This paper introduces the HyForm uncrewed vehicle engineering repository (HUVER), a comprehensive multi-modal dataset of uncrewed aerial vehicle (UAV) designs, complete with performance evaluations, derived from the HyForm UAV design testbed. The dataset includes 6051 unique UAV configurations, each represented using strings adhering to a designed grammar, images, 3D mesh models, and textual descriptions, alongside performance metrics obtained from physics-based simulations. Designed to support data-driven and artificial intelligence (AI)-driven design processes, one area in which this dataset can facilitate research is the surrogate modeling and generative design of UAVs, providing a resource for developing predictive models and supporting human-AI collaboration in UAV design. The dataset adheres to findable, accessible, interoperable, and reusable principles, ensuring it is retrievable, accessible, interoperable, and reusable, and is made available as an online repository for ease of use by the research community.

Hybrid powered locomotives can provide significant savings on the energy consumed to move freight on railways. The SCORE toolset is an open source, web-based application to assess the impact of new powering technologies on railroad performance, specifically technologies capable of both putting power into the rail (motive force) and taking power from the rail (regeneration). SCORE’s primary goal is generate trade studies to analyze different powering and train make-up options to minimize energy usage and greenhouse gas generation. At the core of generating these trade spaces is calculating the optimal powering policy for the train given the makeup of the train, the route, and time constraints. This paper presents details on the algorithms used in SCORE to compute this powering policies that is fast and accurate, discusses its implementation in an Energy-Longitudinal Train Dynamics (E-LTD) model, compares it to naïve approaches, and demonstrates its use across a variety of train/route pairs.

Exploring the opportunities for incorporating Artificial Intelligence (AI) to support team problem-solving has been the focus of intensive ongoing research. However, while the incorporation of such AI tools into human team problem-solving can improve team performance, it is still unclear what modality of AI integration will lead to a genuine human-AI partnership capable of mimicking the dynamic adaptability of humans. This work unites human designers with AI Partners as fellow team members who can both reactively and proactively collaborate in real-time toward solving a complex and evolving engineering problem. Team performance and problem-solving behaviors are examined using the HyForm collaborative research platform, which uses an online collaborative design environment that simulates a complex interdisciplinary design problem. The problem constraints are unexpectedly changed midway through problem-solving to simulate the nature of dynamically evolving engineering problems. This work shows that after the unexpected design constraints change, or shock, is introduced, human-AI hybrid teams perform similarly to human teams, demonstrating the capability of AI Partners to adapt to unexpected events. Nonetheless, hybrid teams do struggle more with coordination and communication after the shock is introduced. Overall, this work demonstrates that these AI design partners can participate as active partners within human teams during a large, complex task, showing promise for future integration in practice.

Adding battery locomotives to traditional diesel locomotives to form a hybrid consist has the potential to reduce fuel consumption and emissions in freight rail operations. This paper provides route-based estimates of diesel (hpd/ton) and battery (hpb/ton) power requirements and diesel (gal/ton) and battery (hpbhr/ton) energy requirements for hybrid consists. Given a route-specific power profile, the power split between the battery pack and the diesel engine is optimized to minimize fuel consumption by running the diesel engine at maximum efficiency during the entire route. Simulations for 200-mile round trips between Chicago and Harrisburg show that the diesel power is maximum in low-gradient regions at 1.05 (hpd/ton) and battery power is maximum in hilly regions at 2 (hpb/ton) for discharge and 6.5 (hpb/ton) for charge. Maximum fuel efficiency gains of approximately 60% are seen in the mountains, where 2.09 (hpbhr/ton) of battery energy is needed. Minimum fuel economy gains are observed in flat regions, with 20% fuel consumption reduction. With battery current and voltage limits, the battery pack size increases, and fuel savings decrease.

