Mobile Sensor Networks based on Autonomous Platforms for Homeland Security

Transcription

1 Mobile Sensor Networks based on Autonomous Platforms for Homeland Security A. Buonanno (+), M. D Urso (+),G.Prisco (+), M.Felaco (+,*) (+)SELEX Sistemi Integrati S.p.A. Via Circumvallazione Esterna di Napoli, zona ASI, I Giugliano, Napoli, Italy {anbuonanno, mdurso }@selex-si.com E. F. Meliadò (*), M. Mattei (*), F. Palmieri (*), D. Ciuonzo (*) (*)Seconda Università degli Studi di Napoli Via Roma 29, 81031, Aversa, Caserta Abstract The development of intelligent surveillance systems is an active research area of increasing interest. In recent years, autonomous or semi-autonomous mobile robots have been adopted as useful means to reduce fixed installations and number of devices needed for surveillance of a given area. In this context SELEX Sistemi Integrati is investigating the possibility to use robots-sensors systems to improve the monitoring of large and populated indoor areas. In particular, the joint use of the Swarm Logic and heterogeneous sensors, some installed at strategic points of the infrastructure and other installed on mobile robots, allows the creation of a very dynamic network of cooperating sensors, that is able to ensure a high level of protection and a fast reaction to threats. In this paper an integrated intelligent system based on swarm logic to improve monitoring performance of large critical infrastructures such as airport terminals, warehouses, railway stations, production facilities is presented. The adopted system architecture, consisting of two hierarchical levels, is introduced and discussed. In each of these levels novel aspects, developed by the team, are present. T I. INTRODUCTION AND MOTIVATIONS HE terrorism events of the last decade shifted the attention from the boundaries to the heart of nations, focusing on all those infrastructures that are central for economic, political, cultural or religious interests. These circumstances push toward the development of new security systems, more sophisticated and effective but also cheaper, due to the increasing number of critical infrastructures to be kept under control. In indoor environments such as airports, warehouses, production plants, etc., the need for automated surveillance systems has stimulated the development of intelligent systems based on mobile sensors, possibly mounted on autonomous or semi-autonomous robots. Differently from traditional nonmobile surveillance devices, those based on mobile robots are still in their initial stage of development, and many issues are under investigations [1-2]. Robots expand significantly the capabilities of surveillance systems with their active role in the environment sensing, consisting of the interaction with both humans (to perform adaptive tasks) and other robots (to perform coordinated tasks). The use of mobile robots allows to obtain an easily reconfigurable solution able to suit the characteristics of different operating scenarios, as: different structural constraints; different human behaviours; different threats to detect; different reactions; different environments. Furthermore, the use of mobile platforms as active mobile sensors needs for a complex system with a modular architecture, flexible and open to integration with different sensors and functionalities. To achieve this goal, SELEX Sistemi Integrati is developing a new easily integrable platform into the existing security systems. Particular attention will be devoted to the development of mobile robotic sensor networks for monitoring large and densely populated areas and for increasing the capability of human operators working in critical situations like anti-fire, antiterrorism, police missions and on the battlefield (Figure 1). Different hardware and software components are under evaluation (Figure 2), while parallel investigations on suitable algorithms and methodologies are being carried out. Figure 1 New semi-autonomous robotic systems. One possible architecture for monitoring of large and populated area, such as airport terminal, is the subject of the first part of the paper. In this part, the algorithms for mobile robot guidance and navigation in a partially known 2012 CNIT Tyrrhenian Workshop

2 environment, already introduced in [3], are briefly recalled. In the second part, the decision fusion algorithms to track risky targets in a dynamic environment through distributed wireless sensors are introduced. The proposed algorithms are the basis for an optimal allocation of the mobile sensing resources to minimize the probability of false alarms and missed detections. Figure 2 Some Robotic platforms evaluated by the team for the development of the surveillance system. II. THE SYSTEM ARCHITECTURE The possibility to monitor wide and densely populated areas such as airport terminals, train and naval stations, warehouses, production plants, sport stadiums, etc., requires the concurrence of different technologies and a high level of integration between them. Modern systems can take advantage of networks of sensors installed both on infrastructures and mobile devices as robots or human operators. In particular, data fusion systems can maximize the effective information, and increase the situational awareness of the operators to take rapid and efficient decisions. SELEX-SI is studying new technologies and methodologies based on mobile robot platforms to introduce into a novel surveillance system including the command and control systems architectures already developed. The proposed architecture of such a system is composed of two hierarchic layers (Figure 3). The higher level layer, based on a net-centric topology, is the layer closest to the security human operators and is devoted to the following tasks. Data fusion: extraction of synthetic information for risk evaluation; Decision making and interface with human operators to increase situational awareness; Classification of targets and tracking of the risky ones. The lower layer (see also Figure 4), distributed among fixed sensors and the robot swarm, is involved in very important tasks: Sensing resource optimization: optimization of mobile devices position to minimize false or missed alarms in a cooperative approach; Tasks Management: assigning and managing dynamically tasks between different platform according to priority of the task and platform capability; Detection of system faults: detection and management of the faults reassigned appropriately the tasks. Furthermore, each robot of the swarm must be able to perform the following tasks: Map Building; Path planning; Obstacle avoidance; Self localization; Motion Control; Detection of robot faults. We note that this is a crucial part of the system. Indeed, to provide the previous capability to each robot of the swarm is essential to obtain the proposed goal, and this is an hard problem due to the complexity of considered scenarios. The presence of different tasks to be accomplished by the robotic agents can produce conflicts in decision making. Resolution of conflicts can be obtained by using competitive or cooperative approaches. A possibility under analysis is the adoption of the Real Time Swarm Intelligence Platform (RT- SIP) for resolution of conflicts. Figure 3 The hierarchic layers of the proposed architecture for the autonomous surveillance system.

