Next-Generation Cognitive Radar Systems

Cover
Contents
About the editors
List of editors
List of contributors
List of reviewers
Preface
Acknowledgments
Part I Fundamentals
	1 Beyond cognitive radar
		1.1 Aspects of cognition
		1.2 Key technology enablers
			1.2.1 Convex and non-convex optimization
			1.2.2 Control-theoretic tools
			1.2.3 Learning techniques
			1.2.4 Operationalization
		1.3 Organization of the book
		References
	2 Adversarial radar inference: inverse tracking, identifying cognition, and designing smart interference
		2.1 Introduction
			2.1.1 Objectives
			2.1.2 Perspective
			2.1.3 Organization
		2.2 Inverse tracking and estimating adversary’s sensor
			2.2.1 Background and preliminary work
			2.2.2 Inverse tracking algorithms
			Example: inverse Kalman filter
			2.2.3 Estimating the adversary’s sensor gain
			2.2.4 Example. Estimating adversary’s gain in linear Gaussian case
		2.3 Identifying utility maximization in a cognitive radar
			2.3.1 Background. Revealed preferences and Afriat’s theorem
			2.3.2 Beam allocation: revealed preference test
			2.3.3 Waveform adaptation: revealed preference test for non-linear budgets
		2.4 Designing smart interference to confuse cognitive radar
			2.4.1 Interference signal model
			2.4.2 Smart interference for confusing the radar
			2.4.3 Numerical example illustrating design of smart interference
		2.5 Stochastic gradient-based iterative smart interference
			2.5.1 Smart interference with measurement noise
			2.5.2 Algorithms for solving constrained optimization problem (2.41)
		Acknowledgment
		References
	3 Information integration from human and sensing data for cognitive radar
		3.1 Integration of human decisions with physical sensors in binary hypothesis testing
			3.1.1 Decision fusion for physical sensors and human sensors
			3.1.2 Asymptotic system performance when humans possess side information
		3.2 Prospect theoretic utility-based human decision making in multi-agent systems
			3.2.1 Subjective utility-based hypothesis testing
			3.2.2 Decision fusion involving human participation
		3.3 Human–machine collaboration for binary decision-making under correlated observations
			3.3.1 Human–machine collaboration model
			3.3.2 Copula-based decision fusion at the FC
			3.3.3 Performance evaluation
		3.4 Current challenges in human–machine teaming
		3.5 Summary
		References
	4 
Channel estimation for cognitive fully adaptive radar
		4.1 Introduction
		4.2 Traditional covariance-based statistical model
		4.3 Stochastic transfer function model
		4.4 Cognitive radar framework
		4.5 Unconstrained channel estimation algorithms
			4.5.1 SISO/SIMO channel estimation
			4.5.2 MIMO channel estimation
			4.5.3 Minimal probing strategies
		4.6 Constrained channel estimation algorithm
			4.6.1 Cosine similarity measurement
			4.6.2 Channel estimation under the cosine similarity constraint: non-convex QCQP
			4.6.3 Performance comparison using numerical simulation
		4.7 Cognitive fully adaptive radar challenge dataset
			4.7.1 Scenario 1
			4.7.2 Scenario 2
		4.8 Concluding remarks
		References
	5 
Convex optimization for cognitive radar
		5.1 Introduction
			5.1.1 Waveform design problems in cognitive radar
		5.2 Background and motivation
			5.2.1 Principles of convex optimization
			5.2.2 Challenges of optimization problems for cognitive radar
		5.3 Constrained optimization for cognitive radar
			5.3.1 SINR maximization
			5.3.2 Spatio-spectral radar beampattern design
			5.3.3 Quartic gradient descent for tractable radar ambiguity function shaping
		5.4 Summary
		References
Part II
 Design methodologies
	6 
Cognition-enabled waveform design for ambiguity function shaping
		6.1 Introduction
