Development of Autonomous Aero-Robot and Its Application to Safety and Disaster Prevention

本研究的目标是开发一种完全自主的飞行机器人，能够在大地震、火灾、火山喷发等灾害场合支持各种活动。第一年，不仅进行了飞行机器人硬件的基本设计和开发，还进行了飞行模拟器的开发。研制了先进飞行控制系统所必需的GPS-INS组合导航系统，使定位误差降至2 cm。此外，还提出了利用神经网络设计针对各种不确定性的鲁棒控制系统的方法。对于控制界面，开发了语音指挥系统，并对其进行了评估。第二年，针对先进的飞行控制系统，对飞行机器人的硬件进行了改进，并利用神经网络或自适应控制器成功地实现了自主飞行，结果表明，所设计的控制器明显优于传统的飞行控制器。此外，还对语音指挥系统进行了改进，使其能够足够稳定地应用于嘈杂的环境中，从而使指挥系统在实际应用中变得更加实用。第三年，开发了适用于灾害现场信息收集的系统。由于采用了基于Web的数据报传输，所开发的航空视频和照片采集和传输系统比以前更易于使用，成本也更低，飞行实验表明了其有效性。在此基础上，提出了利用立体相机和GPS-INS组合导航系统制作三维地形图的方法，结果表明该方法制作的三维地形图精度高，分辨率高，可用于灾情现场勘察等。

The objective of this research is to develop a complete autonomous aero-robot, which can support various activities in an occasion of disaster such as large earthquakes, fires, and volcanic eruption.In the first year, not only the basic design and development of the hardware of the aero-robot but also development of the flight simulator were carried out. GPS-INS hybrid navigation system, which is necessary for advanced flight control systems, was developed and the error of the position is reduced to only 2 cm. Moreover methods to design robust control system against various kinds of uncertainties by use of neural networks were proposed. For a control interface, voice command system was developed and evaluated. Then the difference between novice operators and skilled operators became clear as a result of analysis of the flight data.In the second year, the hardware of aero-robot was improved for advanced flight control systems, and we made success in autonomous flight by use of controllers by use of neural networks or adaptive controller, and it is shown that controllers designed by the proposed methods are much superior to the conventional flight controllers. Moreover, the voice command system was also improved so that it can be used in noisy environment with enough stability, then the command system became much more useful in practical applications.In the third year, systems, which were useful in information gathering at disaster site, were developed. Because web based datagram delivery is adopted, the developed system to gather and deliver aerial video and photograph is much easier to use and much cheaper than before and flight experiments showed its effectiveness. Moreover a method to make 3D terrain map by use of a stereo camera and GPS-INS hybrid navigation system together is developed and it is shown that the 3D map is very accurate and has enough resolution for investigation at disaster site and so on.