Multidimensional Network Modeling and Analysis of Complex Sociotechnical Systems

Complex sociotechnical systems are always constituted of multiple types of stakeholders from different levels who exhibit complex relations and make multiple types of decisions that are dependent with each other. Moreover, there might be a large number of stakeholders, each of which has different decision-making preferences, i.e., the heterogeneity. Using vehicles as an example, customers have various tastes towards vehicle attributes and features, thus have different behavior activities, such as consideration versus purchase. Those behaviors drive designers and engineers’ decisions on vehicle design, which forms the customer-product interactions. Such interactions are complex not only because there exist multiple types of decision-making relations between customers and products, but also because there are complex relations between the customers (e.g., social interactions) and amongst the products (e.g., market segmentation or product family). In addition, constrains from car manufacturers and policies and/or incentives (e.g., the hybrid incentives and rebate) made by government agencies from other levels of perspective all affect the engineering and design strategies.  The final structure of the vehicle market system is an emergent outcome of all these complex relations, decisions and interactions. These unique features call for a systematic modeling framework for complex sociotechnical systems. In this project, we leverage the recent development in network theory and social science to develop a multidimensional network-based approach to model, analyze and predict stakeholders’ decision-making behaviors in supporting complex systems engineering and design. The framework has been successfully applied in vehicle market systems and online design competition system, such as grabCAD Challenges.

Multi-Robot Systems with Spatiotemporal Constraints – Research on Cooperative Additive Manufacturing System

While significant developments have been made over the past decade in additive manufacturing, it still trails behind the conventional manufacturing methods due to two major limitations: slow production rates due to slow print speed and lack of scalability. In this project, we have been working with Dr. Wenchao Zhou to propose a novel concept – Cooperative 3D Printing (C3DP) that aims to address both of those issues by combining two techniques: multi-robot system mimicking swarm behaviors and 3D printing, into one complex system. The key to C3DP is our chunk-based printing technique, which extends the traditional layer-based printing. The idea is to use multiple mobile 3D printing robots to complete a large print job by subdividing it into smaller chunks and assigning each of these chunks to different printers. These printers can work on these chunks simultaneously, mimicking the colony of ants, thus significantly reducing the total print time. The development of such system involves research on multi-level decision-making, evolving geometric constraints, optimization and design heuristics etc. Once the printing is complete, if additional assembly is required, assembly robots will take the task of doing so, which results in minimal to no human intervention while completing the manufacturing of a part in minimal time. Our long-term vision is to develop such a digital mega factory where a customized/personalized design can be realized with the least degree of human intervention.

Game Theory and Behavioral Experimentation in Engineering Design

Game-theoretic models have been used to analyze design problems ranging from multi-objective design optimization to decentralized design and from design for market systems to policy design. However, existing studies are primarily analytical in nature, which start with a number of assumptions about the individual decisions, the information available to the players, and the solution concept (generally, the Nash equilibrium). There is a lack of studies related to engineering design, which rigorously evaluate the validity of these assumptions or that of the predictions from the models. Hence, the usefulness of these models to realistic engineering systems design has been severely limited. In this project, we take a step toward addressing this gap. Using an example of crowdsourcing for engineering design, we illustrate how the analytical game-theoretic models and behavioral experimentation can be synergistically used to gain a better understanding of design situations. Analytical models describe what players with assumed behaviors and cognitive capabilities would do, and the behavioral experiments shed light on how individuals actually behave.

Realizing Artificial Design Intelligence in Supporting Human-Computer Collaborative Design

