SAMURAI: A Streaming Architecture for Mobile and Ubiquitous RESTful Analysis and Intelligence

SAMURAI is a scalable event-based stream mining architecture that integrates and exposes well-known software building blocks for event processing (Esper), machine learning (Weka) and knowledge representation (Parliament) as RESTful services.
  • Tags: , , , , ,
  • Latest update: October 24, 2014
  • Developed by: iMInds-DistriNet, KU Leuven in the scope of BUTLER project
  • Contact name: Davy Preuveneers
  • Contact email: [email protected]
  • Relations:
    • BUTLER Localization Manager SmartServer (more details)
      • Relation type: Interoperability relationship
      • Validation: The localization manager sends the geo-localization (x,y co-ordinates) to SAMURAI to enrich the raw positions into into semantic locations (e.g. Home, Living_Room, etc.) and semantically link locations with activities (e.g., Watching_TV with Living_Room). Although the standard approach is to access the data through the Context Manager, we have demonstrated the inter-operability between these smart servers and validated them in various BUTLER proof-of-concepts and
      • Completeness of the relation: 100%
    • BUTLER Smart Office Trial (more details)
      • Relation type: Re-use relationship
      • Validation: In the inno deployment, the User Behavior Smart Server is used to analyze the data it receives in order to determine the meeting room availability based on multiple factors. Only the remote Smart Service is used in the Smart Office trial at inno. The User Behavior Server is an essential part of the Smart Office scenario as it creates a richer experience to all end users.
      • Completeness of the relation: 33%
      • BUTLER Smart Health Trial (more details)
        • Relation type: Re-use relationship
        • Validation: When participants are indoor, the coordinates are also sent to the User Behavior manager. As an answer, Tecnalia’s OSGI framework receives a semantic location of the position, i.e., the name of the room where the participant is currently standing. This semantic location is shown on the smart mobile application, and it is more comprehensible than the raw position coordinates. The reliability of the Behavior Manager was very high during the trials, considering that the semantic location had to be obtained immediately in order to be handled by the Smart Mobile application, so that intelligible information about the participant location could be read on the screen.
        • Completeness of the relation: 100%

    Intellectual property rights (IPR)

    Copyright © 2012-2013 Davy Preuveneers, iMinds-DistriNet, KU Leuven, Belgium

Key features of SAMURAI include

  • Feature extraction: Convert raw low-level data and events into features that are more meaningful for comparison
  • Information fusion: Aggregate data from different sources to increase the confidence in the quality of the inferred information
  • Domain knowledge: Leverage background information to narrow down likely activities
  • Probabilistic correlations: Identify frequent co-occurrences in event streams to derive event patterns of interest

The overall flow would be to (1) define complex patterns of events with the Esper query language, (2) wrap semantic and machine learning classification as custom Esper operators, (3) use complex events for training and classification purposes and (4) add statements and listeners to subscribe to complex events of interest.

For full specification of SAMURAI with the latest updates, please refer to https://butler.cs.kuleuven.be/samurai/

Service level

SAMURAI will be continuously updated with new research insights and software extensions. The framework will be further developed and maintained in the frame of ongoing and new research projects. Hosting of the service will be guaranteed on a best effort basis.  For any support or enquiries, please send an e-mail to the contact person in the general information page.

Technology Readiness Level

7 - system prototype demonstration in operational environment

Reuse Readiness Levels

4 - Reuse is possible; the software might be reused by most users with some effort, cost, and risk.

Security

The current version supports state-of-practice security solutions such as SSL for communication, and while not mandatory it also offers support for encrypted storage. Additionally, the framework is integrated with Gemalto's Trust Manager, which was also developed in the frame of the FP7 BUTLER project.