Target groups:First responders in fire fighting and flood oeprations, officers in maritime operations

Benefit: Real-time processing and visualization of data involving critical information with embedded visualization of future forecasting as well as uncertainty, for first responders on the go

AR Visualization and Server: An innovative head-worn AR flood management system designed to support situational awareness and way-finding by enhancing decision-making for emergency responders on-the-go. 

AR for Fire Operations: An innovative head-worn AR system designed to enhance situational awareness for firefighting team leaders. 

Eye Tracking Prediction System for Mental Fatigue of First Responders: A low-cost, custom-built eye tracking system designed as a modular add-on for AR headsets. 

AR Maritime Visualization: An innovative head-worn AR navigation and event-forecasting system for maritime operations, emphasizing situational awareness, real-time data visualization, and robust calibration across variable ship infrastructures. 

Target groups: Data scientists and AI engineers, IoT solution providers, System integrators, Industrial AI platform vendors, Research and innovation organizations

Benefit:Simplifies deployment of distributed AI pipelines, enables low-latency analytics, reduces integration and maintenance costs, improves scalability and oeprational efficiency

The Edge-to-cloud extension extends and modifies the RapidMiner Streaming Extension, which was originally created by RapidMiner GmbH. This extension provides operators that allow RapidMiner users to design and automatically deploy arbitrary jobs that can operate either at Apache Flink Big Data clusters, or to edge nodes. Streaming operators will be deployed to Apache Flink and non-streaming jobs to edge devices running RTS Agents.

Target groups: Business experts, AI workflow designers, cloud & edge infrastructure operators, IoT platform providers, data-intensive service providers, digital transformation consultancies 

Benefit: Improves execution efficiency and responsiveness, reduces deployment and tuning effort, enhances system robustness and scalability, supports adaptive and dynamic infrastructures

The Edge-to-Cloud Optimization component is responsible for: 1) Receiving RapidMiner workflow definitions and infrastructure topologies via REST API, 2) Performing the optimization using an algorithm described in Deliverable D4.2 , 3) Returning the optimized workflow via WebSocket to the AI Studio front-end, 4)Saving the optimized workflow to elastic for usage outside of AI Studio

Target groups: Researchers, industry

Benefit: Academic curiosity, early capitalization of opportunities

Wayeb is a Complex Event Processing and Forecasting (CEP/F) engine written in Scala. It is based on symbolic automata and full- or variable-order Markov models. RTCEF is an extended version of Wayeb with the ability to adapt its forecasting models to concept drifts.

The source code of the system components for Explainable Artificial Intelligence (XAI) and their integration into the overall CREXDATA system.

Target groups: Researchers, students, system integrators

Benefit: End-to-end, optimized training and deployment of Federated Learning processes on real network infrastructures

Implementation of the Functional Dynamic Averaging (FDA) state-of-the-art meta-algorithm for extremely low-communication Federated Learning. The plain FDA includes explicit tuning for hyperparameters, suitable for cross-silo training. FedOPT supports auto-tuning and is particularly efficient in cross-device settings.

The suite also includes a module for rapid simulation of communication latency and throughput for Federated Learning training deployments on edge devices, using the NS3 simulator. The simulations can guide the low-cost deployment of federated learning processes.

Target groups: Data Scientists & Analysts, IoT & Systems Architects, Operations Managers & Situational Awareness Leads

Benefit: Users can instantly join live sensor data (GPS pings) with static data (city zoning maps) without complex geometry-matching code, reduced latency resulting in faster decision-making, faster time-to-market thanks to AI studio extensions that can prototype fusion models and deploy them in the AI hub

Scalable, high-throughput architecture built upon Apache Kafka Streams, designed to overcome the complexity of fusing heterogeneous sensor data in real-time. Handles location through the H3 spatial index, which utilizes uniform adjacency and consistent surface area to ensure unbiased situational awareness and accurate movement modeling.This hierarchical indexing strategy enables the dynamic binding of live streaming data with static batch records. The system enhances generalization by combining a compatible workflow designer with a microservices-based execution model, allowing users to intuitively design complex fusion logic visually

