Active Jammer Localization via Acquisition-Aware Path Planning

TL;DR

The paper tackles localizing a single static GNSS jammer in urban environments using crowdsourced RSS measurements. It introduces an active framework that combines Bayesian optimization with acquisition-aware path planning (A-UCB*), employing a Gaussian Process surrogate and a UCB acquisition to guide an autonomous agent through obstacle-rich cities. Key contributions include a novel acquisition-aware BO framework, a path planner that integrates acquisition gain into movement costs, and demonstrated sample-efficient localization with limited measurements across realistic urban layouts. Results show faster convergence and robust performance across environments, highlighting practical potential for resilient PNT operations in complex urban settings.

Abstract

We propose an active jammer localization framework that combines Bayesian optimization with acquisition-aware path planning. Unlike passive crowdsourced methods, our approach adaptively guides a mobile agent to collect high-utility Received Signal Strength measurements while accounting for urban obstacles and mobility constraints. For this, we modified the A* algorithm, A-UCB*, by incorporating acquisition values into trajectory costs, leading to high-acquisition planned paths. Simulations on realistic urban scenarios show that the proposed method achieves accurate localization with fewer measurements compared to uninformed baselines, demonstrating consistent performance under different environments.

Figures (3)

• Figure 1: Overview of the proposed active localization framework. Top: GP posterior mean $\mu_n(\mathbf{x})$ and uncertainty $\sigma_n(\mathbf{x})$ over $\mathcal{F}$. Bottom-right: UCB acquisition $\alpha_{\mathrm{UCB}}(\mathbf{x})$ used to select the next sensing target $\mathbf{x}^{c}_{n}$ ($\star$). The black curve is the past trajectory and the green dashed line is the path planning toward $\mathbf{x}^{c}_{n}$. Gray polygons are buildings (non-traversable) and black dots are initial crowdsourced samples $\mathcal{D}_0$.
• Figure 2: Evolution of the BOE in the Chicago Downtown (top panel) and Boston Common (bottom panel) scenarios across 80 Bayesian optimization iterations. Boxplots at each iteration show the distribution of localization errors across 100 independent trials. The inset zooms into the last 30 iterations highlight last convergence behavior.
• Figure 3: Sensitivity analysis of the exploration–exploitation parameter $\kappa$ in the UCB acquisition function. The plot shows the BOE after 30 iterations for both datasets. Each point corresponds to the median BOE across 100 trials, with error bars denoting the IQR.