Effective characterizing designer’s cognitive capability and modeling their decision-making strategies (collectively called design intelligence) will essentially improve design thereby greatly reducing the rate at which a better design is achieved. But fundamentally understanding design intelligence is scientifically challenging because design itself is a complex human activity both mentally and physically. The primary objective of this project is to develop a systematic framework to quantitatively characterize and analytically model design intelligence while designers concurrently design and engineering products/systems. We believe that design is fundamentally about the transformation of information into something tangible. Therefore, the information that contains design actions is valuable for discovering patterns of effective design thinking, cognition and decision-making behaviors. With such a belief, the project will initiate a full-scale data-driven study that transforms the way of doing design research into a process of discovery and exploration. Specifically, we leverage the modern CAD technology to record the naturally occurred design activities, and use big data mining and machine learning techniques to distinguish ineffective vs. effective design decisions and iterations. We also mathematically model how designers perceive and respond to different internal or external design-assisted information, such as experts’ advice, training notes, engineering analysis results and scientific insights. Successful modeling of such a decision-making process will help create artificial design intelligence enabled agents to support the human-computer collaborative design environment in which the designers can be promoted to revise their designs and spurred to increase useful design iterations triggered by those artificial agents. Designers and manufacturers can, therefore, avoid wasteful design iterations and higher cost while retaining relevant data for post-product activities and new product development.

Data-driven Analysis and Modeling of Community-based Digital Manufacturing System

Digital manufacturing, such as 3D printers, has dramatically changed the way we manufacture, we design, and even we create things. The low-cost and rapid prototyping process provide real-time realization of design thinking and establish near-seamless connection between ideation and making, thereby greatly promote invention and creation. Despite so many advantages of 3D printers, “the industry was not delivering on the promise of 3D printing to decentralize manufacturing” [https://www.3dhubs.com/about#about-start] and did not make it accessible to everyone who possess the capability of designing and creating yet not possess the 3D printing resource. It is therefore the driving force of the emergence of the Community-Based Digital Manufacturing System (CBDMS). One of the most successful business models which realize such a concept is 3DHubs.com, which missions to “connect all 3D printers globally into one online platform and make them locally accessible.” So far they have provided more than 1 Billion people around the world to access 3D printing service within 10 miles away from the customers. Their growth implies many useful and valuable experiences and insights to the digital manufacturing of future, yet, not discovered. Using such a tangible globally distributed design-manufacturing system as an observable case, we leverage data-driven analysis and modeling approaches to obtain a better understanding of the worlds-largest CBDMS. Especially, we are interested in understanding what are the critical factors that make a 3D hub thrives or deceases; how to model the decision-making of designers (customers) in the CBDMS; what are the key drivers that shape the system as we observed today; and are there any trends or patterns in the evolution of the CBDMS. The knowledge generated from this study will help develop new auxiliary systems that can facilitate the sustainable development of CBDMS, and on the other hand, provide implications to the development of existing distributed and/or cloud-based manufacturing systems and the establishment of new systems.

Spatio-Temporal Analysis of Customers Behaviors in Supporting Product Design and Development

Evidences have shown that the customers’ preferences are not only influenced by the products (e.g., the products’ attributes) and social interactions (e.g., the social media connections), their preferences are also correlated to their locations and resulting behaviors can diffuse from one place to another. For example, in certain states, the customers adopt vehicles equipped with new technologies faster than other states, and such preferences can quickly diffuse to the adjacent states because states in certain regions share similar culture, have similar economic status and social structures. Therefore, at which states/cities to launch new products leads to distinct diffusion patterns. Moreover, customers’ preferences are evolving over time because factors such as the appreciation of beauty, thereby the complexity of customer-product relations is further compounded by the spatiotemporal effect in addition to the socio-technical effect. In this project, we hypothesize that the space and time effects play a significant role in affecting customers’ decision-making in purchasing vehicles. From enterprises’ point of view, the critical question is: what is the critical region on a market for launching new products so that the new products can be quickly adopted throughout the market. Towards answering this question, we aim to develop a framework which leverages network theory, spatial econometrics and time series analysis to analyze, model and predict the diffusion patterns of customers’ behaviors across different market regions in supporting engineering design decisions.

