Unmanned Aerial Vehicle-Aided Wireless Sensor Network Deployment System for Post-disaster Monitoring

Transcription

1 Unmanned Aerial Vehicle-Aided Wireless Sensor Network Deployment System for Post-disaster Monitoring Gurkan una 1, arik Veli Mumcu 2, Kayhan Gulez 2, Vehbi Cagri Gungor 3, and Hayrettin Erturk 4 1 rakya University, Department of Computer Programming, Edirne, URKEY 2 Yildiz echnical University, Electrical-Electronics Faculty, Control and Automation Eng. Dept., Istanbul, URKEY 3 Bahcesehir University, Faculty of Engineering, Department of Computer Engineering, Istanbul, URKEY gurkantuna@trakya.edu.tr, cagri.gungor@bahcesehir.edu.tr, 4hayrettinerturk@gmail.com, {tmumcu,gulez}@yildiz.edu.tr Abstract. his paper presents design strategies of using unmanned aerial vehicles (UAVs) to deploy wireless sensor networks (WSNs) for post-disaster monitoring. Natural disasters are unforeseeable events which cannot be prevented. But some recovery procedures can be followed to minimize their effects. Post-disaster monitoring is important to estimate the effects of disasters, which in turn is used to determine recovery procedures to be followed. We propose an UAV-aided unattended WSN deployment system. he system is a post-disaster solution which can be used anywhere required. In this study, we mainly evaluate the efficiency of localization and navigation performance of the proposed system. Our simulation studies with an AirRobot quadrotor helicopter in Unified System for Automation and Robot Simulation (USARSim) simulation platform show that UAVs can be used to deploy WSNs after disasters to monitor environmental conditions. Future work includes implementing the system using a hexarotor helicopter. Keywords: Wireless sensor networks, unmanned aerial vehicles, localization and navigation systems, Global Positioning System, inertial navigation system, Kalman filter. 1 Introduction Although natural disasters cannot be prevented, their effects can be minimized through proper warnings and post-disaster recovery procedures. After the disasters monitoring of environmental conditions is a vital process to prevent possible hazardous effects to human health. Unmanned Aerial Vehicles (UAVs) are used for various civilian and non-civilian tasks including surveillance and reconnaissance operations, monitoring, and aerial photography. Another envisioned use of UAVs is the deployment of wireless sensor networks (WSNs). D.-S. Huang et al. (Eds.): ICIC 2012, CCIS 304, pp , Springer-Verlag Berlin Heidelberg 2012

2 Unmanned Aerial Vehicle-Aided Wireless Sensor Network Deployment System 299 WSNs are distributed systems of sensor nodes which are interconnected over wireless links [1], [2]. During a deployment phase, sensor nodes can be placed by using different attended/unattended methods such as such as dropping from an aerial vehicle [3], [4], delivering in an artillery shell, or rocket, and placing one by one by either a human or a mobile robot [5], [6], [7], [8]. Deployment phase is the first step of making WSNs operational. In this study, we propose an unattended autonomous WSN deployment system by using an UAV. Different from existing studies in the literature which mostly explain theoretical background and practical uses of UAV-aided wireless sensor node deployment; in this paper we mainly investigate localization and navigation performance of an UAV responsible for WSN deployment. Since localization and navigation capabilities are the main requirements of all autonomous vehicles, successful implementation of the proposed system depends on these capabilities. he paper is organized as follows. UAV aided WSN deployment system is explained in Section II. Simulation studies are given in Section III. Conclusions of the paper and future work are given in Section IV. 2 Unmanned Aerial Vehicle Aided WSN Deployment In this study, we propose using UAVs for unattended deployment of wireless sensor nodes. In this system, an UAV follows a predetermined trajectory and drops wireless sensor nodes at predetermined intervals at predetermined locations. Due to several factors a location where a wireless sensor node is dropped for deployment cannot be determined accurately during flight multiple nodes are deployed together. In addition, some nodes may become damaged during deployment. Also communication cannot be established with some nodes since they are out of range. hese are common problems in unattended sensor node deployments. he localization and navigation system plays an important role in the design of the proposed system. It is based on the Kalman Filter (KF). o improve the system s performance we couple Global Positioning System (GPS) receiver and Inertial Navigation System (INS) sensors. Since GPS and INS sensors are common in aerial vehicles, in this study we preferred these sensors. GPS receivers provide absolute information of position and speed, and they do not need information about their previous states to produce a navigation solution [9], GPS signals carry information to determine a GPS receiver s position. he signal is composed of navigation data and a pseudo-random code [10]. GPS receivers cannot be used alone for navigation systems due to the requirement of having an open view of at least four GPS satellites and their low sample rates [10]. he INS consists of an inertial measurement unit (IMU) with an accelerometer and a gyro in addition to a navigation computer. It provides position, velocity and attitude angles. he accelerometer provides a non-gravitational acceleration, and the gyro keeps track of the orientation. Different from GPS receivers, INS systems are not subjected to interferences and reception limitations. But when they are used alone, some drifts are experienced. We combine the advantages of these sensors. An INS sensor keeps the track to the actual position, velocity and attitude with the aid of a GPS receiver. Long term accuracy of the GPS receiver is utilized to reduce the drifts in the INS outputs.