3 RT-SIP is an innovative multi-agent middleware which provides cooperation and coordination services based upon the Swarm Intelligence model. Thanks to the adoption of a datacentric, server-less architecture, based on the Real Time Data Distribution Services (RT-DDS) standard, the RT-SIP is interoperable and can be included in the network-centric systems. The RT-SIP complements the adaptive control of Swarm Intelligence to the scalability, dependability and predictability of the RT-DDS, thus resulting into a valuable component of systems that needs to operate in complex and dynamic scenarios. environmental information. According to our definition, a map is represented by a set of geometric primitives, suitable to describe most of structured indoor environments. In the paper we propose the use of line segments as basic geometrical primitive. In Figure 6 the flow-chart of the proposed algorithm is showed. Start LS acquisition Local segment based map extraction Estimation of the laser pose and its orientation Fusion of the Local map and Global map last scan? N Y Stop Figure 4 Task of the lower layer. Figure 6 - Flow-chart of the proposed algorithm III. THE SYSTEM ARCHITECTURE In this section we briefly re-introduce the key points of the algorithms for mobile robot guidance and navigation in a partially known environment. A. Map Building via Laser Rangefinder The purpose of this algorithm is to obtain a Global Map that represents the whole environment, exploiting Local Maps that are extracted from the laser scanner measures [4-7] (Figure 5). More detail are in [3]. B. Robots Self Localization A reliable estimation of the robot position is another key point in the implementation of the proposed hybrid sensing system. The presence of a dynamic environment with a lot of fixed and moving obstacles makes this problem very complex. One of the most common methods adopted to perform this fusion exploits the Extended Kalman Filter (EKF, [8]). Unscented Kalman Filter (UKF [3]) has been demonstrated to work much better than EKF when system dynamics and output function are nonlinear. We have implemented a dynamic variant of the UKF that has shown to be robust to occasional changes of the environment such as opened/closed doors or walking people [3]. Figure 5 Superimposition of the obtained Global map with an available building map More specifically, a local map is built from a scanner acquisition and it is used to update the global map, so that in each step the global map contains all the acquired C. Path Planning In the proposed approach, based on potential fields [9-11], a fluid mechanics similitude to generate the control vector valued function is used for both obtaining the nominal control sequence to reach the target positions in presence of known obstacles, and updating the control vector map in presence of unknown obstacles (Figure 7). Control actions are computed by solving in the geometrical domain of interest, with Finite Elements Methods, the following problem: 2 φ = 0