		6.2 Preliminaries to AF and optimization methods
			6.2.1 Ambiguity function and its shaping
			6.2.2 MM and Dinkelbach’s algorithm
		6.3 Waveform design for AF shaping via SINR maximization
			6.3.1 System model and problem formulation
			6.3.2 Waveform design via MM
			6.3.3 Convergence analysis and accelerations
			6.3.4 Numerical experiments
		6.4 Waveform design via minimization of regularized spectral level ratio
			6.4.1 Regularized SLR and problem formulation
			6.4.2 Approximate iterative method for spectrum shaping
			6.4.3 Monotonic iterative method for spectrum shaping
			6.4.4 Numerical experiments
		6.5 Conclusions
		Appendix
			A.1 Proof of Lemma 2
			A.2 Proof of Lemma 4
			A.3 Proof of Lemma 5
			A.4 Proof of Lemma 6
			A.5 Proof of Lemma 8
			A.6 Proof of Lemma 9
		References
	7 
Training-based adaptive transmit–receive beamforming for MIMO radars
		7.1 Introduction
			7.1.1 Background
			7.1.2 Contributions
		7.2 System model
			7.2.1 Target contribution
			7.2.2 Clutter contribution
			7.2.3 Noise model
		7.3 Adaptive beamforming
			7.3.1 Receive beamforming
			7.3.2 Transmit beamforming: known covariance
			7.3.3 Transmit BF: estimating the required covariance matrix
		7.4 Reduced-dimension transmit beamforming
		7.5 Transmit BF for multiple Doppler bins
		7.6 Numerical results
			7.6.1 Random phase radar signals
			7.6.2 Airborne radar
		7.7 Conclusion
		Acknowledgment
		References
	8 
Random projections and sparse techniques in radar
		8.1 Introduction
		8.2 A critical perspective on sub-sampling claims in compressive sensing theory
			8.2.1 General issues of non-stationarity
			8.2.2 Sparse signal in intermediate frequency (IF)
			8.2.3 Temporally sparse signal in baseband
		8.3 Random projections STAP model
			8.3.1 Computational complexity and a “small” data problem
			8.3.2 Random projections
			8.3.3 Localized random projections
			8.3.4 Semi-random localized projection
		8.4 Statistical analysis
			8.4.1 Probabilistic bounds
		8.5 Simulations
			8.5.1 Integration as low-pass filtering
			8.5.2 CS: sinusoid in IF example
			8.5.3 CS: rectangular pulse example
			8.5.4 Realistic examples of CS reconstructions
			8.5.5 Random projections with different distributions
			8.5.6 Random and random-type projections
		8.6 Discussion and conclusions
		Acknowledgment
		References
	9 
Fully adaptive radar resource allocation for tracking and classification
		9.1 Introduction
		9.2 Fully adaptive radar framework
		9.3 Multitarget multitask FARRA system model
			9.3.1 Radar resource allocation model
			9.3.2 Controllable parameters
			9.3.3 State vector
			9.3.4 Transition model
			9.3.5 Measurement model
		9.4 FARRA PAC
			9.4.1 Perceptual processor
			9.4.2 Executive processor
		9.5 Simulation results
		9.6 Experimental results
		9.7 Conclusion
		Acknowledgment
		References
	10 
Stochastic control for cognitive radar
		10.1 Introduction
		10.2 Connection to earlier work
		10.3 Stochastic optimization framework
			10.3.1 General problem components
			10.3.2 Partial observability
		10.4 Objective functions for cognitive radar
			10.4.1 Task-based reward functions
			10.4.2 Information theoretic reward functions
			10.4.3 Utility and QoS-based objective functions
		10.5 Multi-step objective function
			10.5.1 Optimal values and policies
			10.5.2 Simplified multi-step objective functions
		10.6 Policies and perception–action cycles
			10.6.1 Policy search
			10.6.2 Lookahead approximations
			10.6.3 Discussion
		10.7 Relationship between cognitive radar and stochastic optimization
			10.7.1 Problem components
			10.7.2 Typical cognitive radar solution methodologies
			10.7.3 Cognitive radar objective functions
		10.8 Simulation examples
