JARVIS banks on a symbiosis between humans and AI to serve two functions. First, it helps system designers explore the space of the possible to find innovative new designs. Using the tool’s effective human-machine interface (HMI), designers can provide feedback to the AI to guide its exploration of the design space. Second, JARVIS helps designers refine specific components in new or existing designs to fine-tune each design’s ability to meet requirements. The tool provides a way for designers to interact with and understand the results of AI.

The advantage of using AI‑human symbiosis as the basis for the tool is that it harnesses the best of both. AI can churn through millions of design candidates and conduct early performance assessments to nudge designers toward new and potentially promising design directions. On the other hand, humans are good at understanding the limitations and practicality of the designs and using intuition to guide the design process. Working together enables wider and more optimized choices.

While traditional design follows a linear iterative path, JARVIS lets designers explore designs, then go back to previous designs and try different choices, and even merge great ideas from two streams of exploration. JARVIS enables optimization for any relevant parameter, whether that be material cost, speed and maneuverability of a vehicle, weight, or power consumption. The tool is agnostic to the specific domain of the design, so it is equally suited to exploring designs for air vehicles, underwater vehicles, or other systems.

Early use cases of JARVIS for DARPA explored the efficient “correct-by-construction” design of cyber-physical systems like aerial and underwater vehicles. In aerial vehicles, for example, JARVIS allowed design teams to explore what kinds of designs would deliver a sharper turning radius or faster acceleration. The correct-by-construction approach ensures that design candidates are put through their paces, so the final choice is sure to meet required criteria.