COMPANY RESTRICTED NOT EXPORT CONTROLLED NOT CLASSIFIED Tomas

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X MATEMATIK, SIGNALBEHANDLING OCH AI INOM RADAR EXEMPEL Milena Anguelova, Adam Andersson Radar Solutions, SAAB Surveillance This document and the information contained herein is the property of Saab AB and must not be used, disclosed or altered without Saab AB prior written consent. 1

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X ORGANISATION SAAB och Radar Solutions Kort om radar C 2 Stör- och klotterundertryckning Målföljning (Target tracking) Target tracking Klassificering Glesa antenn-arrayer AI (Adam) Signal processing Frågor Antenna

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X SAAB OCH VÅRA GRUPPER Försvar och säkerhet Både militära och civila produkter SAAB -> Surveillance -> Radar Solutions -> New Concepts (Milena) SAAB -> Surveillance -> Radar Solutions -> Airborne Early Warning: System design, radar and digital sensors (Adam)

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X KORT OM RADAR RAdio Detection And Ranging • Mäter avstånd till ett mål genom att sända radiovågor och mäta tiden det tar för ekot att återvända • Mäter riktning genom att fokusera energin i smala lober med en direktiv antenn (array) • Detektion och målföljning, i dagsljus och mörker, regn och dimma 4

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X THE RADAR DATA CUBE C 2 Target tracking x xxxx xx x x xxx Detections Signal processing Can process each dimension independently or more than one dimensions together Antenna Constant or data-dependent coefficients • Adaptive signal processing

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X BASIC SIGNAL PROCESSING Beamforming BMF PCP DPL DET • Weights on antenna elements Pulse compression PCP • Matched filtering Doppler processing (including clutter suppression) DPL • FFT, MTI-filter Detection DET • CFAR

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X 7 JAMMER SUPPRESSION target jammer Jammed Antenna array Zoom on target After jammer suppression

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X 8 JAMMER SUPPRESSION – HOW? target jammer • Antenna array 1 2 m … Target Jammer + noise

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X JAMMER SUPPRESSION - CONTEXT Adaptive signal processing Spectral analysis Parameter estimation in colored noise Cramér-Rao bounds Maximum likelihood methods Iterative methods for covariance-based spectral estimation Sparse estimation 9

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X TARGET TRACKING C 2 Target tracking x xx x x x xx x Signal processing Detections Antenna Scene x x x T 1 x x T 2 x x xx x x T 3 Tracks

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X TARGET TRACKING Detections Tracks

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X MULTI-HYPOTHESIS TRACKING consider multiple associations hypotheses over time difficult decisions postponed until more data available T 2 T 1 [] 1 2 Hypothesis = sequence of associated detections Hypothesis tree describes possible associations for each target T 1 M 1 Prediction errors T 2 M 3

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X MULTI-HYPOTHESIS TRACKING T 2 T 1 Each hypothesis updated with new detection (if any) [] 1 T 1 M 4 T 2 M 5 New measurements 2

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X MULTI-HYPOTHESIS TRACKING Problem: number of hypotheses grows fast! T 2 T 1 Each hypothesis predicted to new time-point and updated with new detection (if any) [] [] 4 5 [] 4 1 [] 1 5 [] 4 5 T 1 M 4 T 2 M 5 [] 4 5 2 [] 4 5

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X TRACK FILTERING Filtering for each track hypothesis Multi-mode filtering – filter bank • Constant velocity, coordinated turn, … Interacting multiple models (IMM) Each individual filter e. g. Kalman filter (Extended, Unscented, …)

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X TARGET TRACKING - CONTEXT State New measurement Predicted Target state Updated target state Measurement KALMAN FILTERING Posterior probability distribution of target state

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X TARGET CLASSIFICATION Small UAV’s are becoming an increasing threat How to distinguish small and slow UAVs from birds? Classical methods Machine learning Master’s thesis ”Classification between birds and UAVs using recurrent neural networks” • Henrik Andersson and Chi Thong Luong • Classify track segments - UAV vs bird • Recurrent neural networks (RNN) - long-short term memory (LSTM) - gated recurrent unit (GRU) 17

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X ANTENNA ARRAY DESIGN 18

COMPANY RESTRICTED | NOT EXPORT CONTROLLED | NOT CLASSIFIED Tomas Bryllert | Document number | Issue X SPARSE ANTENNA ARRAYS • Fewer elements ‒ Price ‒ Weight • Can achieve the same resolution as the filled array • Higher side-lobe levels • Design methods for sparse arrays 1. 2. 3. 4. Difference and almost difference sets Fractals Genetic algorithms … 19