			10.8.1 Adaptive tracking example
			10.8.2 Target resource allocation example
		10.9 Conclusion
		References
	11 
Applications of game theory in cognitive radar
		11.1 Introduction
			11.1.1 Research background
			11.1.2 Literature review
			11.1.3 Motivation
			11.1.4 Major contributions
			11.1.5 Outline of the chapter
		11.2 System and signal models
			11.2.1 System model
			11.2.2 Signal model
		11.3 Game theoretic formulation
			11.3.1 Feasible extension
		11.4 Existence and uniqueness of the Nash equilibrium
			11.4.1 Existence
			11.4.2 Uniqueness
		11.5 Iterative power allocation method
		11.6 Simulation results and performance evaluation
			11.6.1 Parameter designation
			11.6.2 Numerical results
		11.7 Conclusion
		References
	12 
The role of neural networks in cognitive radar
		12.1 Cognitive process modeling with neural networks
			12.1.1 Background and motivation
			12.1.2 Situation awareness and connection to perception–action cycle
			12.1.3 Memory and attention
			12.1.4 Knowledge representation
			12.1.5 A three-layer cognitive architecture
			12.1.6 Applications of machine learning in a cognitive radar architecture
		12.2 Integration of domain knowledge via physics-aware DL
			12.2.1 Physics-aware DNN training using synthetic data
			12.2.2 Adversarial learning for initialization of DNNs
			12.2.3 Generative models and their kinematic fidelity
			12.2.4 Physics-aware DNN design
			12.2.5 Addressing temporal dependencies in time-series data
		12.3 Reinforcement learning
			12.3.1 Overview
			12.3.2 Basics of reinforcement learning
			12.3.3 Q-Learning algorithm
			12.3.4 Deep Q-network algorithm
			12.3.5 Deep deterministic policy gradient algorithm
			12.3.6 Algorithm selection
			12.3.7 Example reinforcement learning implementation
			12.3.8 Cautionary topics
			12.3.9 Angular action spaces
			12.3.10 Accuracy of environment during training
		12.4 End-to-end learning for jointly optimizing data to decision pipeline
			12.4.1 End-to-end learning architecture
			12.4.2 Loss function of the end-to-end architecture
			12.4.3 Simulation results
		12.5 Conclusion
		Acknowledgments
		References
Part III 
Beyond cognitive radar—from theory to practice
	13 
One-bit cognitive radar
		13.1 Introduction
		13.2 System model
		13.3 Bussgang-theorem-aided estimation
		13.4 Radar processing for stationary targets
			13.4.1 Estimation of stationary target parameters
			13.4.2 Time-varying threshold design
		13.5 Radar processing for moving targets
			13.5.1 Problem formulation for moving targets
			13.5.2 Estimation of moving target parameters
		13.6 Other low-resolution sampling scenarios
			13.6.1 Extension to parallel one-bit comparators
			13.6.2 Extension to p-bit ADCs
		13.7 Numerical analysis for one-bit radar signal processing
			13.7.1 Stationary targets
			13.7.2 Moving targets
		13.8 One-bit radar waveform design under uncertain statistics
			13.8.1 Problem formulation for waveform design
			13.8.2 Joint design method: CREW (one-bit)
		13.9 Waveform design examples
		13.10 Concluding remarks
		References
	14 
Cognitive radar and spectrum sharing
		14.1 The spectrum problem
			14.1.1 Introduction
			14.1.2 Spectrum and spectrum allocation
			14.1.3 Cognitive radar definition
			14.1.4 Target-matched illumination
			14.1.5 Embedded communications
			14.1.6 Low probability of intercept (LPI)
			14.1.7 Summary
		14.2 Joint radar and communications research
			14.2.1 Applications of joint radar and communication
			14.2.2 Co-existence radar and communication research
			14.2.3 Single waveform tasked with both radar and communication
			14.2.4 LPI radar and communication waveforms
			14.2.5 Adaptive/cognitive radar concepts and examples
