Designing for Grand Challenges: An Interactive Method for Tackling Coupled Device-System Designs

Abstract When designing a system of devices to accomplish a large task, there is an inherent relationship between the capability of each device and the overall system effectiveness. If one is designing the devices from scratch, the engineering specifications depend on a budget and hence the number of devices in the fleet. Determining these engineering specifications depends on some unknown system strategy where the devices must cooperate to accomplish the overall goals. For grand challenges like ocean cleanup or forest fire fighting, the coupled problem of system effectiveness and device specifications can be impossible to determine. One may specify the capabilities of a device which may not be achievable by a design team or may be prohibitively expensive. To overcome this problem, we present an approach that guides an optimization process; one that can both simulate the fleets as well as incorporate designer feedback on design feasibility. This research proposes an interactive framework to address these issues by using Agent-Based Modeling to simulate fleets, utilizing Gaussian Process models to capture user feedback on feasibility and applying constrained Bayesian optimization to guide the design process. This method was tested on two case studies: a simplified ocean cleanup fleet and a higher-fidelity model of firefighting drones. This study provides insight into how performance parameters can be managed simultaneously at the device-level and the system-level.