehicles, ensuring the machine learning algorithms that will power them have the optimal amount of data to work with. Infrared wavelengths, particularly the SWIR waveband, are largely unaffected by things like fog and can pick out obstacles even in low light conditions. Crucially, it can even be used to detect and differentiate between dry, wet, and icy road surfaces as ice has a slightly different SWIR absorption coefficient to water. This data, collected in real-time onboard a vehicle, will enable it to automatically adjust its steering configuration to navigate hazards safely.
SWIR sensors are predominantly made with SiGe and InGaAs compounds. These materials both offer advantages over silicon-based CMOS sensors in terms of absolute performance as they are sensitive to light in the SWIR spectrum. InGaAs sensors offer absorption of up to 2.1µm, however, they are prohibitively expensive, with a cost of around $10,000USD per unit. SiGe sensors are more affordable, but are held back by the band gap of germanium. Their pixel pitch, spectral range, and dark current also mean their performance is not at the level required to provide the reliable visual data a vehicle needs to drive safely on the roads.
QDs represent the most promising solution to these problems. These tiny nanoscale semiconductor materials can be tuned to absorb almost any wavelength of light across the visible, NIR, and SWIR spectra. Most importantly, they offer a high level of performance, with excellent pixel pitch and spectral range, at a fraction of the cost of either InGaAs or SiGe solutions. This makes QDs viable for production at the scale needed to make autonomous vehicles a reality.
At Quantum Science, our pioneering work on INFIQ® QDs will help the automotive industry overcome the many challenges it faces as it drives autonomous vehicle innovations closer to the market.
With INFIQ® technology rapidly approaching the scale-up phase, the missing piece of this puzzle has been found. It has put us within touching distance of the first mass-produced, fully self-driving car – an innovation more than a century in the making.