RF-based Direction Finding of UAVs Using DNN Samith Abeywickrama∗†, Lahiru Jayasinghe†, Hua Fu†, Subashini Nissanka†, and Chau Yuen† †Singapore University of Technology and Design, Singapore ∗National University of Singapore, Singapore Email: [email protected], {aruna jayasinghe,hua fu,yuenchau}@sutd.edu.sg Abstract—This paper presents a sparse denoising autoencoder MUSIC [10] and ESPRIT [11] are considered to be the (SDAE)-based deep neural network (DNN) for the direction most popular algorithms. However, MUSIC and ESPRIT are finding (DF) of small unmanned aerial vehicles (UAVs). It is inherently multi-channel techniques because those algorithms motivated by the practical challenges associated with classical DF algorithms such as MUSIC and ESPRIT. The proposed DF require a snapshot observation. This means, the base-band scheme is practical and low-complex in the sense that a phase data from all antenna elements should be extracted simulta- synchronization mechanism, an antenna calibration mechanism, neously so that a data correlation matrix can be formulated. and the analytical model of the antenna radiation pattern are Therefore, multiple channels should be coherent. However, in not essential. Also, the proposed DF method can be implemented most receivers, the digital down converter (DDC) chain uses a using a single-channel RF receiver. The paper validates the proposed method experimentally as well. coordinate rotation digital computer (CORDIC), which has a random start-up position on power up. The CORDIC therefore creates a random phase each time when the channels of the Index Terms—Drone surveillance, direction finding, UAV receiver are initialized, but remains constant throughout the tracking operation [12]. Therefore, calibrating this start-up phase values of each RF channel becomes necessary to realize a coherent I. INTRODUCTION multi-channel receiver. Clearly, this increases the hardware The use of civilian drones has increased dramatically in complexity and the power consumption. recent years. Likewise, drones are fast gaining popularity Most of the civilian drones use WiFi-like OFDM for their around the world [1–5]. However, drone use in a problematic communication. They are usually unknown, wideband, and manner has stirred public concerns. For example, in January transmitted in burst-mode. Such signal characteristics pose 2015 a drone crashed at the White House [6], raising concerns challenges with classical DF techniques. However, if only the about security risks to the government building; in March 2016 signal power measurements are utilized, performing DF is a Lufthansa jet came within 200 feet of colliding with a drone practically feasible even with such signals [13–16]. In [16], near Los Angeles International Airport [7]; and drones have signal power measurements that are obtained from a switched been accused of being used to violate the privacy and even beam antenna array are utilized to estimate the direction of a carry criminal activities [8]. These events give ample self- WiFi transmitter. As the actual radiation pattern of the antenna evident examples that developing a surveillance system for is vital for these methods, still it bounds with some practical suspect drones is of paramount importance. challenges. Therefore, we propose a practical and a low- The authors of [9] sought to detect a drone using Radio complex drone DF method in this paper, and our contributions Frequency (RF) as it can work day and night and at all can be summarized as fallows. weather conditions. Most of the commercial drones commu- arXiv:1712.01154v4 [eess.SP] 28 Dec 2018 To the best of authors’ knowledge there has not been nicate frequently with their controllers, and the downlink, any other method that involves deep neural network in the i.e., video signal and telemetry signals (flight speed, position, context of drone DF. We focus on a system which com- altitude, and battery level), between the drone and its controller prises a directional antenna array having N antennas, and is always present. To this end, this paper presents a drone a single channel receiver. By processing the signals that surveillance system, by eavesdropping on the communication are transmitted from the drone to its ground controller, the between a drone and its ground controller. The system can single channel receiver measures the received signal power estimate drone’s direction (or bearing) by processing the data at the each antenna using a RF switching mechanism. Then, transmitted from the drone to its controller using a single the obtained power values are fed to the proposed sparse channel wireless receiver. Therefore, no dedicated transmitter denoising autoencoder (SDAE)-based deep neural network is required at the surveillance system. (DNN). More precisely, the first hidden