Target groups: Researchers, students, system integrators

Benefit: Near real-time updating of urban Neural Radiance Fields (NERFs) with low-communication via Federated Learning

Scalable system for training, adapting and visualizing city-scale NeRFs under time-critical emergency conditions. The system supports rapid adaptation of pre-existing NeRFs by decomposing large scenes into spatially localized NeRFs that enable incremental updates as new aerial imagery becomes available, preserving previously learned structure and allowing for the reliable deployment of NeRF-based scene understanding in large real-world environments

Target groups: Application developers, system designers

Benefit: Quick discovery of optimal parameter values, configurable resolution of parameter values exploration, flexible integration into application workflows

A generic and robust optimization component that implements a Genetic Algorithm. The component can be configured to optimize parameter value vectors of arbitrary length (set by the user upon initiation), and utilizes custom user-defined objective functions (such as the outcomes of simulations).
In the CREXDATA emergency use-case this component is integrated to optimize the height of water barriers placed in various locations by emergency responders to render a point of interest safe.

The source code of additional RapidMiner Extensions for integrating additional data sources and CREXDATA system components.

Target groups: Robotics researchers, UAV engineers, decision-makers, policymakers, Epidemiological modelers, data scientists, public health researchers, industry (commercial shipping, traders), insureres, logistics

Benefit: Risk-free evaluation of drone behaviour under varying conditions, Enables reproducible, scalable epidemic simulations integrating mobility, demography and interventions, enables audtomated discovery of adaptive intervention strategies under uncertainty, improved ETA accuracy and supply chain reliability, ehnahnced desicion-making with scenario planning tools

The Weather Emergencies Use Case Simulator HyperSuite (EmCaseSHS) is designed as a set of interoperable and extendable operators for data processing workflows in RapidMiner AI Studio, and extended by a specific web application to enable easy configuration access for end users through a widget started from ARGOS. The EmCaseSHS enables the use of real-time data from the field, from current forecasts, or injected by the Data Injector in simulations.

The Health Emergency Use Case uses two simulators: Alya (models air to lungs, prefusion of the vascular and lymphatic vessels, transportaiton of the virus throughout the brochiole) and PhysiBoss (models different cell types of the tiessue, cell response to O2, state of alveolus).

The Maritime Use Case simulator fuses AIS data, historical mobility patterns, and NOAA/Copernicus forecasts. It providers multi-criteria routing (weather-optimized, traffic-optimized, shortest path) and a modular architecture with H3 grid and A* algorithm for scalable rerouting.

Target groups: Companies & organizations wanting to use Social Media Monitoring and Text Analytics in weather emergencies within Rapid-Miner workflows

Benefit: Users can efficiently detect social media posts sharing relevant and actionable information on ongoing weather emergencies, ability to answer targeted queries using the information provided in posts, e.g. “what are the affected areas?” 

System components for social media crawling (i.e., for crawling X), text analytics, and information extraction (e.g., answering relevant queries) from texts and their integration into the overall CREXDATA system  

Target groups: Domain experts in high-stakes areas (e.g., healthcare, epidemiology, policy, infrastructure), machine learning practitioners and model developers, model auditors & stakeholders responsible for model governance and regulatory compliance

Benefit: Systematic, logic-centered model auditing beyond accuracy metrics, increased transparency and calibrated trust in model behavior, early detection of logically flawed or unsafe rules, and support for domain knowledge discovery by revealing latent patterns in the learned decision logic

A visual analytics framework and software system for inspecting and validating the internal logic of rule-based machine learning models. Enables users to evaluate whether a model’s decision rules are consistent with domain knowledge, regulatory constraints, and expected reasoning patterns and whether the model is likely to behave sensivly on new, unseen data. 

Supports multi-level visual exploration of large rule sets and enables users to interactively filter and drill down by features, predicted outcomes and value ranges and apply computational analyses. The system works primarily on rule sets extracted from models and does not require access to original training data, while offering additional capabilities when labeled ground truth data are available.