Search papers, labs, and topics across Lattice.
This paper introduces a hierarchical perception-aware exploration framework for UAVs in feature-limited environments, explicitly coupling exploration progress with feature observability to improve visual-inertial odometry (VIO) performance. The framework prioritizes visually informative subgoals using a global feature map and optimizes continuous yaw trajectories to maintain stable feature tracks. Experimental results in simulation and real-world indoor environments demonstrate that the proposed method maintains more reliable feature tracking, reduces odometry drift, and achieves significantly higher coverage compared to baselines.
By explicitly considering feature observability, UAVs can achieve 30% higher coverage in feature-limited environments before odometry error exceeds acceptable thresholds.
Autonomous exploration in unknown environments typically relies on onboard state estimation for localisation and mapping. Existing exploration methods primarily maximise coverage efficiency, but often overlook that visual-inertial odometry (VIO) performance strongly depends on the availability of robust visual features. As a result, exploration policies can drive a robot into feature-sparse regions where tracking degrades, leading to odometry drift, corrupted maps, and mission failure. We propose a hierarchical perception-aware exploration framework for a stereo-equipped unmanned aerial vehicle (UAV) that explicitly couples exploration progress with feature observability. Our approach (i) associates each candidate frontier with an expected feature quality using a global feature map, and prioritises visually informative subgoals, and (ii) optimises a continuous yaw trajectory along the planned motion to maintain stable feature tracks. We evaluate our method in simulation across environments with varying texture levels and in real-world indoor experiments with largely textureless walls. Compared to baselines that ignore feature quality and/or do not optimise continuous yaw, our method maintains more reliable feature tracking, reduces odometry drift, and achieves on average 30\% higher coverage before the odometry error exceeds specified thresholds.
