Automatic Docking System for Recharging Home Surveillance Robots

Transcription

1 428 IEEE Transactions on Consumer Electronics, Vol. 57, No. 2, May 2011 Automatic Docking System for Recharging Home Surveillance Robots Guangming Song, Member, IEEE, Hui Wang, Jun Zhang, and Tianhua Meng Abstract This paper presents the development and characterization of a surveillance robot with automatic docking and recharging capabilities for home security. The proposed system is composed of a surveillance robot and a docking station. The palm-sized surveillance robot has a triangular shape with three wheels. It communicates with the general wireless home router through WiFi. It communicates with the docking station through ZigBee and serves as a mobile wireless sensor network gateway. The docking station has a trapezoidal structure with an arc-shaped docking interface. A docking method based on the self-localization of the robot and the infrared detectors of the docking station is proposed. The robot can return to the docking station for recharging operations when the on-board battery is too low. The experimental results show that the prototype robot achieved a success rate of 90% after 60 different docking attempts 1. Index Terms Home security, home robot, surveillance robot, docking. I. INTRODUCTION With the rapid development of microelectronics and wireless communication technologies, mobile robots are being widely used in industrial automation, home automation, hospitals, entertainment, space exploration, military, etc [1]. In recent years, as the size and the cost of mobile robots have decreased significantly, they are finding increasing uses in home environments. More and more mobile robots are now working around us and they will help us a lot in our daily lives [2] [3]. Various home robots have been proposed to do housework such as cooking, cleaning, house plant watering, pet feeding and taking care of children [4] [5]. Home security is one of the typical applications of home robots. In traditional home security systems, monitoring devices are usually mounted on fixed locations such as doors, windows and walls. A home surveillance system based on an embedded system with multiple ultrasonic sensor modules has been presented in [6]. If any intruder passes through the ultrasonic sensing area, the ultrasonic transmission will be blocked by the human body. The authors use a Majority Voting Mechanism (MVM) to process the output signals from multiple ultrasonic receivers. An automatic video-based 1 This work was supported in part by Natural Science Foundation of China under Grant Guangming Song, Hui Wang, Jun Zhang and Tianhua Meng are with the School of Instrument Science and Engineering, Southeast University, Nanjing , China ( mikesong@seu.edu.cn). Contributed Paper Manuscript received 04/14/11 Current version published 06/27/11 Electronic version published 06/27/ /11/$ IEEE human motion analyzer for consumer surveillance system has been proposed in [7]. It is a framework for semantic analysis of human behavior from a monocular surveillance video captured by a consumer camera. The proposed framework captures the human motion, classifies its posture, infers the semantic event, and conducts the 3-D scene reconstruction. The design and implementation of a low cost GSM/GPRS based wireless home security system has been presented in [8]. The proposed system can response rapidly to alarm incidents and has a friendly user interface including a LCD and a capacitive sensor keyboard. A wireless access monitoring and control system based on the digital door lock is proposed in [9]. The proposed system is based on ZigBee network protocol, which includes the ZigBee module, the digital door lock module, the human detection module, and the ZigBee relay module. Although most of the current home security systems can work normally, it is inconvenient to deploy and maintain a lot of sensors and accessories everywhere in the rooms. Due to irregular room structures and various physical limitations of sensors, there often exist some regions that can not be covered by the sensors. In view of these drawbacks, a more flexible and more efficient solution for home security is to deploy a mobile robot equipped with surveillance devices such as pyroelectric infrared sensors and cameras. Many researchers worldwide are now engaged in designing various mobile surveillance robots for home security applications. The development of a patrol robot system for home security with some interaction functionalities has been presented in [10]. The system integrates a variety of sensors to gather environmental information and detect abnormal events such as fire alarm, intruder alert and gas leakage. The authors want to implement some dedicated human-robot interactions that will help create pleasant experiences when the robot gets along with people in the home environment. A home security system composed of a mobile robot and an ultrasonic sensor network is proposed in [11]. The deployed sensor nodes provide the robot with location information for path planning. The design and implementation of an intelligent home security robot based on off-the-shelf components has been presented in [12]. The authors aim to implement a low cost and small home security robot that is suitable for patrolling tasks in narrow indoor environments. A home security system that includes an intelligent security robot and multiple remote interfaces has been presented in [13]. A user can get access to the host computer to get security information of the home or office rooms and control the mobile robot through Internet.

