字幕列表 影片播放 列印英文字幕 Roger. Xwing control ready to taxi. Xwing caravan clear for takeoff. This plane just took off on its own. Don't have to do a thing. There's a pilot in the cockpit but everything on this Cessna Grand Caravan is automated. Xwing, a start-up in the Bay Area took CNBC up for a test flight, where the pilot didn't touch the controls once. Xwing is a technology startup that develops technology to automate aircraft. Over the past 100 years, the technology inside airplanes has become more and more advanced, from jumbo jets to smaller cessnas. Most of the flying happening today has a high level of automation, but pilots are still doing a lot of work. On commercial flights, for example, pilots are telling the autopilot what to do, you don't just set it and forget it. But some see the next step to full automation as removing the pilot completely. Reliable Robotics is another Bay Area start-up working on doing just that. We started Reliable Robotics to create a new type of airline. An airline where instead of having pilots in a cockpit, you have pilots in a control center. Rose previously worked directly with Elon Musk on Tesla's first version of autopilot. We're definitely going to see autonomous aircraft well before autonomous driving at scale. Driving has a huge number of technical challenges as well as regulatory challenges. Whereas with aviation, it's a much more tractable technical problem and you have a regulatory environment that is ready to go. Boeing recently demonstrated a successful autonomous flight of its loyal wingman fighter jet, and Airbus recently performed fully autonomous test flights on one of its larger commercial aircraft. The technology is here, we want it to explore how it works, and when pilotless planes will become the norm. Before we look at what's happening today, let's take a quick look back at how we got here. In 1903, the Wright brothers took their first powered flight, and a few short years later, the first iteration of autopilot was created. Laurence Sperry's gyro stabilizer helped automatically balanced the plane so the pilot wouldn't have to. Fast forward to the 1950s and the rise of commercial aviation. At the time, five people were needed in the cockpit to fly the plane. Two pilots, a flight engineer, a radio operator and a navigator. That number decreased over the next few decades. In the late 70s and early 80s. We went from three crew to two crew. The level of automation today, of course is much greater in terms of what the workload of that crew actually looks like. One of the first digital autopilot systems was used in the lunar module that landed Apollo astronauts on the moon. Although Neil Armstrong had to turn it off to land in a safer area on the moon surface. Then in the 70s, NASA introduced digital fly-by-wire systems. Fly-by-wire is basically utilizing electronics rather than cables and pulleys, so to speak, to manipulate flight controls. It's practically instantaneous as it has to be. The F 16 fighter jet was the first mass production fly-by-wire jet. And that was a major breakthrough and it is now inconceivable that a jetliner would be designed without fly-by-wire flight controls Autopilot evolved into higher levels of automation. The military has flown remotely operated drones since World War Two, and has continued to improve the technology. But these military drones are not built to fly people and often require video streams for remote operators to fly them. One reason the military has continued to work on this is for safety. The constant adjustments a computer makes versus a human ensures for a much smoother and safer ride. 2017 was the safest year in aviation worldwide on record. In the US, there has not been a fatal crash since 2009. Here's how Reliable Robotics and Xwing's technology works. Both companies have developed software that allows them to remotely operate the plane from the ground, if anything were to go wrong. There's no joystick, you're not remotely sticking the plane with our system, we give the remote pilot essentially an interface that lets them specify all of this information. Load it all into the system and then press execute. And then that is then transferred through a SATCOM link up to the aircraft, You can make an aircraft autonomous, that aircraft still needs to communicate to the to a human, that human is the air traffic controller. Those operators don't fly the aircraft. Initially, they're going to be pilots because they need to know how to communicate with air traffic control and how these aircrafts integrate in the airspace. But they don't need to know how to fly or land or take off. That's done entirely by our system. Reliable Robotics already did a test flight without anyone in the aircraft and we got to go up with Xwing on one of its test flights. The remote operator initiated our takeoff and off we went. Do your instincts want to grab the yoke. To be honest it was very alarming the first time it landed itself. Even though the plane is flying on its own, a safety pilot must still be on board. So what we've done here is every time a pilot is involved interacting with the aircraft, we've automated that interaction, whether it be the throttles the brakes, the control surfaces. So we just got back from a fully autonomous flight with Xwing. And my first impressions are that it was pretty normal, the controls and the yoke were moving around by themselves. But besides that, it felt like a normal flight. You know, I was a little anxious going into this, knowing the pilot wasn't going to actually do anything. But overall, it wasn't scary at all. The planes are rigged with cameras, LIDAR and software among a few things that can identify other aircraft and detect potential issues. Rather than build a new plane, both companies converted a Cessna Grand Caravan. It's a lot easier and quicker to start with a tried and true platform and make modifications to it to convert into an unmanned aircraft than is to start from scratch. You can already make all the assumptions around the airplanes air worthiness, it has a track record of safety. And we can really focus with the regulator on what modifications were made to the aircraft to convert into an unmanned aircraft. There obviously is a lot of energy right now going into electric, vertical takeoff and landing urban Air Mobility, advanced term abilities, but those aircraft have a ways to go and they have their own technical problems that need to be overcome. When we started the company, we decided that solving the regulatory challenges was what was going to be our unique differentiator. And you don't need to invent a new type of aircraft to solve the regulatory challenges related to autonomous operations. I think a great deal has to do