layer of the network RF based direction finding (DF) techniques have been well extracts a robust sparse representation of the received power studied, and the classical high-resolution techniques such as values. Then, the rest of the network utilizes this sparse The support of NSFC 61750110529 and SUTD-MIT International Design representation to classify the direction of the drone signal. Center is gratefully acknowledged. The first two authors contributed equally. It should be noted that a phase synchronization mechanism, where Gn(θ) is the real numbered antenna gain for the azimuth N-element angle θ, λ is the signal wavelength, and Directional Antenna Array 2π(n − 1) β (θ)= d cos − θ , (3) n N Single-pole-N-throw " # er Antenna Switch ¡ where d is the radius of the circular antenna array [18]. Since Single Channel we consider a practical DF method in this paper, s(k) and SDR . an(θ) are assumed to be unknown. Therefore, our objective . is to recover the azimuth angle θ, while the parameters s(k) . Softmax Classi Data Preprocessing . and an(θ) are unknown. We focus on a power measurements based approach. To this end, the ensemble averaged received signal power at the n−th Encoder antenna element can be given as SDAE-DNN 2 Pn =E |rn(k)| Fig. 1. The System Model. h i ∗ =E an(θ)s(k)+ nn(k) an(θ)s(k)+ nn(k) h 2 2 2 i an antenna gain calibration mechanism, and the analytical = |an(θ)| E |s(k)| +E |nn(k)| model of the antenna radiation pattern are not essential for 2 h 2 i h i = Gn(θ)Ps + σ , (4) this single channel implementation. The paper validates the proposed method experimentally through a software defined where 2 2 radio (SDR) implementation in conjunction with TensorFlow Gn(θ)= |an(θ)| , [17]. Furthermore, such an experimental validation for drone 2 DF is not common in the literature, and can be highlighted as Ps =E |s(k)| , another contribution of this paper. h i 2 2 The paper organization is as follows. The system model is σ =E |nn(k)| , presented in Section II. Section III discusses the proposed deep h i and E[·] denotes the expectation operator. Here, (4) follows architecture. Then, in Section IV, we validate the proposed from the fact that s(k) and n (k) are independent and method using experimental results. Section V concludes the n uncorrelated, i.e., paper. ∗ ∗ To promote reproducible research, the codes for generating E[s(k)nn(n)] = E[s (k)nn(n)]=0. most of the results in the paper are made available on the website: https://github.com/LahiruJayasinghe/DeepDOA. It can be observed that the received power values at the antenna elements ni and nj , where i, j ∈ {1, ··· ,N} and i 6= j, are 2 2 II. SYSTEM MODEL not identical, since Gi (θ) 6= Gj (θ). We have this property thanks to the gain variation of the directional antenna array We consider a system which consists of a single channel in [0 2π). Therefore, it is desirable to have an underlying receiver, and a circular antenna array equipped with N direc- N relationship (or a pattern) between {Pn}n=1 and θ. tion antennas, see Fig. 1. The antenna array is connected to the The proposed method is as follows. The receiver sequen- receiver using a non-reflective Single-Pole-N-Throw (SPNT) tially activates one antenna element at a time using the RF switch. The switching period is Ts. Suppose that a far- SPNT RF switch, and measures the corresponding received field drone signal impinges on the antenna array with azimuth power value. During the activation of n−th antenna, Pn is angle θ ∈ [0 2π). The received signal at the n−th antenna measured, where n ∈{1, ··· ,N}. A single switching cycle is element can be given as equivalent to N activations, starting from the first antenna to the N−th antenna. Let p = [P , ··· , P ]⊤ denote the power r (k)= a (θ)s(k)+ n (k), (1) 1 N n n n measurements corresponding to a single switching cycle. As ⊤ it is depicted in Fig. 1, x = [x , ··· , xN ] is obtained during where k is the sample index, an(θ) is the n−th antenna 1 response vector for the azimuth angle θ, s(k) is the drone the preprocessing stage, where transmitted signal as it arrives at the antenna array, nn(k) is P x n . (5) circularly symmetric, independent and identically distributed, n = N Pi complex additive white Gaussian noise (AWGN) with zero i=1 2 This means, x is the ratio between P and the summation mean and variance σ , and n ∈ {1, ··· ,N}. Here, an(θ) n P n follows the form of of all power values within the same switching cycle. In the next section, we discuss how the proposed network recovers 2π j βn(θ) an(θ)= Gn(θ)e λ , (2) θ from x. where T is the size of the training data set, β is a hyper parameter1, ρ is the sparsity parameter, T 1 ρ = h (xi) m T m er i=1 X is the average activation level of the m-th hidden unit where xi Corruption hm( ) denotes the activation of the m−th unit for the input .