Realtidsanalys av simulerade sensordata från truckar med hjälp av AWS
(2020) In CODEN:LUTEDX/TEIE EIEL05 20201Industrial Electrical Engineering and Automation
- Abstract (Swedish)
- Syftet med examensarbetet har varit att utveckla och utvärdera ett realtidssystem som utför analys i molnet och lokalt med hjälp av AWS. De viktigaste målen har varit att skapa ett system som är så snabbt som möjligt och att analysera vilken del av analysen som kan göras i molnet och vilken del som måste göras lokalt. Den förväntade nyttan med systemet är de minskade utvecklings- och underhållskostnader som kommer av att placera funktionalitet i molnet.
Examensarbetet har utgått ifrån tre användarfall inom truckhantering och använt simulerade sensordata från en trucksimulering i Node-RED som indata som ska analyseras. Det första användarfallet handlar om att förhindra att truckar tippar genom att analysera gafflarnas höjd och
lutning.... (More) - Syftet med examensarbetet har varit att utveckla och utvärdera ett realtidssystem som utför analys i molnet och lokalt med hjälp av AWS. De viktigaste målen har varit att skapa ett system som är så snabbt som möjligt och att analysera vilken del av analysen som kan göras i molnet och vilken del som måste göras lokalt. Den förväntade nyttan med systemet är de minskade utvecklings- och underhållskostnader som kommer av att placera funktionalitet i molnet.
Examensarbetet har utgått ifrån tre användarfall inom truckhantering och använt simulerade sensordata från en trucksimulering i Node-RED som indata som ska analyseras. Det första användarfallet handlar om att förhindra att truckar tippar genom att analysera gafflarnas höjd och
lutning. För att förhindra att trucken tippar måste systemet varna när lutningen blir för hög eller lasten lyfts för högt. Om lutningen överstiger en viss gräns ska systemet skicka en instruktion till trucken som sänker gafflarna tills den stabiliseras.
Det andra användarfallet handlar om att hålla koll på truckars tidsanvändning för att kunna informera kunden som hyr truckarna om hur mycket de har använt av sin hyrda tid. Syftet är bl.a. att kunden ska kunna prioritera vilka truckar som ska vara igång och vilka som ska stängas av när den märker att den hyrda tiden håller på att ta slut.
Det tredje användarfallet bygger på att en truckuthyrare hyr ut truckar till kunder vid behov och transporterar dem mellan olika lager i en industriell zon. När en truck har arbetat färdigt för en kund ska systemet omfördela den till en annan kund så snabbt som möjligt. Ju snabbare omfördelningen kan ske desto högre blir truckarnas utnyttjandegrad och kunden kan utnyttja mer av sina hyrda tid.
Resultatet blev ett system som hanterar alla tre användarfallen. För att hantera användarfall 1 skickar truckarna regelbundet telemetri som analyseras av en lambda-funktion på en GGC (GGkärna).
Om trucken är nära att tippa skickas en instruktion tillbaka till trucken. Kommunikation sker via MQTT.
För att hantera användarfall 2 skickar truckarna ut statusuppdateringar som berättar att de har arbetat en viss tid. Uppdateringarna föraggregeras på GG-kärnan och skickas sedan vidare till en KDA-applikation som har integrerats med Apache Flink. Applikationen adderar tidsåtgången till en tidspott som är unik för varje kund.
För att hantera användarfall 3 körs en java-applikation på en EC2:a som analyserar
bokningar och utloggade truckars ID-nummer. Varje gång truckar loggar ut försöker systemet omfördela dem till en bokning som vill ha dem. Utloggade truckars ID-nummer skickas från GGkärnan till applikation och instruktioner som sätter truckar i arbete skickas från applikationen till GG-kärnan. (Less) - Abstract
- The purpose of the degree project has been to develop and evaluate a real-time analysis system which performs analysis in the cloud and locally using AWS. The most important goals have been to create a system which is as fast as possible and to reflect over which part of the analysis can be made in the cloud and which part has to be made local. The expected utility of the system is a consequence of the lowered development and maintenance cost associated with placing functionality in the cloud.
The degree project has focused on three use cases related to forklift management, and used simulated fork-lifter data from a forklift simulation in Node-RED as data to be analyzed. The first use case is about preventing fork-lifters from tipping... (More) - The purpose of the degree project has been to develop and evaluate a real-time analysis system which performs analysis in the cloud and locally using AWS. The most important goals have been to create a system which is as fast as possible and to reflect over which part of the analysis can be made in the cloud and which part has to be made local. The expected utility of the system is a consequence of the lowered development and maintenance cost associated with placing functionality in the cloud.