4 subject to the Dirichlet s boundary conditions: ( ) = 1 obstacle φ. φ ( arg ) = 0 t The control action map is directed along the gradient of the solution v = λ φ. et leads to the so called Parallel Range Limited Marginalization (PRLM) algorithm. These algorithms have already shown good performances when they are used with fixed sensors, we will extend their utilization in the case of mobile sensors (Figure 8). The proposed algorithms have the further advantage that they allow to decentralize part of the fusion control decisions to fusion devices covering different zones of the monitored area. Figure 1. Figure 7 Example of vector field of control actions driving mobile robots to targets without obstacles Figure 2. Figure 8 :Block diagram of the considered system. IV. RISKY TARGETS DETECTION ALGORITHMS An important aspect for the high level automation system is the detection of potentially risky targets moving in the monitored area. The possibility to increase the reliability of this detection by means of mobile sensing devices has to be evaluated starting from the analysis of possible risk classification algorithms. Assume that the high level system has to track and classify N different targets moving into the area of interest. Two sub-optimal decision fusion algorithms, namely the Range Limited Marginalization (RLM) and its parallelized form (PRLM) [12] are under investigation, which are based on the separation of the fusion process of each target. An optimal decision fusion algorithm for multi-target classification requires [13] to calculate the Bayes recursive update, this optimal approach cannot be implemented with reasonable time responses. We proposed two sub-optimal approach to obtain a reduced complexity, first of all we update the classification of each target separately from the others, this clearly leads to sub-optimal performance related to the error probability, but that can be implemented in practice. This proposed algorithm, named Range Limited Marginalization (RLM), starts from the exact computation formula of the N marginal posteriors through the Bayes update obtained by the conditioning chain rule [14], due to space-time independency of decisions which are transmitted from the sensor. After, identifying a subset of targets which are inside the maximum detection range of the sensor at each instant of time, we approximate that each conditional distribution for each sensor can be implemented in parallel with each other. This V. CONCLUSIONS SELEX Sistemi Integrati S.p.A is studying new technologies and methodologies to be implemented into a novel surveillance system based on swarm logic that will be integrated in the higher level command and control systems architectures already developed by the company or under development such as in that of the Airport Turn-Key system. The proposed architecture makes use of wireless sensor networks and sensing devices mounted on mobile autonomous or semi-autonomous, terrestrial or flying robots. Modern techniques and technologies will be integrated in such a system. An overview of the proposed hierarchic architecture and of some investigations carried out in the field of mobile robot guidance, navigation, and localization, and in the field of indoor map building is provided. Future work will be devoted to investigate the possibility of using alternative techniques for path generation of robot, and particle filters localization algorithms. Moreover, all the developed algorithms will be tested and optimized in terms of computational cost proving the effectiveness in real operative scenarios. To improve the efficiency of the surveillance system and to optimize the task assignment among different robots, also solutions diverse of the use of RT-SIP and logic swarm will be studied. This last problem finds several points of contact with the decision fusion algorithms proposed to classify risky targets in a multi sensor dynamic environment. All the proposed methodologies are under assessment by using two experimental set-up, one located in Centro Eccellenza Grandi Sistemi of SELEX Sistemi Integrati and the other located in the Test Bed room of the SELEX Sistemi integrati

5 in Giugliano in Campania, with the support of Mechanical Engineering Department of the Second University of Naples and of Sistemi Software Integrati. REFERENCES [1] H. Everett, Robotic security systems, IEEE Instruments and Measurements Magazine, vol. 6, no. 4, pp , December [2] T. Duckett, G. Cielniak, H. Andreasson, L. Jun, A. Lilienthal, P. Biber, T. Martínez, Robotic Security Guard - Autonomous Surveillance and Remote Perception, Proceedings of IEEE International Workshop on Safety, Security, and Rescue Robotics, Bonn, Germany, May [3] POLARIS Innovation Journal N 8 [4] F. Amigoni, S. Gasparini, M. Gini, Building Segment-Based Maps Without Pose Information, Proceedings of the IEEE, vol.94, no.7, pp , July [5] Zezhong Xu, Jilin Liu, Zhiyu Xiang, Han Li, Map building for indoor environment with laser range scanner, Proceedings of IEEE 5th Int. Conf. on Intelligent Transportation Systems, pp , [6] L. Zalud, L. Kopecny, T. Neuzil, Laser proximity scanner correlation based method for cooperative localization and map building, Proceedings of th Int. Workshop on Advanced Motion Control, pp , [7] S. F. Hernandez-Alamilla, E. F. Morales, Global Localization of Mobile Robots for Indoor Environments Using Natural Landmarks, Proceedings of 2006 IEEE Conf. on Robotics, Automation and Mechatronis, pp.1-6, Dec [8] Crowley, J.L., (1989). World Modeling and Position Estimation for a Mobile Robot Using Ultrasonic Ranging. IEEE Int. Conf. on Robotics and Automation (ICRA), Scottsdale, AZ, USA, 1989 [9] G. Antonelli, and S. Chiaverini, Kinematic control of platoons of autonomous vehicles, IEEE Trans. Robotics, Vol. 22, No.6, pp , [10] Hwand, Y., K., and Ahuja, N., A Potential Field Approach to Path Planning, IEEE Trans. Robot. Automat., Vol. 6, 1992, pp [11] Cen, Y., Wang, L., and Zhang, H., Real-time Obstacle Avoidance Strategy for Mobile Robot Based On Improved Coordinating Potential Field with Genetic Algorithm, Proceedings of 16th IEEE Intern. Conf. on Control Appl., Singapore, 1-3 October [12] D. Ciuonzo, A. Buonanno, M. D Urso and A.N. Palmieri, Distributed Classification of Multiple Moving Targets with Binary Wireless Sensor Networks. International Conference on Information Fusion, Chicago, USA, 5-8 July [13] J.H. Kotecha, V. Ramachandran, and A.M. Sayeed. Distributed multitarget classification in wireless sensor networks. IEEE Journal on Selected Areas in Communications, 23(4): , april [14] T.M. Cover and J.A. Thomas. Elements of Information Theory (Wiley Series in Telecommunications and Signal Processing). Wiley- Interscience, 2006.