with where these systems are being operated and less about the systems themselves. If we're talking about operations over long rural, relatively uninhabited regions, you know, that's going to be a lot easier than vertical operations in urban areas. So when you see systems proven, it's probably going to be more in that rural setting, rather than in something urban. Both companies are not able to operate its pilotless planes commercially yet, because they're still working on getting approval from the Federal Aviation Administration. The nice thing about the FAA is that it does provide some flexibility for you to be able to design a system to get to those levels of safety and still fly it as you're designing. This used to be a type certified aircraft or commercial aircraft, that we downgraded to an experimental aircraft so that we could do a lot of flight tests without needing to demonstrate that level of safety prior to doing some of those flight tests. We gave the FAA a very thorough rundown of every aspect of our system, and how we intended not only to develop it, but also operate it and then show through test and analysis that it would be safe to do so. Similar to self driving cars, self flying planes will also need lots of data. We've been flying commercially with piloted aircraft on the cargo side for larger logistics companies since December of last year. This allows us to collect data along commercial routes at scale that feeds into the training or algorithms, but also the certification program. Data helps algorithms learn how to deal with all situations, including emergency landings. One of the problems that you need to solve in automating an aircraft through all phases of flight is you need to look at what could possibly go wrong in each phase, and you have to methodically break down all of these potential failures and develop a mitigation for every single one of those failures. One advantage of an automated aircraft system is it can recognize anomalous behavior much faster, then an onboard human might recognize and then respond to it much faster. So if you have for example, an engine failure, we can trip an alarm immediately and then begin executing an emergency response procedure. So the aircraft will immediately get into a most desirable or safe configuration. Notify the remote pilot who then in turn can notify air traffic control, and then the remote pilot can help guide the aircraft towards a desired emergency landing location. Both companies aim to bring autonomous planes to the air cargo industry first. The Global Air Cargo market is expected to grow from $56.48 billion in 2020 to 99.67 billion in 2025. Both plan to first fly cargo to remote areas in smaller planes. Cargo stands out as the biggest opportunity space, specifically small cargo planes, these planes tend to serve communities that without air wouldn't be served otherwise. And for these types of groups, we can serve them at a lower cost and higher frequency than what they'd enjoy today. Reliable Robotics is testing with a FedEx owned prototype. FedEx, UPS, DHL, etc. have actually been looking at automating these planes for quite a long time and they've been looking for solutions. That aircraft like the Cessna Caravan is by the numbers the world's most popular cargo plane. FedEx operates over 230 of these in the US alone. UPS through a bunch of partner companies has several 100 themselves as well. Rose, who is also a pilot worked at SpaceX before he worked on Tesla's autopilot. I previously had automated fully autonomous spacecraft that that went to the ISS at SpaceX, it could go to the International Space Station with no human intervention. I'd worked on self driving cars at Tesla that could go 70 miles an hour down the 280, thinking, why hasn't anybody done this in aviation? Clearly, we possess the technology. Why can't regular people experience automated flight? And the simple answer is, it's really the regulatory environment is not there yet. Though the companies are starting with cargo, the long term plan is to carry passengers as well. There's currently about 30 airports that make up 70% of air travel, we're gonna see smaller aircraft flying at higher frequencies. Out of more municipal regional airports, we actually have 5000 airports in the US available for public use today. But in order to use those, you either need to be a private pilot or you need to be exceptionally wealthy. And what automation will enable you to do is by lowering the cost of getting into air travel will be able to build systems that work much like an on demand ground transportation system like Uber or Lyft. This technology has the potential to open up a brand new market and significantly scale, you know, grow that market. Neither company knows when they will get approval, but both are optimistic it will be in the next few years. But when will we see commercial flights without a pilot? Though autopilot is very advanced commercial flights still require two pilots in the cockpit. The level of automation in commercial aircraft, even in a corporate aircraft is highly sophisticated, we can program the airplane to take off and and especially to land all by itself, we need to put the gear down and we need to put the flaps down and that's essentially all the pilot really has to do if it's programmed appropriately. Airbus has conducted fully autonomous flight tests on it's a350, which is a lot larger than the cessnas Xwing and Reliable Robotics are testing. The wide body jet achieved autonomous taxiing, takeoff and landing. Boeing has a number of defense platforms that are autonomous and publicly stated it's looking into pilotless aircraft. Safety is a big reason as to why. In the 1960s, the world's airline fleet had a rate of 27.2 accidents per 1 million departures. Today with the current generation of aircraft the rate is 1.5 accidents per 1 million departures. There are situations where an autopilot is better than actually flying your plane yourself. There are certain weather requirements that require us to use the autopilot in certain situations visibility down to a certain level and a certain approach requires that we do use an autopilot. While rare, 80% of all aviation accidents are caused by humans. That figure includes pilots, air traffic controllers, and mechanics. But the misunderstanding of automation has been implicated in accidents. For example, the 2009 Air France flight that crashed into the Atlantic Ocean on its way to Paris, investigators concluded that autopilot had turned itself off, and the pilots weren't able to take manual control over the plane. More recently, the flight control software of the two 737 max planes that crashed in 2018 and 2019 were cited as one of the reasons for the accident.