Exploring the opportunities for incorporating Artificial Intelligence (AI) to support team problem solving has been the focus of intensive ongoing research. However, while the incorporation of such AI tools into human team problem solving can improve team performance, it is still unclear what modality of AI integration will lead to a genuine human-AI partnership capable of mimicking the dynamic adaptability of humans. This work unites human designers with AI Partners as fellow team members who can both reactively and proactively collaborate in real-time towards solving a complex and evolving engineering problem. Team performance and problem-solving behaviors are examined using the HyForm collaborative research platform. The problem constraints are unexpectedly changed midway through problem solving to simulate the nature of dynamically evolving engineering problems. This work shows that after the shock is introduced, human-AI hybrid teams perform similarly to human teams, demonstrating the capability of AI Partners to adapt to unexpected events. Nonetheless, hybrid teams do struggle more with coordination and communication after the shock is introduced. Overall, this work demonstrates that these AI design Partners can participate as active partners within human teams during a large, complex task, showing promise for future integration in practice.

Adding battery locomotives to traditional diesel locomotives to form a hybrid consist has the potential to reduce fuel consumption and emissions in freight rail operations. This paper provides round trip, route-based estimates of diesel (hpd/ton) and battery (hpb/ton) power requirements and diesel (gal/ton) and battery (hpbhr/ton) energy requirements for hybrid consist. Given the route-specific power profile, the power split between the battery pack and the diesel engine is optimized to minimize fuel consumption by running the diesel engine at maximum efficiency during the entire trip. The results from 100-mile out/back routes between Chicago and Harrisburg indicate that the required diesel power for hybrid consist is maximum in low-gradient regions at 1.17 (hpd/ton) and the required battery power is maximum in hilly regions at 2.04 (hpb/ton) for discharge and 9.43 (hpb/ton) for charge. Maximum fuel efficiency gains of approximately 60% are seen in hilly routes, where 2.09 (hpbhr/ton) of battery energy is needed. Minimum fuel economy gains are observed in flat regions, with a 20% fuel consumption reduction. This ideal case analysis of hybrid consist can serve as an easy-to-use tool for the US freight rail sector to estimate fuel savings and sort routes based on the potential results for electrification, without requiring substantial computation.

Although necessary for complex problem solving, such as engineering design, team agility is often difficult to achieve in practice. The evolution of Artificial Intelligence (AI) affords unique opportunities for supporting team problem solving. While integrating assistive AI agents into human teams has at times improved team performance, it is still unclear if, how, and why AI affects team agility. A large-scale human experiment answers these questions, revealing that, with appropriately interfaced AIs, AI-assisted human teams enjoy improved coordination and communications, leading to better performance and adaptations to team disruptions, while devoting more effort to information handling and exploring the solution space more broadly. In sum, working with AI enables human team members to think more and act less. (c) 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-ncnd/4.0/).

Human subject experiments are performed to assess the impact of artificial intelligence (AI) agents on distributed human design teams and individual human designers. In the team experiment, participants in teams of six develop and operate a drone fleet to deliver parcels routed to multiple locations of a target market. Among the design teams in the experiment, half of the design teams are human-only teams with no available AI agent. The other half of the design teams, designated as hybrid teams, have drone design and operation AI agents to advise them. Halfway through the team experiment, team structure is changed unexpectedly, requiring participants to adapt to the change. In the individual experiment, participants develop drones based on given design specifications, either on their own or with the availability of a drone design AI agent to advise them. During these experiments, participants configure, test, and share their designs and communicate with their teammates through an online research platform. The platform collects a step-by-step log of the actions made by participants. This article contains data sets collected from 44 teams (264 participants) in the team experiment and 73 participants in the individual experiment. These data sets can be used for behavioral analysis, sequence-based analysis, and natural language processing.

Human subject experiments are often used in research efforts to understand human behavior in design. However, such research is often time-consuming, expensive, and limited in scope due to the need to experimentally control specific variables. This work develops an initial digital simulation of team-based multidisciplinary design, where the actions of individual team members are simulated using deep learning models trained on historical human design trends. The main benefit of this work is to simulate design session events and interactions without human participants, developing a complimentary method to rapidly perform digital team-based experiments. This research merges the benefits of purely data-driven modeling with minimal assumptions about process, along with the strengths of agent-based modeling in which it is possible to tailor agent behavior. Initial results show that the simulated design team sessions are able to replicate trends and distributions compared to human-based team sessions, but run approximately 21 times faster than equivalent human subject studies. The multi-disciplinary design problem currently simulated is loosely coupled, in the sense that agent behaviors can be modeled in isolation of other agents and yet replicate the behavior of the ensemble. Future work will extend the agents to sense and respond behaviors that can be used to model tightly coupled problems, and truly evaluate team formulations.

