Indoor Localization System For Quadrotors

Nowadays, with the rapid development of remote control and autonomous technology, the usage of Unmanned Aerial Vehicles(UAVs) in civilian or military application has started to become common. Meanwhile, the expanding application range of UAVs makes new demands on the localization technique for UAVs. In the field of indoor localization, since GPS is not available, how to locate fast-moving UAVs such as quadrotors is still a challenging topic. Most of the previous works are based on vision or ultrasound detection. However, we have to notice that there exist some drawbacks such as high cost and high energy consumption. Thus we aim to apply the Wi-Fifingerprint-based method on quadrotors. The main benefit of this method is that it requires no additional infrastructure besides pre-existing WLAN APs, and thus is easy to deploy.However, existing Wi-Fi fingerprint-based indoor localization systems cannot be directly used to locate quadrotors. In practice use, both the unstable access points and the high-speed ?ight will obviously increase location error. Thus, according to the features of quadrotor’s ?ight, we design two localization methods: Localization for Static Quadrotors, and Localization for Dynamic Quadrotors. We implement two subsystems for the two methods respectively.For Localization for Static Quadrotors, we mainly leverage emerging smart access points(APs) equipped with two-mode antennas. A smart AP can change the working mode of its antennas, and thus it can generate two different RSS fingerprint maps. Based on this feature, a single smart AP can act as two normal APs in normal WiFi-RSS based indoor localization systems. Thus we can reduce the number of necessary APs, and maintain the same localization accuracy. To realize this method, we design 1) a synchronous localization strategy to manage the synchronization between the mode switches of APs and the fingerprint measurements of quadrotors, and 2) a self-adaptation adjustment method to adapt to the unstable wireless transmission and diversity of quadrotors. Experimental results demonstrate that while maintaining the same accuracy as normal WiFi-RSS-based indoor localization systems, this subsystem is able to reduce the number of necessary APs by 50%.For Localization for Dynamic Quadrotors, we propose to make use of historical localization results to help amend the ?ight path of quadrotors. Moreover, quadrotors can collect the data about their ?ight states using their sensors and thus help the localization process. Moreover, since the workload of RSS training data collection in 3D case is huge, we use a method based on interpolation to reduce the sampling rate of RSS. To realize this method, we design 1) a series of methods including path estimation, pathfitting and location prediction to reduce the adverse impact of high-speed ?ight; 2) a 4-D RSS interpolation algorithm to reduce the efforts needed to collect RSS training data with consideration of their statistical properties. Experimental results demonstrate that this subsystem reduces the average location error by more than 50% compared with normal WiFi-RSS-based systems, and the workload of RSS training data collection is reduced by more than 80%.