This paper presents an automated method for creating spatial maps of soil condition with an outdoor mobile robot. Effective soil mapping on farms can enhance yields, reduce inputs and help protect the environment. Traditionally, data are collected manually at an arbitrary set of locations, then soil maps are constructed offline using Kriging, a form of Gaussian process regression. This process is laborious and costly, limiting the quality and resolution of the resulting information. Instead, we propose to use an outdoor mobile robot for automatic collection of soil condition data, building soil maps online and also adapting the robot's exploration strategy on-the-fly based on the current quality of the map. We show how using Kriging variance as a reward function for robotic exploration allows for both more efficient data collection and better soil models. This work presents the theoretical foundations for our proposal and an experimental comparison of exploration strategies using soil compaction data from a field generated with a mobile robot.

Soil moisture monitoring is a fundamental process to enhance agricultural outcomes and to protect the environment. The traditional methods for measuring moisture content in the soil are laborious and expensive, and therefore there is a growing interest in developing sensors and technologies which can reduce the effort and costs. In this work, we propose to use an autonomous mobile robot equipped with a state‐of‐the‐art noncontact soil moisture sensor building moisture maps on the fly and automatically selecting the most optimal sampling locations. We introduce an autonomous exploration strategy driven by the quality of the soil moisture model indicating areas of the field where the information is less precise. The sensor model follows the Poisson distribution and we demonstrate how to integrate such measurements into the kriging framework. We also investigate a range of different exploration strategies and assess their usefulness through a set of evaluation experiments based on real soil moisture data collected from two different fields. We demonstrate the benefits of using the adaptive measurement interval and adaptive sampling strategies for building better quality soil moisture models. The presented method is general and can be applied to other scenarios where the measured phenomena directly affect the acquisition time and need to be spatially mapped.

This paper presents a novel method for introducing time into discrete and continuous spatial representations used in mobile robotics, by modelling long-term, pseudo-periodic variations caused by human activities or natural processes. Unlike previous approaches, the proposed method does not treat time and space separately, and its continuous nature respects both the temporal and spatial continuity of the modeled phenomena. The key idea is to extend the spatial model with a set of wrapped time dimensions that represent the periodicities of the observed events. By performing clustering over this extended representation, we obtain a model that allows the prediction of probabilistic distributions of future states and events in both discrete and continuous spatial representations. We apply the proposed algorithm to several long-term datasets acquired by mobile robots and show that the method enables a robot to predict future states of representations with different dimensions. The experiments further show that the method achieves more accurate predictions than the previous state of the art.

Driven by the increasing food demand and the need for higher-quality cultivation, precision agriculture grows steadily during the last decade. It involves the application of mobile robots and intelligent robotic technologies in various agricultural field tasks, concerning a variety of crop types. Aiming at compensating for the lack of selective robotic harvesting solutions regarding the high-value crop of grapes, the EU-funded project BACCHUS ¹ develops an intelligent mobile robotic system, comprising two independent and cooperative robots: one for the grape inspection and collection of valuable data regarding their maturity level, and one for the bimanual harvesting of grapes in a human-inspired manner. Validated via real-field trials, the proposed autonomous system pushes forward the precision agriculture application for a particularly sensitive crop type in the challenging and heavily cluttered environment of vineyards, facilitating the selective harvesting of high-quality grapes.

Long-term studies with autonomous robots “in the wild” (deployed in real-world human-inhabited environments) are among the most laborious and resource-intensive endeavors in human-robot interaction. Even if a robot system itself is robust and well-working, the analysis of the vast amounts of user data one aims to collect and analyze poses a significant challenge. This letter proposes an automated processing pipeline, using state-of-the-art computer vision technology to estimate demographic factors from users' faces and reidentify them to establish usage patterns. It overcomes the problem of explicitly recruiting participants and having them fill questionnaires about their demographic background and allows one to study completely unsolicited and nonprimed interactions over long periods of time. This letter offers a comprehensive assessment of the performance of the automated analysis with data from 68 days of continuous deployment of a robot in a care home and also presents a set of findings obtained through the analysis, underpinning the viability of the approach.