2 G. Song et al.: Automatic Docking System for Recharging Home Surveillance Robots 429 Fig. 2. A prototype of the surveillance robot and the docking station. Fig. 1. Home security system architecture based on the proposed surveillance robot and the docking station. In most of the home security applications, the mobile robots often have to work continuously for several days or even several weeks without human intervention. In addition to being robust in their hardware and software design, such robots must be capable of long-term autonomy with sustainable energy. Although some methods exist for energy conservation of mobile robots, a docking station remains widely used for recharging as most modern robots use batteries for their energy resources [14]. At the docking station, the mobile robot connects to the charger autonomously for recharging. There have been some successful attempts in implementing the docking capability on autonomous mobile robots. The first work on robot recharging has been done by W. Grey Walter [15]. He developed two mobile robots, Elsie and Elmer, which used light to find their way to a hutch where their power supplies were installed. More recently, Hada and Yuta presented a battery recharging system for long term activity of autonomous mobile robots [16]. They used some infrared sensors and a reflective tape on the floor as the target to position the robot for docking. Silverman developed a recharging system and a docking mechanism with two degrees of freedom to increase the tolerance for docking errors [17]. The docking mechanism has been integrated with a commercially available mobile robot. Vision and laser ranger-finder are used to find the docking station. R. C. Luo presented an intelligent recharging system for mobile security robots [18] [19]. An adaptive fusion method is used to detect and diagnose the power sensor status for recharging decision. Y.-C. Wu developed a robot docking station for automatic battery exchanging and charging [20]. A movable carrier allows ±9 degree misalignment between the autonomous robot and the docking station. This paper presents the design and implementation of a novel home surveillance robot with automatic docking and recharging capabilities. The proposed system is composed of a surveillance robot and a docking station. The surveillance robot is 15cm 15cm 9cm (L W H) with a triangular structure. The wheel-based mobile robot with a USB camera is specifically designed for home security usage. It communicates with the general wireless home router through WiFi. It communicates with the docking station through ZigBee and serves as a mobile wireless sensor network gateway. The docking station has a trapezoidal structure with an arc-shaped docking interface. The robot can return to the docking station for recharging operations when the on-board battery is too low. The rest of this paper is organized as follows. Section II introduces the overall architecture of the home security system based on the proposed surveillance robot and the docking station. The locomotion mechanism and the docking method are presented in Section III. The experimental results on the locomotion and docking performance of the prototype robot are given in Section IV. Concluding remarks are given in Section V. II. SYSTEM DESIGN The conceptual architecture of a home security system based on the proposed surveillance robot and the docking station is shown is Fig. 1. The surveillance robot can work in three modes according to user requests and task properties: patrolling mode, first responder mode and remote control mode. In the patrolling mode, the surveillance robot wanders around in the rooms or follows predefined routes autonomously. When security related information is acquired, it will be sent to the home server for further analysis. In the first responder mode, the surveillance robot is programmed to work in cooperation with other fixed monitoring devices. When one of those devices reports a security event to the surveillance robot, it will navigate to the target region to Fig. 3. CAD model of the proposed surveillance robot and the docking station.