Human-computer hybrid teams can meet challenges in designing complex engineered systems. However, the understanding of interaction in the hybrid teams is lacking. We review the literature and identify four key attributes to construct design research platforms that support multi-phase design, hybrid teams, multiple design scenarios, and data logging. Then, we introduce a platform for unmanned aerial vehicle (UAV) design embodying these attributes. With the platform, experiments can be conducted to study how designers and intelligent computational agents interact, support, and impact each other.

As computer technology advances, graphical design environments (GDEs) and visualization tools to support engineering design and decision making are gaining prominence and recognition, particularly in the area of multiobjective design and optimization. In this paper, we discuss an experiment in two graduate courses that was designed to evaluate GDEs through inclass student assignments. For this first set of experiments, a GDE was developed for designing an I-beam cross section with two competing objectives. Within the GDE, students were allowed to vary the values of the design variables and view the corresponding performance graphically in an effort to obtain an optimal design based on a weighted sum of the objectives. Methods for evaluating student efficiency, effectiveness, and satisfaction within a GDE are discussed, and preliminary results from the experiment verify that graphical design environments can improve design quality and overall satisfaction with the design. The importance of rapid graphical feedback in a GDE is also investigated by incorporating time delays in the performance response. The use of graphical design environments to improve student understanding of design tradeoffs in the classroom is discussed, and results from the I-beam experiment are compared with a previous assignment wherein students had to choose an optimal design without the use of a graphical design interface.

The authors present preliminary results on successfully training a recurrent neural network to learn a spatial grammar embodied in a data set, and then generate new designs that comply with the grammar but are not from the data set, demonstrating generalized learning. For the test case, the data were created by first exercising generative context-free spatial grammar representing physical layouts that included infeasible designs due to geometric interferences and then removing the designs that violated geometric constraints, resulting in a data set from a design grammar that is of a higher complexity context-sensitive grammar. A character recurrent neural network (char-RNN) was trained on the positive remaining results. Analysis shows that the char-RNN was able to effectively learn the spatial grammar with high reliability, and for the given problem with tuned hyperparameters, having up to 98% success rate compared to a 62% success rate when randomly sampling the generative grammar. For a more complex problem where random sampling results in only 18% success, a trained char-RNN generated feasible solutions with an 89% success rate. Further, the char-RNN also generated designs differing from the training set at a rate of over 99%, showing generalized learning.

The authors present a generative adversarial network (GAN) model that demonstrates how to generate 3D models in their native format so that they can be either evaluated using complex simulation environments or realized using methods such as additive manufacturing. Once initially trained, the GAN can create additional training data itself by generating new designs, evaluating them in a physics-based virtual environment, and adding the high performing ones to the training set. A case study involving a GAN model that is initially trained on 4045 3D aircraft models is used for demonstration, where a training data set that has been updated with GAN-generated and evaluated designs results in enhanced model generation, in both the geometric feasibility and performance of the designs. Z-tests on the performance scores of the generated aircraft models indicate a statistically significant improvement in the functionality of the generated models after three iterations of the training-evaluation process. In the case study, a number of techniques are explored to structure the generate-evaluate process in order to balance the need to generate feasible designs with the need for innovative designs.

A novel method has been developed to optimize both the form and behavior of complex systems. The method uses spatial grammars embodied in character-recurrent neural networks (char-RNNs) to define the system including actuator numbers and degrees of freedom, reinforcement learning to optimize actuator behavior, and physics-based simulation systems to determine performance and provide (re)training data for the char-RNN. Compared to parametric design optimization with fixed numbers of inputs, using grammars and char-RNNs allows for a more complex, combinatorial infinite design space. In the proposed method, the char-RNN is first trained to learn a spatial grammar that defines the assembly layout, component geometries, material properties, and arbitrary numbers and degrees of freedom of actuators. Next, generated designs are evaluated using a physics-based environment, with an inner optimization loop using reinforcement learning to determine the best control policy for the actuators. The resulting design is thus optimized for both form and behavior, generated by a char-RNN embodying a high-performing grammar. Two evaluative case studies are presented using the design of the modular sailing craft. The first case study optimizes the design without actuated surfaces, allowing the char-RNN to understand the semantics of high-performing designs. The second case study extends the first by incorporating controllable actuators requiring an inner loop behavioral optimization. The implications of the results are discussed along with the ongoing and future work.

