Hybrid drones that can be converted from vertical take-off and landing (VTOL) to fixed-wing flight are becoming increasingly popular in military and civilian applications. However, because the aerodynamics of such aircraft can change dramatically during flight, it is very difficult to control this design. Now scientists are using artificial intelligence (AI) to help automatically design the remote control of any such hybrid drone. Fixed-wing aircraft can be more efficient than multi-rotor aircraft, resulting in a longer range, greater endurance and higher speed at similar power. On the other hand, multi-rotor aircraft can not only hover and fly at low speed, but also take off and land without a runway or complicated launch and recovery devices. Hybrid aircraft capable of switching between fixed-wing and VTOL flights have many advantages of both, which makes them potentially useful for civilians, for example, they can remove them from the farm without requiring a landing strip Flying down, just like military, they can fly out of the jungle, mountains, ship decks and urban battlefields with the same degree of freedom. There are many types of convertible aircraft, such as tilted wings (the propeller-mounted wings can rotate between vertical and horizontal positions) and tailstock aircraft, whose tails take off and land and land, and can tilt horizontally to fly forward. Conversion evolution Since the 1950s, the U.S. military has begun to study convertible aircraft, but manned prototypes often suffer from mechanical complexity and other problems, such as tilting the wing while converting from one flight form to another.
How often to stall when flying, or how the rear seat bodyguard looks awkward to let the pilot sit there during takeoff and landing. Given that unmanned aerial vehicles generally have less mechanical complexity and payload constraints than manned aircraft of the same size, the emergence of unmanned aerial vehicles (UAVs) has renewed interest in hybrid aircraft. Modern improvements in the efficiency of electric motors and the increasing miniaturization of electronic components have also made hybrid vehicles more feasible than ever. However, remotely controlling hybrid drones is still a challenge. Given how active the rotors and wings are, when aerodynamics are particularly complex, scientists usually develop controllers not only for helicopter and aircraft modes, but also for transitions between these modes. Therefore, designing controllers for hybrid drones currently requires experts to "manually adjust hundreds of parameters," said Xu Jie, a researcher with a doctorate in computer science from the Massachusetts Institute of Technology. In addition, given that there are often huge differences between various types of hybrid drones, researchers often cannot transfer controllers from one type to another, so they need to design controllers for each new hybrid drone from scratch . All of these labor-intensive and time-consuming tasks have helped to explore only a small portion of possible hybrid drones so far. Artificial intelligence solutions Now, Xu and his colleagues have developed a method for automatically designing controllers for hybrid drones. Their system can design a single controller for all different flight modes of hybrid drones, and can be applied to any type of hybrid drones. The researchers adopted an AI system called a neural network, in which components called "neurons" are fed with data and cooperate to solve problems such as face recognition. The neural network repeatedly adjusts the connections between its neurons, and sees whether the resulting behavior patterns can better solve the problem. Over time, the network has discovered which mode is best for computing solutions. Then, it uses these as default values to mimic the learning process in the human brain. In the new system, users first design the geometry of the hybrid drone by selecting components from the data set. Then, the system runs the design through the simulator to calculate the flight performance of the design. The simulator takes into account practical issues, such as random sensor noise and control signal delay. Then, the neural network automatically starts to learn how the UAV controller achieves the best performance in the simulation. The researchers used laser cutting and 3D printing technology to create three different types of hybrid drones, thus validating their system and successfully conducting actual flight tests using the resulting controller. "The biggest advantage of this work is to speed up the design process of hybrid drones," Xu said. "Everyone can use this new drone design." The controller does not need to distinguish between helicopter mode and flight mode, nor does it need to explicitly deal with the transition between modes. For example, the controller will automatically adjust the direction of the tail rotor hybrid UAV purely based on its speed, setting it to helicopter mode at low speed and airplane mode at high speed. Xu cautioned that the system currently only supports simple flights. The researchers plan to study ways to increase the maneuverability of the design, such as the position or shape of the rotor or wing, so that the system "can calculate more complex motions." Researchers have published the code of their system on the software development platform GitHub. "We hope that everyone interested in this technology can share this technology," Xu said.
Contact: Zhuhai SVFFI Aviation Co., Ltd.
Add: Zhuhai City, Guangdong Province, Jinwan District Aerospace New Town Planning Exhibition Hall, 3rd Floor