The degree project has focused on three use cases related to forklift management, and used simulated fork-lifter data from a forklift simulation in Node-RED as data to be analyzed. The first use case is about preventing fork-lifters from tipping by analyzing forks height and incline. To prevent the fork-lifter from tipping the system must send a warning when the incline becomes to high, or the load is lifted to high. If the incline exceeds a certain threshold, the system shall send an
instruction to the fork-lifter which in response lowers its forks until it is stabilized.
The other use case is about keeping track of fork-lifters time usage, to be able to inform the customer renting the fork-lifters how much time has been used. The purpose of this is among other things that the customer shall be able to prioritize which fork-lifters are to be kept on, and which ones should be turned off, once it notices the time rented is running out.
The third use case is based on a fork-lift rental company renting out fork-lifters to customers on demand, and transporting them between different warehouses inside an industrial zone. Once a fork-lift is done working for a customer the system shall reallocate it to another customer as fast as possible. The faster the reallocation the higher the fork-lift usage and the next customer can utilize
more of its rented time.
The result of the degree project is a system which handles all three use cases. To handle use case 1, the fork-lifters regularly send telemetry which is analyzed by a lambda function on a Greengrass core. If a fork-lift is close to tipping, an instruction is send back to the fork-lift. The communication is done using MQTT.
To handle use case 2, the fork-lifters send status updates which indicate that they have worked a certain amount of time. The updates are pre-aggregated on the Greengrass core and then send along to a KDA application which has been integrated with Apache Flink. The application adds the time usage to a time pot that is unique for each customer.
To handle use case 3, a custom made java application runs on a EC2, analyzing bookings and logged out fork-lifters ID numbers. Every time fork-lifters are done working the system tries to reallocate them to another customer. The ID numbers of logged out fork-lifters are sent from the Greengrass core to the application, and instructions putting the fork-lifters to work are sent from the application to the Greengrass core. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9032256
- author
- Ahlgren, Måns LU
- supervisor
- organization
- course
- EIEL05 20201
- year
- 2020
- type
- M2 - Bachelor Degree
- subject
- keywords
- AWS, realtid, real-time, Greengrass, Kinesis, Flink
- publication/series
- CODEN:LUTEDX/TEIE
- report number
- 3092
- language
- Swedish
- id
- 9032256
- date added to LUP
- 2025-02-05 17:23:17
- date last changed
- 2025-02-05 17:23:17
@misc{9032256,
abstract = {{The purpose of the degree project has been to develop and evaluate a real-time analysis system which performs analysis in the cloud and locally using AWS. The most important goals have been to create a system which is as fast as possible and to reflect over which part of the analysis can be made in the cloud and which part has to be made local. The expected utility of the system is a consequence of the lowered development and maintenance cost associated with placing functionality in the cloud.
The degree project has focused on three use cases related to forklift management, and used simulated fork-lifter data from a forklift simulation in Node-RED as data to be analyzed. The first use case is about preventing fork-lifters from tipping by analyzing forks height and incline. To prevent the fork-lifter from tipping the system must send a warning when the incline becomes to high, or the load is lifted to high. If the incline exceeds a certain threshold, the system shall send an
instruction to the fork-lifter which in response lowers its forks until it is stabilized.
The other use case is about keeping track of fork-lifters time usage, to be able to inform the customer renting the fork-lifters how much time has been used. The purpose of this is among other things that the customer shall be able to prioritize which fork-lifters are to be kept on, and which ones should be turned off, once it notices the time rented is running out.
The third use case is based on a fork-lift rental company renting out fork-lifters to customers on demand, and transporting them between different warehouses inside an industrial zone. Once a fork-lift is done working for a customer the system shall reallocate it to another customer as fast as possible. The faster the reallocation the higher the fork-lift usage and the next customer can utilize
more of its rented time.
The result of the degree project is a system which handles all three use cases. To handle use case 1, the fork-lifters regularly send telemetry which is analyzed by a lambda function on a Greengrass core. If a fork-lift is close to tipping, an instruction is send back to the fork-lift. The communication is done using MQTT.
To handle use case 2, the fork-lifters send status updates which indicate that they have worked a certain amount of time. The updates are pre-aggregated on the Greengrass core and then send along to a KDA application which has been integrated with Apache Flink. The application adds the time usage to a time pot that is unique for each customer.
To handle use case 3, a custom made java application runs on a EC2, analyzing bookings and logged out fork-lifters ID numbers. Every time fork-lifters are done working the system tries to reallocate them to another customer. The ID numbers of logged out fork-lifters are sent from the Greengrass core to the application, and instructions putting the fork-lifters to work are sent from the application to the Greengrass core.}},
author = {{Ahlgren, Måns}},
language = {{swe}},
note = {{Student Paper}},
series = {{CODEN:LUTEDX/TEIE}},
title = {{Realtidsanalys av simulerade sensordata från truckar med hjälp av AWS}},
year = {{2020}},
}