Please select the desired project time frame:

• July 2026
• January 2026
• July 2025
• January 2025
• July 2024
• January 2024
• July 2023
• January 2023
• July 2022
• January 2022
• July 2021
• January 2021
• July 2020
• January 2020
• July 2019
• January 2019
• July 2018
• January 2018
• July 2017
• January 2017
• July 2016
• January 2016
• July 2015
• January 2015
• July 2014
• January 2014
• July 2013
• January 2013
• July 2012
• January 2012
• July 2011
• January 2011
• July 2010
• January 2010
• July 2009
• January 2009
• July 2008
• January 2008
• July 2007
• January 2007
• July 2006
• January 2006
• July 2005
• January 2005
• July 2004
• January 2004
• July 2003
• January 2003
• July 2002
• January 2002
• July 2001
• January 2001

Start of funding 01.07.2018

Data Analytics of Soil Health Data

---

Agricultural production is facing tremendous pressure to increase crop production to keep pace with world population growth while at the same time operate in sustainable fashion. This implies that soil quality has to be maintained despite intensive farming use. In this project we extend an existing Seekur robotic platform (FARMBA) to autonomously and continuously take soil quality metrics via instrumentation. In addition, the platform also includes weather data sensors. The platform will generate a lot of continuous sensor data that has to be stored and processed for analytics. We plan on leveraging our existing data processing pipeline and machine learning techniques and extend them to make them applicable to a completely new problem domain.?We have two main goals for our project: (1) leverage the Docker-based processing pipeline for the FARMBA acquired soil sensor data to develop a scalable and flexible storage and analysis backend for the robot data, and (2) develop machine learning models that enable the automatic detection of impending soil quality degradation and the identification of contributing factors.?

Final report:
The originally proposed research focus of providing data analytics for soil health turned out to be infeasible. Unfortunately, the robotic platform that was supposed to generate the soil-related data at UCSB did not become available in time (in fact, as of 9/2019 it is still not available). However, we were able to pivot our efforts in the same area (IoT and data analytics for smart farming) to another focus, which turned out be very successful and resulted in two directly related publications in two top-rated conferences.

The work consequently concentrated on two related efforts: sensor synthesis and edge computing (not only) for smart farms. One problem that popular IoT computation platforms such as the Raspberry Pi or Arduino microcontroller face consists of the limited (port) space for sensors. Therefore, if one desired measurement, e.g., the outdoor temperature can be derived (= synthesized) from other values, the saved space can be used for other sensors or reduce the instrumentation energy cost.

In our ICIOT paper, we were able to successfully synthesize the outdoor temperature based on the Raspberry Pi's CPU temperature combined with an Internet weather service temperature for the farm location to a precision of 1 degree Fahrenheit.

A second significant problem for smart farming is the provision of a durable computing platform at the farm site, i.e., providing an edge computing platform. Edge computing is necessary because Internet connectivity if often inadequate or even unavailable at farm locations, which is the case for the UCSB smart farm. Also, since the farm is at a significant distance to where the computer-savvy personnel is located (UCSB's main campus), the computing platform has to be robust, and selfhealing as any crash would result in a long time to repair and consequently low system availability.

In our Edge Computing paper, we demonstrated techniques on how to implement a durable edge cloud by creating a system (named Mandrake) that enables the hosting of a Hadoop Cluster using Intel NUC devices. Mandrake is very robust and self-healing and keeps running as long as no failure eliminates 50% or more of all computing nodes at the same time. In the case of a node failure, Mandrake gracefully scales down to the remaining compute nodes and re-integrates nodes that have been repaired (e.g., by power-cycling, something that can be performed even by local farm personnel) into the edge computing cluster. This process does not affect actual workloads except for a performance degradation that is commensurate with the amount of lost computing power.