3 430 IEEE Transactions on Consumer Electronics, Vol. 57, No. 2, May 2011 Fig. 4. Hardware components of the surveillance robot. perform on-site inspections. It overcomes obstacles on its routes by making a detour. In the remote control mode, the surveillance robot will be guided to the region of interest under control of a remote user. Users can access the home security system through various terminals such as PCs, PDAs and mobile phones. A prototype of the surveillance robot and the docking station is shown in Fig. 2. The CAD model that displays the detailed structures of the surveillance robot and the docking station is shown in Fig. 3. The docking interface of the robot is a semi cylinder with two charging electrodes on the front. The charging electrodes are installed on elastic supports so that the impact of the docking station in the docking process can be effectively buffered. The docking interface of the docking station has an arc shape that can guide the robot to finish electrical connections with the docking station. The proposed architecture of the docking interface can tolerate a robot position error up to 7cm and an angle error up to 60 degrees when the robot is performing the docking action. A. The Surveillance Robot The hardware architecture of the surveillance robot is shown in Fig. 3. The two DC motors provide power for rotating the two rear wheels. The servo is used for tilting the camera for a wide field of view. The voltage detection module is mainly used for real-time detection of the battery status. In the normal working status, when detecting that the battery voltage is lower than the preset charging voltage, the core board will command the robot to return to the docking station for recharging. In the charging status, when detecting that the battery voltage is higher than the preset working voltage, the core board will command the robot to start to work again. Fig. 6. Software architecture of the surveillance robot. The robot depends on two infrared sensors and two incremental encoders to perform obstacle avoidance, autonomous navigation and other locomotion tasks. The encoding board is a circular card with black and white barcodes on the back of the wheel. The encoder detection circuit reads and stores the voltage pulses caused by the rotating wheels. The moving distance of robot therefore can be calculated. The software architecture of the surveillance robot is shown in Fig. 6. The embedded program in the robot can implement basic locomotion behaviors and other high-level behaviors such as video transmission, cruising, docking and recharging. The robot communicates with the docking station through ZigBee and serves as a mobile wireless sensor network gateway. More docking stations can be added to the system by serving as additional wireless end nodes. Therefore it will be more flexible for docking and recharging the robot. The graphic user interface at the host computer communicates with the robot through WiFi. In the WiFi network, the robot changes its role to an end device. B. The Docking Station The hardware components of the docking station include a charging module, a current detection module, a wireless communication module and two IR sensors, as shown in Fig. 5. The current detection module uses a comparator to compare the input voltages to determine the presence or absence of Fig. 5. Hardware components of the docking station. Fig. 7. Outline of the docking area.

4 G. Song et al.: Automatic Docking System for Recharging Home Surveillance Robots 431 Fig. 8. Locomotion model of the surveillance robot. charging current. The wireless communication module exchanges data with the surveillance robot through the ZigBee network. The IR sensors are used for precise docking when the robot is in the docking area. The outline of the docking area is shown in Fig. 7. Each side of the docking station has an IR sensor to detect obstacles ahead. According to the outputs of the IR sensors, the relative position between the robot and the docking station can be determined. Therefore the robot will connect to the docking station automatically. During the docking process, the IR sensors on the robot will temporarily stop working to avoid interfering with the IR sensors on the docking station. Detailed docking method will be introduced in the following section. III. DOCKING METHOD If the surveillance robot wants to recharge by itself whenever the battery voltage is low, it must be able to navigate back to the docking area and connect with the docking station automatically. Some key techniques include self-localization, global and local path planning, docking and charging status detection, and fault-tolerant processing. Before reaching the docking area, the robot mainly depends on its own locomotion capabilities to work. In the final docking process, the robot and the docking station work cooperatively to complete the task. A. Locomotion Control The proposed surveillance robot is designed with a rear-wheel differential drive structure. Its locomotion model is shown in Fig. 8. We select a center point O that is located in the center of the rear axle for locomotion modeling. Using the Fig. 9. Grid map of the robot workspace. The shadowed grid cells represent the areas occupied by obstacles. The blue line indicates an example path from the grid coordinates (0, 0) to (7, 5). trajectory of the center point as the robot trajectory, we get the locomotion equations as follows: Vl Vr 1 V R( l r) (1) 2 2 Vl Vr R ( l r ) (2) D D where R is the radius of the robot wheels, V l and V r are the line speeds of the robot wheels, ω l and ω r are the angular velocities of the robot wheels, V and ω are the straight line speed and angular velocity of the robot in the horizontal plane respectively. The location of the robot is expressed as the following vector: x p y (3) The self-localization of the robot is based on the condition that its initial position is known. To simplify operation and improve accuracy, it is necessary to calculate the position coordinate information every t, which is set as the mobile robot operating interval. The incremental moving distance ( x, y, ) is: x scos( /2) (4) yssin( /2) (5) Fig. 10. The diagram of the final docking process when the robot is detected by one of the IR sensors on the docking station. (a) The robot is detected by one of the IR sensors. (b) The robot is detected by both sensors. (c) The robot turns 90 degrees to face the docking station. (d) The robot connects itself to the docking station.

