Radar Analysis and Prediction of Intentions/Behaviour of small droners swarms
Abstract
The objective of this proposed research effort is to establish innovative techniques for analyzing and possibly predicting the behavior of swarms of Unmanned Aerial Vehicles (UAVs) monitored by radar systems. The PI aims to develop a comprehensive simulation environment capable of modelling the radar signatures of multiple cooperating UAVs. This shall be achieved by integrating the existing simulation engine MAVERIC (Modelling of Autonomous Vehicles using Robust Intelligent Computing) developed at the University of Glasgow, together with experimental radar data previously collected, and leveraging currently undertaken experimental work to maximize the realism of Radar Cross Section (RCS), micro-Doppler spectra, and tracks of UAVs. In particular, the goal of the PI and his group is to develop innovative classification techniques to distinguish individual consistent versus random behavior, identify platforms that are acting cooperatively, and infer and possibly predict if this behavior can be associated to hostile intentions. The aim being to establish transformative capabilities to monitor wide areas with key assets where many autonomous platforms can operate, in order to identify patterns of deliberate and/or cooperative behavior, and possibly predict anomalous and hostile activities depending on specific metrics. The realization of these goals will enable safe and reliable deployment and coexistence of many autonomous platforms for a variety of civilian applications, and the capabilities to react promptly and effectively in case of anomalies or suspicious/hostile behavior detected.The PIs focuses on radar as the key enabler monitoring technology for a number of advantages over alternatives, such as its immunity to acoustic noise and external light (day-night) or weather conditions (rain, fog, smoke), and its capabilities to jointly estimate distance and velocity of targets directly at the detection stage using range-Doppler processing algorithms.Specifically, the PI seeks to develop a comprehensive simulation framework to model swarms of autonomous UAVs, their radar signatures for different operational parameters (of the radar itself and of the deployment geometry), and their behavior in different scenarios (urban vs rural scenarios, presence of key assets and no-fly zones that must be respected). The flexibility provided by this model will then be exploited to generate a wide set of data to establish and validate classification techniques to characterize and when possible anticipate the behavior of swarms of autonomous platforms. Innovative machine learning techniques will be utilized to infer and predict the behavior of possibly hostile autonomous agents from radar data, delivering a transformative change from just physical sensing to anomaly prediction and behavioral inference
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 13, 2019
- Source ID
- N629091912073
Entities
People
- Francesco Fioranelli
Organizations
- Office of Naval Research
- United States Navy
- University of Glasgow