Informative Sensor Planning for a Single-Axis Gimbaled Camera on a Fixed-Wing UAV

TL;DR

The paper addresses efficient information gathering with a forward-looking EO sensor on a single-axis gimbal mounted to a fixed-wing UAV. It proposes an informative non-myopic sensor planning framework that computes dynamic yaw bounds $(\psi_1, \psi_2)$ over a planning horizon $h$ by constructing a planning-horizon polygon from the current and projected UAV positions and selecting boundary high-information cells to drive the yaw limits; this sensor planner is cascaded with a global IPP planner (TIGRIS) by widening the effective FOV to reflect gimbal motion. The main contributions are a boundary-high-information cell–driven sensor planning algorithm, integration with a global IPP planner in a cascaded architecture, and end-to-end validation on a fixed-wing VTOL with an EO sensor in simulations and Isaac Sim showing substantial entropy reduction compared with baselines. The framework improves data acquisition efficiency and mission success by prioritizing high-information regions while keeping computation tractable for real-time operation, enabling more robust search missions in large environments.

Abstract

Uncrewed Aerial Vehicles (UAVs) are a leading choice of platforms for a variety of information-gathering applications. Sensor planning can enhance the efficiency and success of these types of missions when coupled with a higher-level informative path-planning algorithm. This paper aims to address these data acquisition challenges by developing an informative non-myopic sensor planning framework for a single-axis gimbal coupled with an informative path planner to maximize information gain over a prior information map. This is done by finding reduced sensor sweep bounds over a planning horizon such that regions of higher confidence are prioritized. This novel sensor planning framework is evaluated against a predefined sensor sweep and no sensor planning baselines as well as validated in two simulation environments. In our results, we observe an improvement in the performance by 21.88% and 13.34% for the no sensor planning and predefined sensor sweep baselines respectively.

Figures (6)

• Figure 1: An example scenario of a UAV sweeping a gimbaled camera to search for a missing hiker. As the UAV follows its straight line path over the mountain, the sensor planner prioritizes sweeping over the river.
• Figure 2: A visualization of how $f_{\text{current}}$ and $f_{\text{current}}$ are created and combined to create a planner horizon polygon.
• Figure 3: An example scenario showing the belief space as well as the output of our sensor planner. You can see the boundary high information grid cells marked with green and blue squares in the top row of images. The actual gimbal yaw is plotted along with the gimbal bounds to show how the gimbal sweeps between the bounds. The gimbal focuses on the regions of high information gain and ignores the low-value areas.
• Figure 4: Sensor model used for implementing the proposed sensor planning algorithm based on (\ref{['3.3']}).
• Figure 5: Results of 100 runs comparing our proposed approach to a system with a system having a predefined sensor sweep and a system having no sensor sweeping. The shaded regions represent the 95% confidence intervals for the mean percent reduction in entropy between the approaches. The second figure shows the percentage difference between our sensor planner and the two approaches.