Fully autonomous focused exploration for robotic environmental monitoring

Abstract

Abstract-Robotic sensors are promising instruments for monitoring spatial phenomena. Oftentimes, rather than aiming to achieve low prediction error everywhere, one is interested in determining whether the phenomenon exhibits certain critical behavior. In this paper, we consider the problem of focusing autonomous sampling to determine whether and where the sensed spatial field exceeds a given threshold value. We introduce a receding horizon path planner, LSE-DP, which plans efficient paths for sensing in order to reduce our uncertainty specifically around the threshold value. We report fully autonomous field experiments with an Autonomous Surface Vessel (ASV) in an aquatic monitoring setting, which demonstrate the effectiveness of the proposed method. LSE-DP is able to reduce the uncertainty around the threshold value of interest to 68% when compared to non-adaptive methods.

Adaptive continuous‐space informative path planning for online environmental monitoring

“…However, even in settings with small λ values, which allow for larger deviations from the global path, the goal is not the accurate classification but the detection of areas of interest. This is one of the essential differences between this present work and similar ones, which aimed at accurately estimating level sets . While accurate knowledge about the position of level set boundaries is as well valuable, it often leads to dense accumulations of measurements along the boundaries (as they are the region of interest).…”

“…Alternatively, greedy approaches with limited look‐ahead can be used . These approaches have shorter computation times, enabling their application in adaptive settings, but provide no guarantees regarding optimality due to their limited planning horizon.…”

“…As the CIPP procedure can provide any‐time solutions, this can easily be implemented. A similar planning‐ahead scheme was used in earlier work . The initialization for the global 3D path was done analog to the 2D version we used in the simulations presented in Section : From the starting point a straight line was used to the furthest corner of the environment and back to the starting point.…”

“…Autonomous mobile robots equipped with sensors emerge as a suitable solution, and are increasingly deployed to collect high‐resolution data over large areas for applications such as gas detection, geological surveys, as well as for monitoring around and within water bodies . However, even though autonomous robots can turn tedious manual data collection into an automated, precise, and repeatable sampling strategy, datasets that provide a representative three‐dimensional (3D) depiction of a given environment remain scarce because, even despite recent developments, it is still arduous to automatically plan efficient informative (that is, information‐rich) paths in large 3D environments.…”

Relaxing the planar assumption: 3D state estimation for an autonomous surface vessel

“…Besides taking measurements within the water column below the boat, such as in the work of Brown et al [2011] and Dunbabin et al [2009], or also prior work with the present ASV [Hitz et al, 2014a], ASVs have been used to take measurements of features located above the surface. Hollinger and Sukhatme [2014] have presented an application of an ASV on a lake to estimate the scalar field of wireless signal strength in the context of planning efficient measuring paths.…”