ABSTRACTDesign decision‐making involves tradeoffs between many design variables and attributes, which can be difficult to model and capture in complex engineered systems. To choose the best design, the decision maker is often required to analyze many different combinations of these variables and attributes and process the information internally. Trade Space Exploration (TSE) tools, including interactive and multidimensional data visualization, can be used to aid in this process and provide designers with a means to make better decisions, particularly during the design of complex engineered systems that have multiple, competing objectives. In this paper, we investigate the use of TSE tools to support decision makers using a Value‐Driven Design (VDD) approach for complex engineered systems. A VDD approach necessitates a rethinking of TSE, and we outline and illustrate four different uses of a VDD approach to TSE. The research leverages existing TSE paradigms and multidimensional data visualization tools to identify optimal designs when using a value function for a system. A satellite design example is used to demonstrate the differences between a VDD approach to design complex engineered systems and a multiobjective approach to capture the Pareto frontier. Ongoing and future work is also discussed.

Design decision-making involves trade-offs between many design variables and attributes, which can be difficult to model and capture in complex engineered systems. To choose the best design, the decision-maker is often required to analyze many different combinations of these variables and attributes and process the information internally. Trade Space Exploration (TSE) tools, including interactive and multi-dimensional data visualization, can be used to aid in this process and provide designers with a means to make better decisions, particularly during the design of complex engineered systems. In this paper, we investigate the use of TSE tools to support decision-makers using a Value-Driven Design (VDD) approach for complex engineered systems. A VDD approach necessitates a rethinking of trade space exploration. In this paper, we investigate the different uses of trade space exploration in a VDD context. We map a traditional TSE process into a value-based trade environment to provide greater decision support to a design team during complex systems design. The research leverages existing TSE paradigms and multi-dimensional data visualization tools to identify optimal designs using a value function for a system. The feasibility of using these TSE tools to help formulate value functions is also explored. A satellite design example is used to demonstrate the differences between a VDD approach to design complex engineered systems and a multi-objective approach to capture the Pareto frontier. Ongoing and future work is also discussed.

This paper develops and explores the interface between two related concepts in design decision making. First, design decision making is a process of simultaneously constructing one’s preferences while satisfying them. Second, design using computational models (e.g., simulation-based design and model-based design) is a sequential process that starts with low fidelity models for initial trades and progresses through models of increasing detail. Thus, decision making during design should be treated as a sequential decision process rather than as a single decision problem. This premise is supported by research from the domains of behavioral economics, psychology, judgment and decision making, neuroeconomics, marketing, and engineering design as reviewed herein. The premise is also substantiated by our own experience in conducting trade studies for numerous customers across engineering domains. The paper surveys the pertinent literature, presents supporting case studies and identifies use cases from our experiences, synthesizes a preliminary model of the sequential process, presents ongoing research in this area, and provides suggestions for future efforts.

In this paper we describe the development of an interactive visualization tool to support the design and evaluation of microgrid architectures in ultra low energy communities. The work is motivated by recent Department of Defense regulations to reduce energy costs at and increase energy conservation at military installations. Using two sets of energy analysis models derived from existing energy modeling software packages, we illustrate how such a design environment can be used to (1) run a fast, low fidelity model to support an initial trade space exploration, (2) understand key trends and relationships, (3) filter microgrid architectures based on desired constraints, (4) identify architectures of interest, (5) run high fidelity analyses for architectures of interest, and (6) select an architecture and use a map view to change device type locations. The process is demonstrated through a web-based design environment that we prototyped and applied to two design examples. In both cases, promising microgrid architectures are identified from an initial set of 500 randomly generated designs. Manual adjustments of the position and location of the device types were used to further improve system performance. The end result in each case was a microgrid architecture that offered low fixed and operating costs based on the assumed electrical and thermal loads. The prototype effectively illustrates how Visual Analysis might be performed during Steps 4 & 5 of the Army’s Real Property Master Planning Process. Future enhancements to support the design decision-making process are also discussed.