Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Timely Target Tracking: Distributed Updating in Cognitive Radar Networks

William W. Howard, Anthony F. Martone, R. Michael Buehrer

TL;DR

This work considers a CRN tracking multiple targets with a goal of providing information which is both fresh and accurate to a measurement fusion center (FC), and presents a centralized AoI-inspired node selection metric and a distributed approach which utilizes the Age of Incorrect Information (AoII) metric.

Abstract

Cognitive radar networks (CRNs) are capable of optimizing operating parameters in order to provide actionable information to an operator or secondary system. CRNs have been proposed to answer the need for low-cost devices tracking potentially large numbers of targets in geographically diverse regions. Networks of small-scale devices have also been shown to outperform legacy, large scale, high price, single-device installations. In this work, we consider a CRN tracking multiple targets with a goal of providing information which is both fresh and accurate to a measurement fusion center (FC). We show that under a constraint on the update rate of each radar node, the network is able to utilize Age of Information (AoI) metrics to maximize the resource utilization and minimize error per track. Since information freshness is critical to decision-making, this structure enables a CRN to provide the highest-quality information possible to a downstream system or operator. We discuss centralized and distributed approaches to solving this problem, taking into account the quality of node observations, the maneuverability of each target, and a limit on the rate at which any node may provide updates to the FC. We present a centralized AoI-inspired node selection metric, where a FC requests updates from specific nodes. We compare this against several alternative techniques. Further, we provide a distributed approach which utilizes the Age of Incorrect Information (AoII) metric, allowing each independent node to provide updates according to the targets it can observe. We provide mathematical analysis of the rate limits defined for the centralized and distributed approaches, showing that they are equivalent. We conclude with numerical simulations demonstrating that the performance of the algorithms exceeds that of alternative approaches, both in resource utilization and in tracking performance.

Timely Target Tracking: Distributed Updating in Cognitive Radar Networks

TL;DR

This work considers a CRN tracking multiple targets with a goal of providing information which is both fresh and accurate to a measurement fusion center (FC), and presents a centralized AoI-inspired node selection metric and a distributed approach which utilizes the Age of Incorrect Information (AoII) metric.

Abstract

Cognitive radar networks (CRNs) are capable of optimizing operating parameters in order to provide actionable information to an operator or secondary system. CRNs have been proposed to answer the need for low-cost devices tracking potentially large numbers of targets in geographically diverse regions. Networks of small-scale devices have also been shown to outperform legacy, large scale, high price, single-device installations. In this work, we consider a CRN tracking multiple targets with a goal of providing information which is both fresh and accurate to a measurement fusion center (FC). We show that under a constraint on the update rate of each radar node, the network is able to utilize Age of Information (AoI) metrics to maximize the resource utilization and minimize error per track. Since information freshness is critical to decision-making, this structure enables a CRN to provide the highest-quality information possible to a downstream system or operator. We discuss centralized and distributed approaches to solving this problem, taking into account the quality of node observations, the maneuverability of each target, and a limit on the rate at which any node may provide updates to the FC. We present a centralized AoI-inspired node selection metric, where a FC requests updates from specific nodes. We compare this against several alternative techniques. Further, we provide a distributed approach which utilizes the Age of Incorrect Information (AoII) metric, allowing each independent node to provide updates according to the targets it can observe. We provide mathematical analysis of the rate limits defined for the centralized and distributed approaches, showing that they are equivalent. We conclude with numerical simulations demonstrating that the performance of the algorithms exceeds that of alternative approaches, both in resource utilization and in tracking performance.
Paper Structure (25 sections, 3 theorems, 39 equations, 14 figures, 1 table, 1 algorithm)

This paper contains 25 sections, 3 theorems, 39 equations, 14 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Cognitive Radar Networks
  3. Single Node Techniques
  4. Age of Information
  5. Problem Summary
  6. Contributions
  7. Notation
  8. Organization
  9. System Model
  10. Spatial Modeling
  11. Motion Modeling
  12. Network Modeling
  13. Measurement Model
  14. Update Policies
  15. Centralized Policy
  16. ...and 10 more sections

Key Result

Lemma 1

Let policy $\phi$ be fixed, so that $\#(\mathcal{N}_\phi(t))=C\; \forall t$. Then, $\phi$ is constrained.

Figures (14)

  • Figure 1: Tracking scenario with node density $\lambda_n=3$ and target density $\lambda_m=7$.
  • Figure 2: The Markov chain motion model exhibited by UAV targets.
  • Figure 3: Example network showing which nodes can observe which UAVs. Some targets are observed by multiple nodes and some targets are not observed at all.
  • Figure 4: A sample fused track from the perspective of the FC.
  • Figure 5: As the update rate for each node increases, the number of targets updated increases more slowly.
  • ...and 9 more figures

Theorems & Definitions (8)

  • Definition 1: Constrained Policy
  • Definition 2: Fixed Policy
  • Lemma 1: Fixed Policies are Constrained
  • proof
  • Lemma 2: Constraint Equivalence
  • proof
  • Lemma 2: Constraint Equivalence
  • proof