5 432 IEEE Transactions on Consumer Electronics, Vol. 57, No. 2, May 2011 sr sl (6) b s r s s l (7) 2 where sl and sr are the moving distance of the left and right wheel respectively, b is the distance between the left wheel and right wheel. Therefore, we can get the updated position P ' : sr sl sr sl cos( ) x 2 2b s s s s (8) 2 2b sr sl b r l r l p' f(x,y, s r, s l) y sin( ) Since the workspace of the home surveillance robot can be known a prior, we can plan a global path for the robot to navigate back to the docking area. We use the grid method to build a 2D grid map of the robot workspace, as shown in Fig. 9. A grid cell with 35cm 35cm in size is big enough to fully accommodate the robot and the docking station. A shortest collision-free path can then be planned by the A* algorithm [21]. The path is composed of straight lines that follow either the X-axis or the Y-axis in order to simplify the calculation. B. Docking Process Since the localization errors during the navigation process can not be avoided, it is most unlikely that the robot will be exactly aligned with the docking station when the navigation process ends. But it is much easier for the robot to reach the docking area that is the grid cell containing the docking station. When the robot reaches the docking area and one of the IR sensors on the docking station detects it, the final docking process begins, as shown in Fig. 10. It should be noted that each time the robot reaches the docking area, it first has to adjust its direction to be parallel with the docking station. Therefore it will be much easier for the two IR sensors to detect the robot. If only one IR sensor detects the robot, the robot is not aligned with the docking station. But it can still finish the docking process with the help of the arc-shaped docking slideway. In order to increase the success rate of the docking process, our docking algorithm commands the robot to keep moving until the other IR sensor detects it. Then the robot stops and turns 90 degrees to reach the docking interface. The flowchart of the robot control in the docking process is shown in Fig. 11. The flowchart of the docking station control in the docking process is shown in Fig. 12. The robot first runs its wireless senor network gateway and initiates a ZigBee network. It waits for the wireless end node of the docking station to join the network. Once the end node joins the network, the gateway sends the docking-start Fig. 11. Flowchart of the surveillance robot control in the docking process. Fig. 12. Flowchart of the docking station control in the docking process.

6 G. Song et al.: Automatic Docking System for Recharging Home Surveillance Robots 433 (a) Fig. 13. Testbed setup for the automatic docking experiments. (a) The testbed. (b) The graphic user interface. (b) command. Then the end node will send back the IR sensor data of the docking station. The robot chooses its next-step actions according to the received IR sensor data types. If neither of the IR sensors detects the robot, it means that the robot fails to reach the docking area at the end of the navigation process. Under the circumstances, the robot starts to search the docking station by following a stepped line path. If the robot fails to find the docking station after two searching steps, it means that the cumulative localization error in the navigation process is too big. The robot must stop the automatic searching actions at once and send requests to the human operator for assistance. IV. EXPERIMENTAL RESULTS A testbed has been built in our laboratory for the automatic docking experiments. The testbed setup is shown in Fig. 13. A laptop computer runs the high-level control programs and provides the graphic user interface. Through the graphic user interface, a user can remote control the surveillance robot and watch the real-time video transmitted back by the surveillance robot. Both the laptop computer and the surveillance robot connect to a wireless local area network router through WiFi. The surveillance robot communicates with the docking station through ZigBee. Some experiments have been devised and performed to evaluate the locomotion performance and the automatic docking performance of the implemented surveillance robot. In the path planning of the surveillance robot, we use only two basic motion components, i.e. going-straight and in-situ turning. We have performed calibration experiments to increase the accuracy of self-localization and navigation. The self-localization results of the surveillance robot are shown in Fig. 14 and Fig. 15. The surveillance robot is programmed to move on a flat table-top surface at its full speed and perform only one basic motion at a time. The full straight moving speed is about 12cm/s and the full turning speed is about 22deg/s. As shown in Fig. 14, the surveillance robot is programmed to move straight along the x-axis from the starting point (0, 0) to several end points respectively. Ten Fig. 14. Calibration results in self-localization when the surveillance robot is commanded to go in a straight line. Fig. 15. Calibration results in self-localization when the surveillance robot performs in-situ turning.

