Inexpensive and no-maintenance biodegradable soil moisture sensors could improve existing knowledge on spatial and temporal variability of available soil water at field-scale. Such sensors can unlock the full potential of variable-rate irrigation (VRI) systems to optimize water applications in irrigated cropping systems. The objectives of this study were to assess (i) the degradation of soil moisture sensor component materials and (ii) the effects of material degradation on maize (Zea Mays L.) growth and development. This study was conducted in a greenhouse at Colorado State University, Colorado, USA, by planting maize seeds in pots filled with three growing media (field soil, silica sand, and Promix commercial potting media). The degradation rate of five candidate sensor materials (three blends of beeswax and soy wax, balsa wood, and PHBV (poly(3-hydroxybutyrate-co-3-hydroxyvalerate))) was assessed by harvesting sensor materials at four maize growth stages (30, 60, 90, and 120 days after transplanting). All materials under consideration showed stability in terms of mass and dimension except PHBV. PHBV was degraded entirely within 30 days in soil and Promix, and within 60 days in sand. Balsa wood did now show any significant reduction in mass and dimensions in all growth media. Similarly, there was no significant mass loss across wax blends (p = 0.05) at any growth stage, with a few exceptions. Among the wax blends, 3:1 (beeswax:soy wax) was the most stable blend in terms of mass and dimension with no surface cracks, making it a suitable encapsulant for soil sensor. All materials under consideration did not have any significant effect on maize growth (dry biomass, green biomass, and height) as compared to control plants. These results indicated that 3:1 beeswax:soy wax blend, PHBV, and balsa wood could be suitable candidates for various components of biodegradable soil moisture sensors.

Successful human-robot collaboration depends on a shared understanding of task state and current goals. In nonlinear or freeform tasks without an explicit task model, robot partners are unable to provide assistance without the ability to translate perception into meaningful task knowledge. In this paper, we explore the utility of multimodal recurrent neural networks (RNNs) with long short-term memory (LSTM) units for real-time subtask recognition in order to provide context-aware assistance during generic assembly tasks. We train RNNs to recognize specific subtasks in individual modalities, then combine the high-level representations of these networks through a nonlinear connection layer to create a multimodal subtask recognition system. We report results from implementing the system on a robot that uses the subtask recognition system to provide predictive assistance to a human partner during a laboratory experiment involving a human-robot team completing an assembly task. Generalizability of the system is evaluated through training and testing on separate tasks with some similar subtasks. Our results demonstrate the value of such a system in providing assistance to human partners during a freeform assembly scenario and increasing humans' perception of the robot's agency and usefulness.

Monitoring of microbiological processes in soil is important for understanding and improving soil health and agricultural productivity. Despite its significance, microbiological measurements are currently difficult to make in-situ and in real time. Evaluation of microbially-mediated soil processes usually involves manual sampling followed by laboratory analysis, which can be costly, time consuming, physically intensive, non-continuous, and reduces capacity for measuring changes at high temporal and spatial resolution, thereby limiting the ability to make prompt informed land management decisions. Low-cost soil sensors manufactured using printing techniques offer a potential scalable solution to these issues, allowing for wide distribution of sensors that collect time-series data for decomposition rates. Here, we tested the use of novel sensors for the evaluation of soil microbial processes. This comprises the first large-scale field deployment of these sensors, which use a biodegradable composite conductor that transduces microbial decomposition of substrates to a change in electrical resistance. Sensors were installed for 50 days across 44 experimental plots of a long-term grassland biodiversity restoration experiment, including treatments with additions of synthetic fertilizer, manure, mixed seed, and red clover, causing significant differences in soil microbial activity. Soil biological properties commonly used as soil health indicators, including microbial biomass and enzymatic activities related to carbon, nitrogen and phosphorus cycling, were measured using standard laboratory methods. Sensor responses were compared to these conventional soil health measures to better understand their potential utility by considering the signals of sensors within a set decomposition window and individual sensor degradation rates across various timescales, and correlating changes in signal variability with soil measured properties. All statistical approaches found positive correlations between the sensor signal and laboratory measurements of microbial biomass carbon and soil organic carbon, and some approaches found correlations with enzymatic measurements. These findings demonstrate the potential for the proxy measurement of soil microbial processes in-situ using readily distributable printed decomposition sensors, thereby supporting their potential for low-cost, high-resolution temporal and spatial monitoring of complex soil biological parameters and their ability to provide new insights into soil health.