Object Detection, Tracking, and Identification in a Congested Environment Using Artificial Intelligence (AI) Enabled Algorithms

OUSD (R&E) MODERNIZATION PRIORITY: Artificial Intelligence / Machine Learning; Network Command, Control and Communications TECHNOLOGY AREA(S): Sensors OBJECTIVE: Develop and validate AI-enabled algorithms and associated software capable of detecting, tracking, and identifying objects in a congested environment using data streams from radio frequency (RF) (e.g., passive, bi-static, synthetic aperture radar (SAR)) detection systems). DESCRIPTION: This topic seeks to develop AI-enabled algorithms and associated software capable of using data streams and/or data collected by sensors to enable detection, tracking, and identification of targets in a congested environment. Software would be expected to determine position and velocity and track objects in the field of view (FOV) or field of regard (FOR) of the sensor(s). Data streams of interest are those other than the traditional RF radar sources. Applicable data streams could include commercially available data streams for example. Ideally, the technology (AI-enabled algorithms and software) would be capable of establishing a fingerprint or signature for individual objects in a given environment through training or modeling in a controlled setting, i.e. in an area around an airport or a navigable waterway where cooperative objects are readily available and can be identified and tracked using online resources (e.g. https://flightaware.com, https://www.adsbexchange.com, and https://www.marinetraffic.com). For missile defense applications, which can include air, sea, and space security around valued assets as well as defense against threats, the technology would require rapid adaptability to diverse environments and extrapolation of data available on cooperative, non-cooperative, or deliberately deceptive targets. Solutions should apply to sensors using RF data streams, such as passive, bi-static, or SAR detection systems. PHASE I: Describe architecture and concept of operations applicable to missile defense applications and missions. Develop initial AI-enabled algorithms and describe their ability to distinguish between similar objects and to track objects of interest. PHASE II: Develop prototype AI-enabled algorithms and associated software. Demonstrate ability of the algorithms to detect, track, and identify objects in a congested environment using available data streams, such as from commercial airport and seaport websites. PHASE III DUAL USE APPLICATIONS: Implement the software into a missile defense relevant sensor system to demonstrate effectiveness. Sensors may be ground, sea, or space-based to detect, track, and identify threat in a congested environment. Additionally, post intercept assessment would be applicable to space-based sensors. Other civilian and commercial uses should be assessed. REFERENCES: P. Lang, X. Fu, M. Martorella, et.al, Comprehensive Survey of Machine Learning Applied to Radar Signal Processing, arXiv:2009.13702v1 [eess.SP]. W. M. Lees , A. Wunderlich , P. J. Jeavons, et.al, Deep Learning Classification of 3.5-GHz Band Spectrograms With Applications to Spectrum Sensing, IEEE Transactions on Cognitive Communications and Networking, Vol. 5, No. 2, June 2019. B. Yonel , E. Mason , and B. Yazici, Deep Learning for Passive Synthetic Aperture Radar, IEEE Journal of Selected Topics in Signal Processing, Vol. 12, No. 1, Feb. 2018. S. Mahfouz, F. Mourad-Chehade, P. Honeine, et.al, Target tracking using machine learning and Kalman filter in wireless sensor networks, IEEE Sensors Journal, IEEE, 2014, 14 (10), pp.3715 3725. F. Santi and D. Pastina, A Parasitic Array Receiver for ISAR Imaging of Ship Targets Using a Coastal Radar, Hindawi Publishing Corporation, International Journal of Antennas and Propagation, Vol. 2016, Art. ID 8485305. http://dx.doi.org/10.1155/2016/8485305. KEYWORDS: Artificial intelligence, AI, machine learning, data fusion