Data Fusion of Probe and Point Sensor Data: A Guide (2024)
Chapter: 7 Additional Fusion Resources
CHAPTER 7
Additional Fusion Resources
A quick search for data fusion techniques will return more results than any professional could hope to read and digest. Many data fusion books, papers, and other resources may be geared toward complex military applications. It can be difficult to sort through what is and is not relevant. Furthermore, there are more and more readily available ML models that are freely accessible from the web. Many developers simply apply these pre-existing models to their own datasets, train them, and then compare several different models to see which ones work the best for their specific application. This “try and see” type of approach is common for researchers and developers in this field and should not be discouraged.
For those looking for broadly applicable use cases, mathematical models, and other fusion algorithms, the following resources may be of support for implementers who are working to fuse point sensor and probe data. This is not an exhaustive list, and it does not include use case-specific papers like those discussed in Chapter 6.
-
Klein, L. (2019) Sensor and Data Fusion for Intelligent Transportation Systems. SPIE, Bellingham, WA.
This book can serve as a reference manual for ITS data fusion implementers. The scope is much larger than just point sensor and probe data; however, it is custom tailored toward ITS applications and data. A second book by the same author, titled Sensor and Data Fusion: A Tool for Information Assessment and Decision Making, Second Edition, is also available, though it is not tailored specifically to ITS applications. There is plenty of useful information specifically for system implementers and those wishing to dive deeper into the details and mathematics of data fusion.
-
Framework to Support Data Fusion, Analysis, and Decision-Making. Federal Highway Administration, U.S. Department of Transportation, Washington, DC (forthcoming).
This FHWA report is expected to be published soon and will provide broad guidance on the fusion of disparate transportation data. Unlike this NCHRP report, which focuses solely on probe-based speed data and point-sensor data, the FHWA report expands to all manner of transportation-related data including, at a high level, events and incidents, traffic performance, traffic signals, weather, CVs, trajectories, and transit data.
- Cocchi, M. (ed.) (2019). Data Fusion Methodology and Applications, 1st Edition. Elsevier, Amsterdam, The Netherlands.
This book has detailed mathematical models and three chapters that act as tutorials for implementing several different fusion methods and models. While this book is not tailored to transportation applications, there are still many relevant examples.
-
Ameisen, E. (2020). Building Machine Learning Powered Applications. O’Reilly Media, Sebastopol, CA.
This publication “demonstrates practical ML concepts using code snippets, illustrations, screenshots, and interviews with industry leaders. Part I teaches you how to plan an ML
- Winder, P. (2020). Reinforcement Learning. O’Reilly Media, Sebastopol, CA.
This book covers how reinforcement learning (RL), a subset of ML, can be used to program software to learn from its environment. The book teaches specific RL algorithms and fundamentals and how to deploy RL to production. It includes topics such as meta learning, hierarchical learning, imitation learning, and multi-agent learning, and specific algorithms including Soft Actor Critic, Proximal Policy Optimization, Twin Delayed Deep Deterministic Policy Gradient, and Rainbow.
- Awesome Production Machine Learning. Institute for Ethical AI and Machine Learning, Berlin (n.d.).
This GitHub repository includes a curated list of open-source libraries that may help with the deployment, monitoring, versioning, scaling, and securing of a use case-specific ML data fusion application. It is available at https://github.com/EthicalML/awesome-production-machine-learning and includes information such as:
- Explaining Predictions and Model
- Privacy Preserving ML
- Model and Data Versioning
- Model Training Orchestration
- Model Serving and Monitoring
- AutoML
- Data Pipeline
- Data Labeling
- Metadata Management
- Computation Distribution
- Model Serialization
- Optimized Computation
- Data Stream Processing
- Outlier and Anomaly Detection
- Feature Store
- Adversarial Robustness
- Data Storage Optimization
- Data Science Notebook
- Neural Search
application and measure success. Part II explains how to build a working ML model. Part III demonstrates ways to improve the model until it fulfills your original vision. Part IV covers deployment and monitoring strategies” (Ameisen 2020).
