Reconnaissance and IoT Data Processing with the Palantir Foundry: Walkthrough and Sample implementation of an MVP with the Palantir Foundry Ontology - Softcover
Inhaltsangabe
Reconnaissance – military or civil – has an interesting similarity to processing IoT (Inter-net of Things) data. In both cases data of different levels of semantic processing are fed into a system to generate a digital twin of the reality.
This book has two main goals: The first goal is to show an application architecture based on the Palantir Foundry for the implementation of an insight-creation and decision-making system. The second goal is to describe a generic data- and processmodel for a Digital Twin of a given subject area. The reconnaissance area is taken as the use-case.
The design and implementation process runs through iterations. At the end of the last iteration an MVP (Minimum Viable Product) is implemented that allows scaling to a real world deployment.
Data architecture "Digital Twin"
The dynamic part of a digital twin is typically built around the processing of all "events" in a given area/factory or other context. A fully qualified "event" answers the W-Questions of what happened:
- Where did it happen?
- When were the start and the end?
- Who or what was acting?
- What was the kind of activity?
Having fully or partly qualified events, higher level questions such as “Why did it happen?”, “Is it ‘normal’?”, and “Will something follow?” might be answered.
The data core of this digital twin is an “Ontology of data” and objects around the “event”. An “Ontology of data” means that each object of the real world (an actor and activity or an intention) has only one representation in the data.
The range of the qualification of the incoming data ranges from “None”, i.e. the data is raw sensor data with a timestamp but has no identified relation to the assets of the digital twin, to fully qualified data where the only aspect of processing it is mapping the names in the source system to the names of the objects in the ontology.
Application Architecture with the Foundry
The Palantir Foundry is a excellent choice for the implementation of a system to cover the requirements:
- The pipeline-based data processing fits perfectly into the idea of creating and transforming incoming data to a central repository of events.
- The Foundry Ontology and the ability to edit objects provide the means to enrich incoming data as well as the possibility to create new insights by operators.
- The basic means of the Foundry Ontology and its Object Explorer are used to create a backbone to work with the objects in an explorative way and to perform all actions on the objects with a common user interface.
- To implement the “business logic” functions, i.e. functions that create new insights or events, the Foundry Contour, Code-Workbook or Code-Repositories can be used.
- The need skillset for implementing the business logic starts with Contour at a level comparable to Excel sheet creation and scales seamlessly to AI programming with Python or Java
- With Workshop, convenient dashboards and workflow applications can be implemented.
- The Alerting use case and the Taurus rule-engine is implemented to manage business rules for alerting
- Using Webhooks, the generated insights and decisions can be propagated to connected systems
The approach is not to go into depth on a given functionality but rather go into the width and show how an application with very different functional parts can be assembled using and combining various elements of the Palantir Foundry.
The expected skill level of the reader of this book is twofold: To understand the strategy of the implementation, a general understanding of programming and databases is needed.
To follow each detail of the implementation, advanced Foundry Know-how is needed.
Die Inhaltsangabe kann sich auf eine andere Ausgabe dieses Titels beziehen.