25-09-2020 | By Sam Brown
Planes have been a key transport technology for more than 100 years and historically have always had human operators at the controls. What are self-flying planes, what technologies would they be reliant on, and what challenges do they face?
Since the dawn of time, humans have looked to find new methods for getting around. The domestication of the horse led to the ability to travel great distances while hauling commodities, and it was this that enabled trade. The development of sea travel allowed for people to travel across vast oceans carrying cargo from all corners of the world, and this led to the rise and fall of many empires. The introduction of the train saw the ability to move across vast amounts of land in a short time, and it was this technology that helped to introduce fresh fish and chips into England's inner cities such as Birmingham. But of all modes of transport that have been invented, the plane is by far the one that stands out the most; the ability to circumnavigate the globe in a day while flying at speeds never thought possible. Before the COVID-19 pandemic, it was estimated that there were a total of 40 million flights per year, and these flights include the transportation of both people and cargo. But, like all technology, the times are changing, and advances in AI, LIDAR, RADAR, and many other sensory techs are seeing smarter and better planes every day.
Just as the name suggests, a self-flying plane is one that pilots itself including take-off, mid-flight, and landing. While self-flying planes are currently in development, auto-pilot features make the vast majority of a plane flight automated. There are even auto-landing features that have been tested, proven, and installed in planes, including Airbus and Boeing. However, because a human pilot is present, these planes cannot be considered self-flying, and the act of placing a human pilot at the controls defeats the goal of self-flying planes; do not require human intervention.
Self-flying planes face the same challenges as those faced by self-driving cars, but also face other challenges too. The challenges faced by self-flying planes fall into two separate categories; human and technological.
The technological challenges faced by self-flying planes are very complex and simple at the same time. Currently, auto-pilot systems exist which can keep a plane in the air, and correctly change its flight path. Planes are also alone in the air and do not need to avoid other planes crossing their path (or at least, not a regularly as cars do on the road), or perform emergency manoeuvrers often. However, planes are also required to avoid areas of turbulence, land, take-off, and take instructions from air traffic control when taxing around the airport. Planes may also be subjected to requests from other aircraft such as military jets, who may want to escort aircraft to a different location, and the plane needs to be able to comply.
The human challenges faced by self-flying planes stem from trusting a system to operate an aircraft correctly. While automated systems can be shown to be statistically safer, there are instances where automated systems make fatal mistakes that would not be done by human operators. For example, the auto-pilot feature in some Tesla cars struggled to correctly read a specific road in the US which cause a fatality, and other instances of the system not detecting vehicles to its side. While the fault of the instances technically falls with the drivers (as Tesla cars are not self-driving by definition), such mistakes can bring distrust. Using statistics alone to demonstrate safety is also partly flawed as it does not take into account emotional trauma and fear. For example, aeroplanes are statistically one of the safest methods of travel, but a plane that fails 35,000 feet in the air that hits the ground not only kill all occupants on board (0% survival), but the time taken to fall can leave passengers in a state of utter fear for several minutes.
The first technology that a self-flying plane will be heavily reliant on is AI. AI will be key to self-flying planes being able to take instructions from human operators at air traffic control, ability to recognise atmospheric conditions to take action if needed (i.e. suspicious cloud formations), recognise debris on runways, and read signs at airports (such as taxi lanes). AI will also allow planes to learn from their flights to understand better how to minimise fuel consumption, avoid areas that provide navigational difficulties, and reduce stresses on the plane. Unlike human pilots, AI driving planes could even actively monitor sensors around the plane to look for structural strain, vibrations, and potential engine failure to take action before problems arise.
Self-flying planes will be heavily reliant on sensor technologies that allow for a complete understanding of the surrounding environment. To start, LiDAR is a technology that allows for mapping, and this will be crucial in landing and take-off. As a plane lands, the runway must be clear of debris, and LiDAR can be used to determine if debris is present. LiDAR is also useful for object avoidance and thus can be used in taxi applications (i.e. planes that move around on the runway). Wireless technologies such as 5G and StarLink that allow the plane to connect to a global network will be invaluable. Unlike cars, planes move incredibly fast, so fast that it is near impossible for a pilot to react in time when seeing a plane coming in the opposite direction. The same applies to self-flying planes, but a connected network with all listed planes will allow for self-flying planes to avoid nearby aircraft as well as receive on-the-fly (pardon the pun), updates. Remote connections also allow for emergency operators to take over planes struggling to cope with a new environment or those that have faulted.
Self-flying planes have a long way to come, but several tests of self-flying planes show that it is more than possible, even practical. The first deployment of such planes will be for transporting cargo; no one minds if their item is lost in the post, but no one wants to be in a falling plane. It is more than likely that pilots will continue to be needed for the next 30 years whereby all self-flying planes require a human operator at the controls if something does go wrong, and they can take over.
Small LiDAR ICs offer higher bandwidth for faster self-driving cars