7 434 IEEE Transactions on Consumer Electronics, Vol. 57, No. 2, May 2011 repeated trials have been done for every end point. The final results of real positions show that the positioning error increases faster in the y-direction. As shown in Fig. 15, when the surveillance robot performs in-situ turning, a steady deviation from the target position still exists. Although the self-localization accuracy of the surveillance robot seems good enough now, the cumulative errors will still hinder the robot from working normally in long-time patrolling tasks. We suggest that the planned path should have as many symmetric turns as possible. Therefore many positive and negative navigation errors will be effectively neutralized. The automatic docking experiments are divided into three types. The first type is the docking-only experiments in which the surveillance robot starts docking directly from within the TABLE I RESULTS OF THE AUTOMATIC DOCKING ATTEMPTS Docking Types Success Failure Docking Only IR1 IR Navigation Only Start from (0, 0) 10 0 Start from (0, 5) 10 0 Navigation and Docking Start from (0, 0) 9 1 Start from (2, 5) 8 2 Total 74 6 Fig. 17. Trajectories of the the surveillance robot during the process of navigating back to the docking area. docking area. The second type is the navigation-only experiments in which the surveillance robot navigates back to the docking area from the other side of the testbed. The third type is the navigation and docking experiments in which the surveillance robot performs complete automatic docking and recharging tasks from a remote starting point on the testbed. The success and failure rates of each type of docking attempts are shown in Table I. The docking test is repeated ten times for each type. IR1 and IR2 represent the outputs of the two infrared sensors on the docking station. The output value of one indicates that the surveillance robot is detected. The output value of zero indicates that the surveillance robot is not detected. The surveillance robot has reached the docking area in all of the Fig. 16. Video frames cut from the video captured during the process of navigating back to the docking area for recharging.

8 G. Song et al.: Automatic Docking System for Recharging Home Surveillance Robots 435 navigation-only attempts. The six failures are all caused by unsuccessful alignment in the final docking stage. The experimental results show that the prototype robot achieved a success rate of 90% after 60 different docking attempts. Nine key frames cut from the video captured in one trial of the navigation and docking experiments are shown in Fig. 16. The sequence of the frames in the figure is in accordance with the time stamps in the lower right corner of the frames. The surveillance robot navigates back to the docking area from frame (t=1s) to frame (t=22s). It performs the final docking actions in the rest of the frames. The trajectories of the surveillance robot during the process of navigating back to the docking area are shown in Fig. 17. The robot returns to the docking area by following the planned path. V. CONCLUSION We have presented the design and implementation of a surveillance robot with automatic docking and recharging capabilities for home security. A docking method based on the self-localization of the robot and the infrared detectors of the docking station is proposed. The robot can navigate back to the docking station for recharging operations when the on-board battery is too low. The prototype robot achieved a success rate of 90% after 60 different docking attempts. Future work will focus on improving the current prototype robot to enable more functions. We plan to address several technical challenges such as visual navigation, adding more docking stations, and the automatic battery replacement mechanism. REFERENCES [1] G. Song, Z. Wei, W. Zhang and A. Song, A hybrid sensor network system for home monitoring applications, IEEE Trans Consum Electron, Vol. 53, No. 4, pp , [2] G. Song, Y. Zhou, Z. Wei and A. Song, A smart node architecture for adding mobility to wireless sensor networks, Sens Actuators A Phys, vol. 147, no. 1, pp , [3] G. Song, K. Yin, Y. Zhou and X. Cheng, A Surveillance Robot with Hopping Capabilities for Home Security, IEEE Trans Consum Electron, Vol. 55, No. 4, pp , [4] C. D. Nugent, D. D. Finlay, P. Fiorini, Y. Tsumaki and E. Prassler, Home automation as a means of independent living, IEEE Trans. Autom. Sci. Eng., Vol. 5, No. 1, pp. 1-8, Jan [5] Yoo Oh, Jae Yoon, Ji Park, Mina Kim and Hong Kim, A name recognition based call-and-come service for home robots, IEEE Trans Consum Electron, Vol. 54, No. 2, pp , [6] Y. W. Bai, L. S. Shen and Z. H. Li, Design and implementation of an