3 300 G. una et al. Fig. 1. Using UAVs to deploy WSNs Centralized architectures designed for GPS-INS integrations generally provide more accurate solutions, but they require intensive computations in addition to requiring to access raw IMU and GPS data. Hence, considering the specifications of the UAV to be used in our field tests, we designed a loosely coupled direct feedback GPS-INS architecture shown in Fig. 2 using the GPS-INS integration architectures in [11] and [12]. he architecture shown in Fig. 2 involves three main steps. hese steps are data preprocessing, filtering, and smoothing. his architecture estimates systematic errors in the IMU outputs. o do this it uses observations obtained from the GPS receiver. In case of GPS outages, very common in urban areas, the IMU error estimates obtained with KF based on the previously accumulated GPS data can compensate inertial navigation errors. In Fig. 2, x and y represent position coordinates. ˆx and ŷ are position estimations. he INS uses θ and ν to calculate a position. Here, θ represents incremental angles derived from gyros, and ν represents incremental velocities derived from accelerometers. he GPS receiver uses s and ϕ to yield a position. Here, s represents measured range, and ϕ represents carrier signal phase. navigation frame. n C represents the rotation matrix from sensor to s 2.1. Filtering Fig. 2. GPS-INS integration architecture In the proposed navigation architecture, the observation delivered to the filter is actually the observed error of the inertial navigation, and the filter estimates the errors

4 Unmanned Aerial Vehicle-Aided Wireless Sensor Network Deployment System 301 in this inertial navigation solution. Hence, the inertial navigation equations were linearised, and the filter took on a linear form shown in Fig. 2. KF is a statistical recursive algorithm which operates recursively in two stages: Prediction and Update [13], [14]. KF provides an estimate of the states, ˆx( k k ), at time k given all observations up to time k. ˆx( k k 1) shows the estimate of the state at time k given information up to time k-1, and is called the prediction. Under the assumption that observation and process noises are zero mean and uncorrelated, KF provides an optimal Minimum Mean Squared Error (MMSE). KF uses a landmark based map and In the prediction stage, the command u(k) and the UAV motion model are utilized to estimate the UAV s location. hen, in the update stage, to update the landmark s position and to refine the estimation of the UAV s location, the new observation z(k) from an exteroceptive sensor is used. Fig. 3 shows the steps of KF based localization and navigation approach. Fig. 3. he steps of KF based localization and navigation In the GPS-INS integration architecture in Fig. 2, when an observation arrives, the filter estimates the error in the vehicle states. he observation is the observed error of the inertial navigation system. Whenever the GPS receiver provides position and velocity data, the observation error becomes, z p( k ) P INS ( k) - P GPS ( k) z( k)= = z v( k ) V INS ( k ) - V GPS ( k ) (1) When position and velocity errors are integrated, the observation model becomes z p( k ) P INS ( k) - P GPS ( k) z( k)= = z v( k ) V INS ( k ) - V GPS ( k ) (P ( k) + δ P( k)) - (P ( k) - v P ( k)) = (V ( k) + δ V( k)) - (V ( k) - v V ( k)) δ P( k) v P ( k) = δ V( k) + v V ( k) (2)

