You are on page 1of 3

Net-Centric, Layered-Sensing Issues in Distributed Tracking and Identification Performance Evaluation

Erik Blasch
Air Force Research Lab WPAFB, OH
erik.blasch@wpafb.af.mil

Chun Yang
Sigtem Tech., Inc. San Mateo, CA
chunyang@sigtem.com

Ivan Kadar
Interlink Sys. Sciences., Inc. Lake Success, NY
ikadar@SystemsSciences.com

Genshe Chen
DCM Research Resources.

Li Bai
Temple Univ. Philadelphia, PA
lbai@temple.edu

Germantown, MD
gchen@dcmresearchresources.com

Abstract Real-world instantiation of distributed target tracking (DTT) systems will require analysis of the scenario (i.e. operational conditions (OC) of the sensor, target, and environment), requirements on data support (e.g. communication channels and a priori data bases), and acknowledged constraints of the system applicability. The overview presents some of the possible bounds to aid in the testing, validation, and usefulness of future distributed tracking solutions. We make three claims in the paper. (1) Scenarios drive the user efficacy of solutions, thus a distributed system evaluation should develop metrics over OCs which highlight the degradation of solutions to afford users (or sensor manager) how and when to best apply the DTT solution. (2) DTT should incorporate communication, distributed-database, and support information to aid in driving information requirements (or requests) for effective targeting (identification who, what, when, where). Finally, (3) DTT can gather information from either localized sensors or a global approach. The global approach is a layered sensing concept, when applied hierarchically, can incorporate various information types for analysis. Keywords: Fusion, Sensor Management, Multiresolution, Tracking, ATR, Performance Evaluation

scenarios, and (3) user-machine information sharing and collaboration. [15] 1.1 Layered Sensing DTT Layered sensing can be multiple sensors differing by spatial/temporal/spectral/polarimetric locations or by measurements (i.e. space, cyber, targeting) supporting full spectrum dominance [16]. Utilizing a horizontal (e.g. cyber support) or vertical fusion (as shown in Figure 1), coordinating the diverse set of sensors, platforms, and users requires accurate timing and hand-off, effective multiresolution scaling and cueing, and manageable frequency and communication bandwidths/capacities [17].

Introduction
Figure 1. Layered Sensing 1.2 Net-Centric DTT

Distributed target tracking (DTT) solutions have been produced for exemplar cases [1, 2, 3, 4, 5]. Many of the approaches to DTT result from a decentralized/distributed solution (i.e. Kalman Filter [6] or Information Filter [7]) to a known centralized approach. [8, 9]. As real-world systems are instantiated under the military philosophy [10, 11, 12] - Centralized Command, Decentralized Control, there exist expectations to successful implementation for machines that are not imparted on users. As a case in point, it is assumed all actors/sensors have the same information. However, not all sensors will have the same global or local information, needs, or protocols. DTT typically implies the coordination of many sensors [13] to track a series of targets in a complex environment. To tackle the complexities a distributed, layered, and netcentric information approach is needed that simultaneously manages the sensors and delivers situational awareness.[14] To move towards the goal of a net-centric, layered environment, there is a need for (1) robust decisions, (2) command and control over different

Net-centric approaches [6] consist of coordination between ground assets (e.g. cameras) and surveillance platforms (e.g. UAVs), thus dictating that there will be a set of nodes (nets), communicating with other nodes for distributed analysis. Most net-centric systems designs pay tribute to data and information fusion, but concentrate on the software issues for Global Information Grid (GIG) [7] and the Joint Battlespace Infosphere (JBI). Analogy to business enterprise issues of distributing logistics tracking, protocols, a new way of thinking [8] is to utilize the right tracking approach for differing scenarios, missions, and scales. Examples of tenets of net-centric solutions include (1) near-real time web enabled collaboration of task, post, process and use, (2) reliable and secure content management, (3) sense-making through decision superiority, and (4) network architectures supporting distributed solutions. [9]

"Smart Weapons for UAVs" Article / Defense Update 1/2007)

Figure 3. Net-Centric Enabled Operators 2.2 Figure 2. Net-Centric Tracking [from 10 Kaplan] A NET-TRACK [11] solution is needed to enable simultaneous tracking and identification of targets in the environment where ID information helps to validate and hand-off tracks cued from detections and tracklets. Important information includes the effective communication and coordination between nets to pass useful complementary information to preserve bandwidth while limiting the passing of and redundant information. Issues associated with tracking include security, communications, and message passing such as techniques of stenography [12], time-division multiplexing, and cognitive radio as well MIMO electro-optical and radar for waveform diversity. Interoperatibility will be a key issue in wireless communications. Net-centric solutions are more about distributed Command and Control, including the ability to conduct a diversity of missions. Robust Set of Net-Centric DTT Metrics