		14.3 Summary and conclusions
		Acknowledgments
		References
	15 
Cognition in automotive radars
		15.1 Introduction
		15.2 Review of automotive radar
			15.2.1 Automotive radar
			15.2.2 FMCW radar
			15.2.3 MIMO radar and angle estimation
		15.3 Cognitive radar
			15.3.1 Perception–action cycle
			15.3.2 Perception
			15.3.3 Learning
			15.3.4 Action
		15.4 Physical environment perception for FMCW automotive radars
			15.4.1 Range–velocity imaging
			15.4.2 Micro-Doppler imaging
			15.4.3 Range–angle imaging
			15.4.4 Synthetic aperture radar imaging
			15.4.5 Radar object recognition based on radar image
		15.5 Cognitive spectrum sharing in automotive radar network
			15.5.1 Spectrum congestion, interference issue, and MAC schemes
			15.5.2 FMCW-CSMA-based spectrum sharing
			15.5.3 FMCW-cognitive-CSMA-based spectrum sharing
			15.5.4 Comments on spectrum sharing for cognitive radar
		15.6 Concluding remarks
		References
	16  A canonical cognitive radar architecture
		16.1 A canonical CR architecture
		16.2 Full transmit–receive adaptivity
			16.2.1 Full transmit adaptivity
			16.2.2 Full receive adaptivity
		16.3 CR real-time channel estimation (RTCE)
		16.4 CR radar scheduler
		16.5 Cognitive radar and artificial intelligence
		16.6 Implementation considerations
		16.7 Advanced modeling and simulation to support cognitive radar
		16.8 Remaining challenges and areas for future research
		References
	17 
Advances in cognitive radar experiments
		17.1 The need for cognitive radar experiments
			17.1.1 Cognition for radar sensing
			17.1.2 Chapter overview
		17.2 The CREW test bed
			17.2.1 The CREW design
			17.2.2 CREW demonstration experiments
		17.3 The cognitive detection, identification, and ranging testbed
			17.3.1 Development considerations
			17.3.2 The CODIR design
			17.3.3 Experimental work with CODIR
		17.4 Universal software radio peripheral-based cognitive radar testbed
			17.4.1 USRP testbed design
			17.4.2 USRP testbed demonstration experiments
		17.5 The miniature cognitive detection, identification, and ranging testbed
			17.5.1 The miniCODIR design
			17.5.2 miniCODIR experiments
		17.6 Other cognitive radar testbeds
			17.6.1 SDRadar: cognitive radar for spectrum sharing
			17.6.2 Spectral coexistence via xampling (SpeCX)
			17.6.3 Anticipation in NetRad
		17.7 Future cognitive radar testbed considerations
			17.7.1 Distributed cognitive radar systems
			17.7.2 Machine learning techniques
			17.7.3 Confluence of algorithms—metacognition
		17.8 Summary
		Acknowledgments
		References
	18 
Quantum radar and cognition: looking for a potential cross fertilization
		18.1 Introduction
		18.2 Cognitive radar
			18.2.1 Cognitive radar scheduler
			18.2.2 Within the cognitive radar
			18.2.3 Verification and validation
		18.3 Quantum mechanics in a nutshell
		18.4 Quantum harmonic oscillator
		18.5 Quantum electromagnetic field
			18.5.1 Single mode
			18.5.2 Multiple modes
		18.6 Quantum illumination
		18.7 An experimental demonstration
		18.8 Hybridization of cognitive and quantum radar: what recent research in neuroscience can tell about
		18.9 Quantum and cognitive radar
		18.10 Conclusions
		Acknowledgments
		References
	19 
Metacognitive radar
		19.1 Metacognitive concepts in radar
			19.1.1 Metacognitive cycle
			19.1.2 Applications: metacognitive spectrum sharing
			19.1.3 Applications: metacognitive power allocation
			19.1.4 Applications: Metacognitive antenna selection
		19.2 Cognition masking
		19.3 Example: antenna selection across geometries
			19.3.1 Cognitive cycle
			19.3.2 Knowledge transfer across different array geometries
		19.4 Numerical simulations
		19.5 Summary
		References
Epilogue
Index
Back Cover