Machine Data Hackathon
This section breaks down the life-cycle of the Machine Data Hack following the steps described in the main guide.
One of the biggest problems farmers face is the interoperability of farming equipment due to different digital standards. This lack of interoperability is a big pain for farmers: it costs time, money and effort to solve non interoperability issues. It explains much of the current inhibitions that farmers feel when it comes to new technologies.
A second issue concerns machine data access. Best case, equipment manufacturers allow access to data via a specific cloud solution. Farmers still end up with a lot of work to gather data and make it actionable, having different equipment on farm. So as long as manufacturers do not all use open standards (or make use of the possibility to send over 90% in some closed proprietary encrypted format), freeing up machine data is a top priority.
When researching possible solutions, we ran into ISOBlue, and it’s creator, the amazing Aaron Ault from Purdue University. ISOBlue is an open source hardware and software project to free up machine data and connect agricultural machines to the cloud.
At the same time much is being done in terms of standardization, such as the ADAPT framework, which is an interoperability solution for field operations data that is being adopted around the world. The exploration also shed light on relevant, mostly publicly financed, research projects targeting machine data. Check for example the CNHi side of things regarding the topic of machine interoperability (presentation).
We mapped out relevant stakeholders that have a stake in machine data:
Farmers interested in machine data, covering different regions, and both arable and dairy farmers. We included contract workers, who have a clear stake in precision farming. The number one question for this group: what is my machine already telling me that I do not know because I am not looking at the data?
Equipment manufacturers, that are offering specific cloud services, adopting open standards (or not) and are also thinking about issues such as privacy, security and liability.
Machine data loggers: there is an increasing amount of initiatives that claim to free up machine data. Most relevant in our case was a project called uCANData, moving sensor data from agricultural vehicles into a big open database in the cloud.
Data Consumers: possible outlets for machine data. So once freed from the machine, where does the farmer want his data to go? Ranging from raw data for the QGis savy farmer to existing platforms and farm management information systems, we wanted to make all of these stops accessible to the farmer.
FarmHackers: finally, we added in some data scientists, visualisation experts, open source experts, hardware/Iot enthusiasts and open source experts, that were willing to devote their free time to the freeing up of machine data.
We organised a small, targeted on-farm hackathon for 4 teams to work on three challenges, the “Trekkerhack”. We managed to mobilise:
Several farmers interested in machine data (fruit trees, arable farming)
A representative of CNH-i, that was also involved in the Farm Machine Interoperability case of IOF2020
A representative of Beijer Automotive, the driving force behind the uCANData pilot
Two representatives of an existing farmer centric database solution for hands free registration called Farm24
10 FarmHackers, with backgrounds ranging from hardware to software, and from front end to back end engineering.
Team 1: Prototyping the first Dutch ISOBlue
ISOBlue 2.0 was an existing US open source hardware + software platform for collecting real-time data from agricultural machines. We needed to build a prototype of ourselves and work our way through updating the linux kernel and incomplete documentation of the decoding process using Apache Kafka. At the end of the hackathon we had a successful test run.
Team 2: Visualizing and layering of multiple datasets
Data that was available during the hackathon was transformed into GeoJson to visualize soil resistance, tractor speed and tractor engine RPM on a heatmap that showed the actual fields. Check the blogpost with results for more detail.
Team 3: Measuring hitch resistance
This team was especially interested in the resistance on the hitch. Combining these parameters with GPS, they got an indication of the degree of soil compaction. They had to overcome the problem that previously the Case tractor data could not be combined with GPS data from TRIMBLE. Long live the ISOBlue ;-)
Based on the Trekkerhack a partnership was built between the ISOBlue team and FarmHack. We committed to use 2019 to run a pilot with 5 ISOBlues in different regions of the Netherlands. We log all machine data, but to minimise the decoding and cleaning of data we initially focus on: Lift position and load, Engine load, speed and torque, Wheel speed, GPS speed, Fuel consumption, Acceleration, PTO, Working hours. For now we produce CSV files, but we are preparing more advanced ways to disseminate the data.
In addition to the technical challenge, we also invested in building multiple partnerships around machine data, including embedding the pilot in a H2020 innovation experiment, collaborating with a farmer association that wants to build a data hub and preparing a new challenge for an upcoming hackathon for the Dutch government, that wants to offer a universal solution for farmers to participate in handsfree registration and autonomous monitoring.
Link to video (Dutch)
Link to Forum (English) (note: for now most of the conversation is in Dutch in a closed part of the forum, as we are including farmers and local stakeholders around the sensitive issue of real time access to machine data).