2
2.1

Distributed Tracking Needs


DTT Evaluation over Scenarios

Managing expectations of system capabilities and constraints is necessary to bound information requirements and information performance. Since future DTT systems will enable collaborative coordination [13], it is essentially to provide scenarios that include a diversity of target types (e.g. cars and people), sensor platforms (e.g. air, ground, space, cyber), environmental conditions (e.g. urban, remote), as well as users (e.g. sensor operators and commanders). Figure 3 shows a case where multiple co-located operators may have diverse missions enabled by net-centric coordination. In addition to known/controllable assets in a scenario design, it is important to consider the complexity of the non-cooperative players. New methods of distributed tracking would include (1) game-theoretic approaches [14] to track active and passive participants [15], (2) sensor management utilizing databases of social-cultural behaviors, and (3) interoperable constraints of security, protocols, and communication constraints [17].

While DTT has focused on decentralized/distributed architectures that support Info Filter/Kalman Filter state and covariance message passing; there are a host of metrics to consider when implementing net-centric approaches such as timeliness, accuracy, confidence, throughput, utilization cost, and security. Users desire quality of service (QoS) [16] which requires information quality [29, 30] and interoperable, but robust performance. Providing both generic QoS and fusion specific QoS [30] capabilities would encourage user trust, reliance, and interaction with the DTT system. The net-centric panels [15] postulate the benefits of designs with enhanced situational awareness, coordinated sensor management, and user understanding; however key tenets of these benefits are robust decision making through timely and actionable sense-making. To deliver on these goals, DTT designs should develop approaches on top of the net-centric protocols, utilize information databases, and planned architectures. These networks have additional constraints of bandwidth on message passing, security for reliable transmission, utilization for distributed users, and data confidence for trust. Thus, DTT metrics should explore a robust set of metrics that determine sensitivity performance for various scenarios to manage user expectations. 2.3 Performance Models Enabling DTT Processing Performance models drive the analysis of distributed processing through content and context understanding (e.g. environmental models of terrain and road networks to provide a common framework), behavior analysis for hand-off (e.g. target models for predictions of where tracks will likely be in other systems), and scaling and cueing (e.g. sensor models to handoff messages between satellite and acoustic ground sensors). Net-centric DTT requires an understanding that different track information is desired over logistics, targets, and engagements. Figure 4 highlights the models needed for effective DTT (which demonstrates a reliance on many support structures to enable DTT).

4
[1]

