Supporting Decision Making in Engineering Design Using Parallel Coordinates in Virtual Reality

Abstract

Computational engineering design methods and tools are common practices in the modern industry. Such approaches are integral in enabling designers to efficiently explore large and complex design spaces. However, they also tend to dramatically increase the number of candidate solutions that decision makers must correctly interpret. Since all candidate solutions can be represented in a digital form together with their assessment criteria, a natural way to explore and understand the complexities of the design problem is to visualize their multidimensional nature. The task now involves the discovery of patterns and trends within a multidimensional design space. In this work, we aim to enhance the design decision making process with immersive Parallel Coordinates Plot (IPCP) in virtual reality. We present the design of this system, which allows representation and exploration of multidimensional scientific datasets. A qualitative validation with two surrogate expert users demonstrated that the system can be used successfully to both detect both known and previously unknown patterns and support learning the decision making process in a shorter time. The results serve as a promising indication of how immersive parallel coordinate plots can enhance decision support in complex engineering design processes.

Virtual Reality-Based Parallel Coordinates Plots Enhanced with Explainable AI and Data-Science Analytics for Decision-Making Processes

“…The way in which our implementation varied from the examples listed above was discussed in detail in Tadeja et al [2][3][4]. However, the IPCP design presented in this paper was refined based on our prior findings revealed by the user studies [2][3][4]. For instance, we restricted the number of available 3D scatter plots to only one.…”

“…For instance, we restricted the number of available 3D scatter plots to only one. This is due to the fact that having one such scatter plot composed of the criteria dimensions was enough to make an exhaustive search of the cluster given the exploration method suggested to the users [2][3][4]. Moreover, as the users can now use their hands [4,9] to manipulate the IPCP and scatter plot alike, they are no longer relying on any type of hand-held controller to select the interactive objects, and the movement facilitated with the controllers was also restricted to the head-mounted display (HMD)-tracking executed with the Oculus Rift sensors [25].…”

“…Therefore, methods that aid experts and users in the process of analyzing the PCP visualization and developing a new understanding of a decision-making process are needed. Thus, in this paper, we present a refinement of a previously developed system called Immersive Parallel Coordinates Plots (IPCP) [2][3][4], which is now augmented with new and improved functionality. In our case, data visualization is presented to the user with the help of an interactive and fully immersive Virtual Reality (VR) interface.…”

“…In our case, data visualization is presented to the user with the help of an interactive and fully immersive Virtual Reality (VR) interface. A previous exploratory and qualitative study carried out using the first version of IPCP demonstrated that users were able to successfully detect patterns in high-dimensional datasets visualized as-is [2][3][4], that is, without any prior analysis or data cleaning, using the IPCP system (see Figure 1) combined with a simple naive clustering algorithm applied to cross-dimensional 3D scatter plots data [2][3][4]. However, this basic IPCP system allowed users to rediscover the knowledge; specifically, patterns with the data items previously found by the domain experts and, more importantly, users were able to discover new knowledge in the dataset compared to original analyses carried out by experts using a more traditional 2D PCP system [3,4].…”

“…In comparison, the version presented in this paper has been extended with features and improvements of functionality bestowed by the IPCP [3,4] and supporting 3D scatter plots. Similar to prior work [3,4] we rely on two datasets to drive the system development process, which we will refer to as the Pareto front data (54 data items with 29 dimensions each) as dataset DS1 and the S-duct design optimization study (166 data items with 39 dimensions each) as dataset DS2.…”