Active visual SLAM for robotic area coverage: Theory and experiment

Abstract

This paper reports on an integrated navigation algorithm for the visual simultaneous localization and mapping (SLAM) robotic area coverage problem. In the robotic area coverage problem, the goal is to explore and map a given target area within a reasonable amount of time. This goal necessitates the use of minimally redundant overlap trajectories for coverage efficiency; however, visual SLAM's navigation estimate will inevitably drift over time in the absence of loop-closures. Therefore, efficient area coverage and good SLAM navigation performance represent competing objectives. To solve this decision-making problem, we introduce perception-driven navigation, an integrated navigation algorithm that automatically balances between exploration and revisitation using a reward framework. This framework accounts for SLAM localization uncertainty, area coverage performance, and the identification of good candidate regions in the environment for visual perception. Results are shown for both a hybrid simulation and real-world demonstration of a visual SLAM system for autonomous underwater ship hull inspection.

Two-Stage Focused Inference for Resource-Constrained Collision-Free Navigation

“…Another popular application in visionbased systems is to down-sample landmarks based on some measure of visual saliency in the hope of improving loop closure detection. Specific applications include active gaze control [14], area coverage [15], and lifelong operation of service robots [16].…”

Application-oriented design space exploration for SLAM algorithms

“…Most active SLAM algorithms are based on maximizing mutual information [8], [31], which is also referred to as maximizing information gain [32], [33]. These algorithms are for various applications such as increased coverage, decreased pose uncertainty, or dense mapping purposes.…”

“…1). Unlike other information theoretic methods, such as those which try to maximize mutual information or information gain [8], [31], [32], [33], in our method we propose to limit the information divergence, to ensure that the SLAM system is robust with respect to the structure of the observed scene.…”

Risk aversion in belief-space planning under measurement acquisition uncertainty

“…These works find locally-optimal solutions to the planning problem and provide promising results for robotics applications, especially point-to-point planning queries. Work in path planning for information gathering [12] and active SLAM systems [13][14][15] focused more on the interaction between planning and SLAM, and how the performance and efficiency of SLAM is improved with intelligent decisions regarding which paths to travel.…”

“…They derived the Kalman filter equations as functions of the stochastic acquisition variables. Kim and Eustice [14] and Indelman et al [2] included acquisition variables in belief-space planning for robotics, but their formulations removed the effect of the acquisition randomness in the resulting belief. Instead, we seek to model the variability in the outcomes with respect to uncertainty in order to design objective functions for planning that are sensitive to risk.…”