How do Autonomous Vehicles use Digital Engineering?

In this post we are going to discover what Digital Engineering is, why it is so practical in today’s world of “Internet of Things,” “Software Defined Vehicles,” and mission-critical applications that are exploding in new industries such as space exploration and drone transportation.

What is Digital Engineering?

Digital Engineering stems from the benefits of systems engineering and Model-Based Design, which is the use of mathematical models to simulate and design complex control systems, physical responses, communication flow, and functionality within embedded software.

Digital Engineering goes beyond Model-Based Design to stretch across two additional directions: behavioral requirements models, and big-data analysis and field-monitoring. Furthermore, Digital Engineering is not just limited to software or control theoretic domains like Model-Based Design, but encompasses all relevant domains including mechanical and human cognitive behavioral models.

In short, Digital Engineering holistically combines these three main well-established disciplines:

  1. Model-Based Design and Systems Engineering

  2. Behavioral Modeling and Formal Requirements

  3. Big-Data Analysis and DevSecOps field monitoring

These are combined in order to maintain a trusted “Authoritative Source of Truth,” that houses the systematic relationships of the models of the system, along with the accompanying probabilistic validation of the parameters in the model, traceable to the data and the various sources from which the data came.

The “Authoritative Source of Truth” is then shared across stakeholders to access the latest and greatest “state of the art” understanding of the whole system or segmented sub-system. This access accelerates and expands optimization activities across organizations, and helps partners develop robustness in their deployed systems.

Where did Digital Engineering come from?

The US Department of Defense published the Digital Engineering Strategy in June 2018 in which it stated, “The DoD vision for digital engineering is to modernize how the Department designs, develops, delivers, operates, and sustains systems. DoD defines digital engineering as an integrated digital approach that uses authoritative sources of system data and models as a continuum across disciplines to support lifecycle activities from concept through disposal.”

From the publication of that strategy, defense contractors, researchers, and associated businesses in aviation and aerospace have been adopting and promoting Digital Engineering best practices.

Similarly, the aviation industry has championed the “Safety Management System” which relies on timely data collection and information sharing to the decision-makers in order to do early detection of problems that need to be fixed in the field or in design.

For example, anomalous flight performance data can be collected across multiple airlines in order to detect an early sign that pilots need to be trained differently, more maintenance needs to be performed, or a manufacturer recall needs to be initiated. The investment and build-up of Digital Engineering assets over time could be leveraged within a SMS to quickly identify the likely causes of data anomalies, and also to explore robust mitigation measures to address the behavior.

Why is Digital Engineering critical to Autonomous Safety Management Systems?

Autonomous surface vehicles, or Automated Driving Systems, must perform complex behavioral requirements in high-risk situations. For example, an autonomous vehicle may have to stop in the presence of an unaccompanied child in the street in one scenario, but in another scenario it may have to continue to drive as it squeezes between a pedestrian standing at the curb on one side, and on-coming vehicle approaching on the other side.

There is no hard and fast rule in autonomous vehicle behavior that says: “Stay 2 m away from a pedestrian,” that works every time. Rather complex behaviors must be defined for sufficiently nuanced conditions. This is just as challenging to define offline, in a requirements model in a Digital Engineering environment, as it is to define in the embedded software in the vehicle, itself.

The product liability risks associated with a “miscalculation” by the autonomous vehicle and their organizations are arguably higher than aviation and aerospace industries, given the close proximity to hazards on public roads. Leveraging the most amount of data from test vehicles, government studies, research institutions, and competitors is essential for ADS developers to have confidence in their safety-critical behavioral models. This is the perfect use case for Digital Engineering.

How would this work?

Digital Engineering could support all facets of autonomous vehicle development and deployment, from business cases, to adding services, and designing new vehicles.

For example, one camera perception company could get access to “Key Performance Indicators” (KPI) of how their units perform in the field. Let’s say they want to monitor how well their cameras dynamically adjust to maximize the contrast of an on road target from the background. They could detect an overall reduction in performance, causing them to redesign their product or inform their customers of the expected performance limitations in the real-world. Or, they could detect that all units work as expected except for one particular customer. They could then talk with that customer to see what may be causing the unexpected degradations, such as where they are placing the camera, what glass is in front of it, etc.

The camera perception company may have to provide their customers with the KPI algorithms, and get permission to use an over-the-air (OTA) gateway to collect the data. The camera perception company may have challenges with different customers having different types of data streams they would need to collect, but in exchange they could offer greater guarantees and warranties on performance over the life of their product.

Closing Thoughts

The challenge with setting up Digital Engineering infrastructure is getting the stakeholder’s permission to collect potentially sensitive data and to power those processes for calculating the KPI algorithms “on the Edge,” and then piping that data back to the organization’s storage. The reward for investing in these capabilities, however, is huge, as the very products that are developed can become the testbeds and engines for future improvements and commercial opportunities.

That is exactly one of the goals of Digital Engineering, and it is the goal of Retrospect to help establish consensus among AV safety stakeholders on the validation of fundamentally safe driving behaviors. Retrospect’s RiskEngineTM safety monitor is a critical “edge” KPI algorithm and behavioral model that measures how well any ADS anticipates and mitigates reasonably foreseeable risks on the road, in a universal format.

As autonomous vehicle companies prepare their commercial deployments, they should ensure they have proven their “Safety Management Systems” and “Digital Engineering” assets to defend themselves against liability, especially with 3rd-party validators. They will also benefit from taking additional time to build up the knowledge-sharing ecosystem that is Digital Engineering to continuously improve while cutting costs and increasing services. If you would like to further review your organization’s Digital Engineering assets and strategy, please contact us here, or leave a comment below. We look forward to hearing your thoughts on the opportunities and challenges with Digital Engineering!

Michael WoonComment