5 302 G. una et al. he observation is the error between the position and velocity obtained from the INS and that of the GPS receiver. he uncertainty in the observation is reflected by the noise of the observation of the GPS receiver. he architecture brings an advantage that tuning the filter is only based on the observation noise matrix Smoothing Optimal fixed-interval smoothers provide optimal estimates using measurements from a fixed interval. here are several commonly used smoothing algorithms. In this study we preferred he Rauch-ung-Striebel (RS) smoother. RS smoother is a fixedinterval two-pass smoothing algorithm which is commonly used for bridging GPS outages in the post-processing mode. In this algorithm, the standard Kalman estimate and covariance are computed in a forward pass, and then the smoothed quantities are computed in a backward pass. Forward pass: Backward pass: xˆ = Axˆ t+ 1 t t t Pt + 1 t = APt t A + Q t+ 1 t+ 1 t t+ 1 t 1 ( ) ( C ) K = P C CP C + R xˆ = xˆ + K y xˆ t+ 1 t+ 1 t+ 1 t t+ 1 t+ 1 t+ 1 t P = P K CP t+ 1 t+ 1 t+ 1 t t+ 1 t+ 1 t L = P A P 1 t t t t+ 1 t ( + + ) xˆ = xˆ + L xˆ xˆ t t t t t 1 t 1 t P = P + L ( P P ) L t t t t t+ 1 t+ 1 t t (3) (4) ˆ t x is the optimal estimate of state at time t, P t is the measure of uncertainty, and L t is gain matrix. 3 Performance Evaluations In this study, we mainly concentrate on the performance of the localization and navigation system. o show the localization and navigation performance of an UAVaided WSN deployment system, we performed simulation studies by using USARSim and Robot Operating System (ROS). We created a ROS interface to USARSim in Python. USARSim is based on the Unreal ournament (U) game engine, and is a simulation of robots and environments. USARSim serves both as a general purpose research tool and as the basis for the RoboCup rescue virtual robot competition [15], [16]. he interface we developed interfaces with the USARSim server, receives input from GPS and INS sensors of the UAV and sends commands to the UAV. We

6 Unmanned Aerial Vehicle-Aided Wireless Sensor Network Deployment System 303 implemented INS and GPS integration by using the KF. Source codes of the simulation application are available upon request. In our simulation studies, we used an AirRobot [17] autonomous quadrotor helicopter. In USARSim simulation platform, USARBot.AirRobot class is used to represent an AirRobot. Simulated AirRobot is equipped with one tilt-only color camera as the exteroceptive sensor by default. Since the system uses GPS and INS measurements to localize and navigate the AirRobot, we placed a GPS receiver and an INS on it. he GPS sensor finds the current AirRobot position in meters and converts it to latitude and longitude. Since USARSim worlds inherently do not have a GPS coordinate associated with them, we edited the USARBot.ini file and added ZeroZeroLocation inside the GPSSensor section to map a virtual location to a real one. he INS sensor simulates a physical INS sensor by using angular velocities and distance traveled [16]. During localization and navigation simulations we only calculated positional errors from the real trajectory. Positional errors from the East and North are shown in Fig. 5 (a) and Fig. 5 (b). Only first 250 seconds of the simulation are shown in the figures. When calculating the positional errors we compared the filter outputs with the ground truth values. Fig. 4. he simulated AirRobot Fig. 5. (a) rajectory Errors - X (in meters), (b) rajectory Errors Y (in meters) rajectory errors of the simulated UAV are around less than or around 0.5 m for X and Y. hese minor deviations can be neglected in UAV-aided WSN deployment system. he results of our simulations show that an UAV can localize and navigate itself successfully during deployment when loaded with a priori map and a waypoint.

