An Open-Source Infrastructure for Real-Time AutomaticAgricultural Machine Data Processing

Yang Wang Sam Noel James Krogmeier Dennis BuckmasterOpen Ag Technology and Systems (OATS) Center

Purdue University

[1] httpswwwisoblueorg [2] httpskafkaapacheorg [3] httpopenagio [4] httpsgithubcomOATS-Groupwheres-my-isoblue [5] httpsgithubcomOpenATKTrialsTracker

Connected Machinery via ISOBlue 20

ISOBlue 20 [1] is an open-source ag IoT It is made tobull Be dust and weatherproofbull Be wake on CAN activitybull Collect machine data via ISOBUS diagnostic portbull Opportunistically stream collected data to Cloud

via a 4GLTE connection

CONNECTED MACHINERY CLOUD WEBMOBILE APPS

Cloud Architecture Powered by OADA API Real-time Data Viz amp Analytics via Apps

bull Ag machine data no longer ldquotrappedrdquo

bull Real-time diagnostic info as-harvested as-applied data

Figure 1 (left to right) an assembled and a deployed ISOBlue 20

Apache Kafka [2] manages the collected log on ISOBlue 20 it is capable of handling massive amount of incoming data

bull Microservice-based (scalable)

bull Handles massive amount of incoming data

bull Stream-processes data continuously

machines

hellipsprayers hellipharvesters

Graph database and linked resources in OADA

helliptiled-mapsdry-yield-map

Non-versioned links

hellipgeohash-index

Resource-watching microservice in OADA

Open Ag Data Alliance (OADA) API [3] is an open-source ag data exchange API that offers data security privacy and interoperability

Low-level scripts Kafka broker

Kafka MirrorMaker

Kafka queue

Workflow amp Software Used Workflow amp Software Used

Connected machinery

ISOBlue 20

OADA conformant Cloud

Workflow amp Software Used

Applications

Kafka brokeruservices

Kafka queue

Graph data store

Websockets

TrialsTracker Where-is-my-ISOBlue

T T

Figure 3 a screenshot of where-is-my-isoblueweb app interface

Where-is-my-ISOBlue [4] is a proof-of-concept web app that displays the real-time GPS tracks and debug info for different deployed ISOBlue 20s

TrialsTracker [5] enables user to ldquoaggregate-by-fingerrdquo to compute average field yield and compare trails with both historic and streaming data

AcknowledgementsMore info on ISOBlue 20

bull Offer intuitive visualizations to end users

bull Process both historic and real-time data

Figure 2 high-level Kafka workflow

Figure 4 a screenshot of TrialsTracker web app interface