Evolving Computers from Tools to Partners in Cyber-Physical System Design

Program seeks to generate AI-enabled tools to streamline design of cyber-physical systems while uncovering more innovative design approaches

Department of Defense (DOD) systems and platforms are composed of numerous integrated cyber-physical subsystems, which create an enormous amount of complexity and makes their engineering a daunting task. Today, designing cyber-physical systems (CPS) requires an army of skilled engineers with the right domain expertise, and hundreds of domain-specific tools. The process used to design these systems is largely manual, creating long design cycles that often result in costly redesigns after building and testing the systems. The flaws in the process are numerous – from balancing predictability with cost-efficiency to operating under tight time constraints to integrating disparate pieces from multiple design teams. Further, teams are often limited to focusing on known design approaches, restricting their ability to create or identify more sophisticated system alternatives or innovative concepts.

“Current approaches to designing cyber-physical systems are largely manual, costly, and inefficient, sometimes taking decades to complete,” said Dr. Sandeep Neema, a program manager in DARPA’s Information Innovation Office (I2O). “Within the defense community, there is increased demand for mission-specific system design that could be stood up quickly. Similarly, updating and advancing large legacy platforms remains a challenge. To support the rapidly evolving defense landscape, we need a way of accelerating and streamlining CPS design – one that could take advantage of new machine learning and automation capabilities.”

DARPA created the Symbiotic Design for Cyber Physical Systems (Symbiotic Design) program to bring intelligent automation into the CPS design process. The program will explore fusing model-based design with machine reasoning and learning to develop a core set of AI-enabled tools that designers can use to accelerate the process from design concept to developed system. The tools will support the search, composition, evaluation, and exploration of knowledge bases and design corpora, and will come together to form an AI-enabled co-designer that provides its human counterparts with a true design partner.

Design corpora already exist but combing through them to find the best options for a set of requirements is an arduous task. Under the Symbiotic Design program, researchers will endeavor to create AI-enabled tools that can mine these knowledge bases to rapidly uncover more sophisticated design options and approaches. The intent is for the tools to propose sets of alternatives to construct a multi-domain design space for the design problem at hand. With the design options identified, the AI tools will also help determine the best way to pick candidate subsystems, and integrate them into a coherent design while identifying any potential gaps to address.

Key to the AI tools will be their ability to observe and learn from past successful designs and design corpora. Ultimately, the exchange of ideas between humans and the AI-enabled co-designer will allow the system to “autocomplete” the design based on past learnings or experience. Once the human designer has selected a design candidate, the AI co-designer will automate the evaluation of the design points using relevant domain-specific analysis and simulation tools.

“While existing approaches to design automation have brought some benefits to the process, they are largely utilizing first wave AI, or rule-based systems. These systems are limited in that they rely on human expertise and an ever-evolving list of ‘if-then’ rules. The Symbiotic Design program seeks to employ second and third wave AI, combining learning with advanced search techniques to help synthesize highly sophisticated designs,” noted Neema.

In addition to developing the tools that will form the AI-enabled co-designer, the Symbiotic Design program seeks to create user-friendly interfaces to facilitate human-machine collaborations. Using automation technologies and human-machine interfaces, the program aims to make the design process accessible to more individuals by reducing the need for specialized skillsets, as well as augmenting the skills of experts.

The Symbiotic Design program is a part of DARPA’s AI Next campaign – a multi-year, $2 billion investment into new and existing programs focused on the development and application of “Third Wave” AI technologies. DARPA views the Third Wave of AI as the development of systems that are capable of acquiring new knowledge through generative contextual and explanatory models.

Source: DARPA
Date: Aug 2, 2019