References
C-Y Chong, S. Mori, & K-C Chang, Distributed MultisensorMultitarget Tracking, Ch 8 in Multitarget-Multisensor Tracking Adv. Appl. Y. Bar-Shalom (Ed) Artech House, 1990. J. J. Leonard and H. Durrant-Whyte, Directed Sonar Sensing for Mobile Robot Navigation, Kluwer, 1992. R. R. Brooks and S. S. Iyengar, Distributed Dynamic Sensor Fusion Ch. 14 in Multisensor Fusion : Fundamentals and Applications, Prentice Hall 1998. C-Y. Chong; F. Zhao, S. Mori, and S. Kumar, Distributed tracking in wireless ad hoc sensor networks, Fusion03, 2003. S. Coraluppi, and C. Carthel Distributed tracking in multistage sonar, IEEE AES, Vol 41, 3, 2005. C. Brown, H. Durrant-Whyte, J. Leonard, et al, Distributed Data Fusion Using Kalman Filtering: A Robotics Application, Ch 7 in Data Fusion in Robotics and Machine Intelligence, M. Abidi and R. Gonzales (Eds.), Academic Press, Inc. 1992. K. C. Chang, T. Zhi and R. K. Saha, Performance Evaluation of Track Fusion with Information Matrix Filter, IEEE Trans. AES, Vol. 38, No. 2, April 2002. Y. Bar-Shalom & X. Li, Multitarget-Multisensor Tracking: Principles and Techniques, YBS, New York, 1995. S. Blackman and R. Popoli, Design and Analysis of Modern Tracking Systems, Artech House Publisher, Boston, 1999. Sun Tzu. The Art of War. Trans. by S. B. Griffith. Oxford University Press, 1971. C. V. Clausewitz, On War, trans. and ed. M. Howard and P. Paret, (Princeton, NJ: Princeton University Press, 1984). M. Van Creveld, Command in War, Harvard Univ. Press, 1985. C-Y. Chong & S. Kumar, Sensor networks: evolution, opportunities, & challenges, Proc. IEEE, Vol 91, 8, 2003. D. Hall, Challenges in Data Fusion: Dirty Secrets, Current State of Technology & Res. Roadmap, 2005. Available at : www.csci.psu.edu/seminars/springnotes/DataFusion2005.pdf Defense Science Board 2006 Study, Information Management for Net-Centric Operations, Vol II, USD for Acquisitions, April 2007. E. Blasch and P. Hanselman, "Information Fusion for Information Superiority," IEEE NAECON, 2000. I. Kadar, Distributed Multisensor Fusion with Network Connection Management, Proc. SPIE 5809, April 2005. S. Biswas, S. Gupta, F. Yu, and T. Wu, Collaborative Multi-target Tracking using Networked Micro-Robotic Vehicles, Proc. SPIE Vol. 6578, 2007. ncow.nps.edu/wiki/index.php/Global_Information_Grid. D. S. Onley, Net-Centric Goal: a different military, http://www.gcn.com/print/22_32/24048-1.html OSD Office of Horizontal Fusion. J. Kaplan, Enterprise-Wide, Net-Centric Systems-of-System Eng.,www.ndu.edu/ctnsp/defense_tech_papers.htm C. Yang, E. Blasch, W. Garber, and R. Mitchell, A Net Track Solution to Pose Angular Tracking of Maneuvering Targets in Clutter, 2007 IEEE Conf. Signals, Systems and Computers, 2007. L Bai, S. Biswas, and E. Blasch, An estimation approach to extract multimedia information in distributed steganographic images, Fusion2007, 2007. G. Seetharaman, M. Talbert, and E. Blasch, Architectural elements of integrated micro-sensors for distributed sensor networks, IEEE ASAP Conf, 2007. D. Shen, et al. Adaptive Markov Game Theoretic Data Fusion Approach for Cyber Network Defense, IEEE MILCOM, 2007. M. Wei, et. al, A Decentralized Approach to Pursuer-Evader Games with Multiple Superior Evaders in Noisy Environments, IEEE Aerospace Conf., 2007. E. Blasch, M. Pribilski, B. Daughtery, B. Roscoe, et al , Fusion Metrics for Dynamic Situation Analysis, Proc SPIE 5429, 2004. M. Johnson and K.C. Chang, Quality of Information for Data Fusion in Net Centric Publish and Subscribe Architectures, Fusion05, 2005. I. Kadar, Research Challenges in Network & Service Management for Distributed Net-Centric Fusion, Proc. SPIE 6235, April 2006.

[2] [3]

[4] [5]

Figure 4. DTT Modeling Support To determine the QoS robust metrics of performance, DTT systems should be tested over the sensitivity of performance relative to the fidelity of models assumed in the solutions. While Figure 4 highlights the needs for a DTT system, there is no mention of the supporting hardware and software constraints that are the focus of net-centric publications. DTT will be applied over different sensors, missions, and users; all which might have different models. The performance for one net might not be the system level performance of the system (which is what users will evaluate). Thus, different DTT solutions should be tested over different scenarios and model sensitivities for robust performance. The intersection would be robust control for model parameter variation, robust forecasting for model predictions, and robust decision making for user acceptance.

[6]

[7]

[8] [9] [10] [11] [12] [13] [14]

[15] [16]

Summary

[17] [18]

Net-centric, layered sensing is the current enabler of global universal situational awareness (USA). DTT can provide location and identification of targets of interest, but is subject to the sensor, target, and environment models residing in different machine nodes and information users. To deliver robust system performance, care must be taken to acknowledge the hardware and software constraints such as communication and interoperability. These constraints would provide a better perspective on fusion specific Quality of Service performance metrics. Characterizing the performance models and measures of merit (e.g. timeliness), DTT systems should be tested over a variety of scenarios and missions to fully deliver on a net-centric transformation in which varying numbers and types of targets, spatial coverage, and message and detection frequencies are demanded of DTT designs. Finally, as much of the netcentric literature is based on the software architectures, it is imperative that DTT mathematical solutions intersect with software needs (e.g. communication) and hardware constraints.

[19] [20] [21] [22] [23]

[24]

[25]

[26] [27]

[28] [29]

[30]

You might also like