embedded home surveillance system by use of multiple ultrasonic sensors, IEEE Trans Consum Electron, Vol. 56, No. 1, pp , [7] W. Lao, J. Han and Peter H.N. de With, Automatic video-based human motion analyzer for consumer surveillance system, IEEE Trans Consum Electron, Vol. 55, No. 2, pp , [8] Y. Zhao and Z. Ye, A low cost GSM_GPRS based wireless home security system, IEEE Trans Consum Electron, Vol. 54, No. 2, pp , [9] I.K. Hwang and J.W. Baek, Wireless access monitoring and control system based on digital door lock, IEEE Trans Consum Electron, Vol. 53, No. 4, pp , [10] C. Chang, K. Chen, H. Lin, C. Wang and J. Jean, Development of a patrol robot for home security with network assisted interactions, in SICE Annual Conference 2007, Kagawa University, Japan, 2007, pp [11] Y. Kim, H. Kim, S. Lee and K. Lee, Ubiquitous home security robot based on sensor network, in IEEE/WIC/ACM Int. Conf. on Intelligent Agent Technology, Hong Kong, China, 2006, pp [12] R.C. Luo, P.K. Wang, Y.F. Tseng and T.Y. Lin, Navigation and mobile security system of home security robot, in IEEE Int. Conf. on Systems, Man and Cybernetics, Taipei, Taiwan, 2006, pp [13] R.C., Luo, T.Y. Hsu, T.Y. Lin and K.L. Su, The development of intelligent home security robot, in IEEE Int. Conf. on Mechatronics, Taipei, Taiwan, 2005, pp [14] Se-gon Roh, J. H. Park, Y. H. Lee, Y. K. Song, K. W. Yang, M. Choi, H.-S. Kim, H. Lee, and H. R. Choi, Flexible docking mechanism with error-compensation capability for auto recharging system of mobile robot, Int. Journal of Control, Automation, and Systems, vol. 6, no. 5, pp , [15] W. Grey Walter, The Living Brain, W.W. Norton, New York, [16] Y. Hada and S. Yuta, Robust navigation and battery re-charging system for long term activity of autonomous mobile robot, Proc. Int. Conf. Advance Robotics, pp , [17] M. C. Silverman, D. Nies, B. Jung and G. S. Sukhatme, Staying alive: a docking station for autonomous robot recharging, in IEEE Int. Conf. on Robotics and Automation, Washington D.C., 2002, pp [18] R.C. Luo, P.K. Wang, Y.F. Tseng and T.Y. Lin, Automatic docking and recharging system for autonomous security robot, in IEEE Int. Conf. on Intelligent Robots and Systems, Edmonton, Canada, 2005, pp [19] Ren C. Luo and Kuo L. Su, Multilevel multisensor-based intelligent recharging system for mobile robot, IEEE Trans Industrial Electronics, Vol. 55, No. 1, pp , [20] Yi-Cheng Wu, Ming-Chang Teng and Yi-Jeng Tsai, Robot docking station for automatic battery exchanging and charging, in IEEE Int. Conf. on Robotics and Biomimetics, Bangkok, Thailand, 2009, pp [21] C. Ismail and S. Lan, Adaptations of the A* algorithm for the computation of fastest paths in deterministic discrete-time dynamic networks, IEEE Trans Intelligent Transportation Systems, Vol. 3, No. 1, pp , BIOGRAPHIES Guangming Song (S 04-M 05) was born in Yichun city, China, on December 14, He received the B.S. and M.S. degrees in mechanical engineering from Hefei University of Technology, Hefei, China, in 1997 and 2001, respectively. He received the Ph.D. degree in control science and engineering from the University of Science and Technology of China, Hefei, China, in From 2004 to 2006, He was a Research Fellow with the Robotic Sensor and Control Laboratory, Southeast University, China. Since 2006, he has been with the School of Instrument Science and Engineering, Southeast University, China. He is currently a Professor with the School of Instrument Science and Engineering, Southeast University, China. His current research interests include wireless sensor networks, distributed measurement and control, and networked robots. Dr. Song is a member of Sensor Network Technical Committee, China Computer Federation. Hui Wang received the B.S. degree in measuring and control technology and instrumentations from China University of Mining and Technology, in He is currently an M.S. candidate in measurement technology and instruments, Southeast University, Nanjing, China. His research interests include wireless sensor networks and embedded systems. Jun Zhang received the B.S. degree in measurement & control technology and instrument from Nanjing University of Science and Technology, Nanjing, China, in He is currently a PhD candidate in measurement technology and instruments, Southeast University, Nanjing, China. His research interests include wireless sensor networks, distributed robotics and bio-inspired robotics. Tianhua Meng received the B.S. degree in energy and power engineering from Yangzhou University, Yangzhou, China, in She is currently an M.S. candidate in measurement technology and instruments, Southeast University, Nanjing, China. Her research interests include wireless sensor networks and embedded systems.