Ground Vehicle Systems Engineering & Technology Symposium 2009

Infinite Prognostics & Diagnostics: A System Architecture to Support P/D Algorithms Before They Even Exist

A combination of real world experience and new research initiatives will open up the universe of prognostic and diagnostic algorithms that can be created in the future. This presents the challenge of creating a system architecture that enables effective support of an infinite set of future algorithms even before they have been conceived, designed, implemented, tested, and approved for use. The Arbor architecture enables five critical elements to meet this challenge: (1) clean integration between legacy and new software, (2) remote, over the air provisioning of algorithms, (3) flexible data structures capable of evolving, (4) control points for the algorithm to report findings to in-vehicle occupants, and (5) a data collection strategy for failure incident reporting. Many algorithms are impossible to develop until we collect real world performance and failure information from on the vehicle. The Arbor system collects this information and feeds it off-board for analysis. Researchers analyze the data and develop diagnostic or prognostic algorithms that can then be deployed to a single vehicle experiencing odd behaviors or to an entire fleet, preemptively. A prognostic algorithm written, or modified, as an Arbor application can define its own outputs, which are then visible to the vehicle operator. These same outputs can be broadcast to a service technician with a diagnostic scan tool or to a remote operational command site, contingent on available communications links. Effective deployment of prognostic algorithms enables costly failures to be predicted ahead of time, thereby improving safety, reducing costs, and minimizing down time for equipment in order to effect more efficient fleet operations.

Get the papers!

Infinite Prognostics & Diagnostics (1.9 MB)
Presentation Slides (1.5 MB)

Bringing Best Practices from Web Development Into the Vehicle

Building embedded systems is nothing like building desktop applications, as the hard real time requirements and relative harshness of the operating environment further constrains design choices to meet real world needs. Those familiar with mainframe or server farm hosted, high volume, wide bandwidth applications know similar harsh computing environments for application development. Given that more man-hours have been devoted to web application development over the past decade than have been devoted to embedded application development, there may be some valuable lessons to be learned that can be adopted by the embedded community for in-vehicle computing. The best web application development teams successfully apply the notions of Representational State Transformation (REST) and Resource Description Framework (RDF) to handle the increasing demands on their sites. We have taken these technologies and applied them to the smaller scale vehicle telematics platforms (PowerPC, ARM, and Atom) to test their viability. This paper describes how we approached the design decisions that enabled us to successfully wrap a commercial J1939 CAN bus with a miniature web server that provides a REST API for applications to interact with an engine control unit. The architecture has been successfully deployed for custom mining and construction equipment.

Get the papers!

Lessons From The Web (324 kB)