Heterogeneous Measurement Selection for Vehicle Tracking using Submodular Optimization

TL;DR

This paper introduces a submodular optimization approach for selecting the most informative heterogeneous measurements from multiple sensors to improve vehicle tracking accuracy under communication constraints.

Contribution

It develops a Fisher information matrix-based measurement selection method that leverages submodular properties for efficient and near-optimal sensor data selection in vehicle tracking.

Findings

The proposed method guarantees at least 63% of the optimal estimation accuracy.

FIM-based selection effectively reduces communication load while maintaining tracking performance.

Application to various sensor types demonstrates versatility and effectiveness.

Abstract

We study a scenario where a group of agents, each with multiple heterogeneous sensors are collecting measurements of a vehicle and the measurements are transmitted over a communication channel to a centralized node for processing. The communication channel presents an information-transfer bottleneck as the sensors collect measurements at a much higher rate than what is feasible to transmit over the communication channel. In order to minimize the estimation error at the centralized node, only a carefully selected subset of measurements should be transmitted. We propose to select measurements based on the Fisher information matrix (FIM), as "minimizing" the inverse of the FIM is required to achieve small estimation error. Selecting measurements based on the FIM leads to a combinatorial optimization problem. However, when the criteria used to select measurements is both monotone and submodular it allows the use of a greedy algorithm that is guaranteed to be within $1-1/e\approx63\%$ of the optimum and has the critical benefit of quadratic computational complexity. To illustrate this concept, we derive the FIM criterion for different sensor types to which we apply FIM-based measurement selection. The criteria considered include the time-of-arrival and Doppler shift of passively received radio transmissions as well as detected key-points in camera images.