7 304 G. una et al. In addition to the simulations, we are planning to conduct field tests with an autonomous hexarotor helicopter Flybox shown in Fig. 6. he specifications of Flybox can be found in [18]. Since it runs Robot Operating System (ROS) [19] on Ubuntu, we are going to develop our application by using ROS. Fig. 6. Flybox autonomous hexarotor helicopter 4 Conclusions his paper focuses on using UAVs to deploy WSNs in case of disasters. Since disasters are mostly unforeseeable events, recovery procedures to minimize the effects of these events need to be planned. Monitoring of the disaster area is important to determining the recovery procedures. Proposed system takes the benefit of UAVs to establish a monitoring system consisted of wireless sensor nodes. herefore a single UAV team can serve many cities in a small region. he proposed system has a practical use after disasters. In this study, we specifically address efficiency of the system s localization and navigation aspects. he results of our simulation studies show that an UAV can follow a predetermined trajectory successfully to deploy a WSN for post-disaster monitoring. Future work includes implementing field tests. Acknowledgment. his research has been supported by Yildiz echnical University Scientific Research Projects Coordination Department. Project Number: ODAP01 and Project Number: KAP05. References 1. Akyildiz, I. F., Su, W., Sankarasubramaniam, Y., Cayirci, E.: Wireless sensor networks: a survey. Computer Networks, 38(4): (2002) 2. Gungor, V. C., Hancke, G. P.: Industrial Wireless Sensor Networks: Challenges, Design Principles, and echnical Approaches. IEEE ransactions on Industrial Electronics, 56(10): (2009)

8 Unmanned Aerial Vehicle-Aided Wireless Sensor Network Deployment System Ollero, A., Bernard, M., Civita, M. L., Hoesel, L. V., Marron, P. J., Lepley, J., Andres, E. D.: AWARE: Platform for Autonomous self-deploying and operation of Wireless sensor-actuator networks cooperating with unmanned aerial vehicles. In: Proceedings of the 2007 IEEE International Workshop on Safety, Security and Rescue Robotics, pp. 1-6 (2007) 4. Corke, P., S. Hrabar, R. Peterson, D. Rus, S. Saripalli, G. Sukhatme: Autonomous Deployment and Repair of a Sensor Network using an Unmanned Aerial Vehicle. In: Proceedings of the 2004 IEEE International Conference on Robotics and Automation (ICRA), pp (2004) 5. Wang, Y., Wu, C. H.: Robot-Assisted Sensor Network Deployment and Data Collection. In: Proceedings of the 2007 IEEE International Symposium on Computational Intelligence in Robotics and Automation, pp (2007) 6. Younis, M., Akkaya, K.: Strategies and techniques for node placement in wireless sensor networks: A survey, Ad Hoc Networks, 6(4): (2008) 7. Suzuki,., Kawabata, K., Hada, Y., obe, Y.: Deployment of Wireless Sensor Network Using Mobile Robots to Construct an Intelligent Environment in a Multi-Robot Sensor Network. In: Advances in Service Robotics, pp (2008) 8. Suzuki,., Sugizaki, R., Kawabata, K., Hada, Y., obe, Y.: Autonomous Deployment and Restoration of Sensor Network using Mobile Robots, International Journal of Advanced Robotic Systems, 7(2): (2010) 9. Guivant, J. E., Masson, F. R., Nebot, E. M.: Simultaneous Localization and Map Building Using Natural Features and Absolute Information, Robotics and Autonomous Systems, 40(2-3):79-90 (2002) 10. Huang, J., an, H.-S.: A Low-Order DGPS-Based Vehicle Positioning System Under Urban Environment, IEEE ransactions on Mechatronics, 11(5): (2006) 11. Sukkarieh, S.: Low Cost, High Integrity, Aided Inertial Navigation Systems for Autonomous Land Vehicles, Ph.D. hesis, University of Sydney (2000) 12. Giremus, A., Doucet, A., Calmettes, V., ourneret, J.-Y.,: A Rao-Blackwellized particle filter for INS/GPS integration. In: Proceedings of the 2004 IEEE International Conference on Acoustics Speech and Signal Processing ICASSP- 04, 3: (2004) 13. Dissanayake, G., Newman, P., Clark, S., Durrant-Whyte, H. F., Csorba, M.: A Solution to the Simultaneous Localization and Map Building (SLAM) Problem. IEEE ransactions on Robotics and Automation, 17(3): (2001) 14. hrun, S., Burgard, W., Fox, D.: Probabilistic Robotics. MI Press, Cambridge (2005) 15. Carpin, S., Lewis, M., Wang, J., Balakirsky, S., Scrapper, C.: USARSim: a robot simulator for research and education. In: Proceedings of the 2007 IEEE Conference on Robotics and Automation, Roma, pp (2007) 16. USARSim (2011), AirRobot (2011), Flybox (2011), %20FlyboXScientificFlyerV2.pdf 19. ROS (2011),