Decision-Centric Modeling of Air Transportation Systems

The air transportation system (ATS) is a complex evolving system with many stakeholders, such as airlines, Federal Aviation Administration (FAA) and passengers. Each of them makes decisions and act in accordance with its own respective objective. For example,  passengers decide which flight to take  depending on the fare, time of transit, in-plane service, etc. The passengers’ decisions contribute to the demand between cities, which drives the airlines’ decisions on whether to add/delete a route or no between a city-pair (i.e, route selection) and what type of aircraft to be served on a specific route (i.e., fleet planning). FAA at a higher level will make policy and regulations in response to decisions made by airlines to ensure the overall system runs safely and efficiently. Therefore, the stakeholders’ decisions collectively drive ATS’s evolution. To model such a complex system, the key is to accurately model airlines’ decision-making on route selection because it directly shapes the network topology, thereby impacts network-wide effects such as the propagation of delays and the robustness of the network to service disruptions. In this project, we first model airlines’ route selection decision-making with discrete choice random-utility theory by considering multiple factors such as market demand, operation cost, hub/non-hub characteristic of airports, etc. Our developed approach is capable of regenerating US domestic ATN that has less than 5% error to the observed real network. To further improve the model, separately modeling airlines’ decision is not enough due to the multi-stakeholder nature of the ATS. Therefore, we are working towards cohesively taking the decision-making of other stakeholders’ into consideration. Our aim is to develop a multi-level multi-stakeholder decision-cenric modeling framework to address the interactions among different entities, with the consideration of combined effect of technology, regulatory, and economic activities, in a holistic way.

Network-based Modeling Framework for Open-Source Product Evolution

Modeling the structure and evolution of products is important from the standpoint of improving design quality and maintainability. With the increasing popularity of open-source processes for developing both software and physical systems, there is a need to develop computational models of product evolution in such a dynamic product developments scenario. Existing studies on the evolution of products involve modeling products as networks, taking snapshots of the structure at different time steps, and comparing the structural characteristics. Such approaches are limited because they do not capture the underlying dynamics through which products evolve. In this project, we take a step toward addressing this gap by presenting a generative network model for product evolution. The generative model is based on different mechanisms though which networks evolve –  addition and removal of nodes, addition and removal of links. The model links local network observations to global network structures. The developed model is applied to model and analyze the evolution of a software product (Drupal) and a physical product (RepRap) developed by the open-source process. The proposed model has three general applications: a) longitudinal studies of a product’s evolution; b) cross-sectional studies of evolution of different products, and c) predictive analyzes.

Estimating Local Decision-making Behaviors in Autonomous System Level Internet

The Internet is a system-of-systems composed of devices and networking technologies operated by autonomous systems (AS). Typical ASes include Internet Service Provider, e.g., the Comcast and AT&T, and Universities and Companies who need to purchase Internet service. The decisions made by an AS on which target ASes to link with in order to access Internet service directly affect the topology of the Internet, that consequently affect the performance of the Internet, e.g., the traffic of routing and system robustness. The objective of this project is to estimate AS linking preferences in such a decision-making process. We have adopted the discrete choice modeling approach which utilizes the observed Internet topology as the input to estimate preferences of ASes by considering not only the network metrics but also the geographic and economic aspects, such as the ASes’ geographic locations, inter-domain traffic, business role and provider-customer relations. The approach provides a way to quantify the impact of different variables on the ASes’ decisions and the resulting linking probabilities. The uniqueness of the proposed approach is that it provides a theoretical and explanatory framework for ASes’ decision-making through the lens of utility-maximization principles. We are now developing improved decision-making models of ASes based on the estimated preference structures, and aim at developing AS-level Internet topology generator/simulator with such improved decision models.

Towards the Design of Complex Networks with High Robustness

Robustness quantifies the degree of persistence of a system’s characteristic behavior under perturbations. It is one of the most important design objective for large-scale complex networked systems, such as power grid and transportations systems. However, achieving desired robustness of large-scale systems, such as the smart grid with more 10,000 nodes, is scientifically challenging. This is because such systems’ performance is not under the direct control of the designers, but is an emergent outcome of self-directed individual entities (nodes). So the relations between node-level behaviors and system-level performance is highly nonlinear and complex. In this project, we aim to develop a decision-centric approach based on random-utility theory to model the evolution of complex networked systems. It accounts for the decision making preferences of each node, thereby establishing a direct mapping from node-level preferences to the network structure and performance. By controlling the node-level preferences, the proposed model is capable of modeling a large variety of network topologies with different desired robustness properties. However, it is observed that the robustness of networks is extremely sensitive to the nodes’ linking preferences. Such observation guide the design of incentives and/or mechanisms with proper selection of parameters for the design of complex networks so that the high robustness can be achieved.