Intelligent Data Understanding for Entry, Descent, and Landing, Architecture Analysis
Abstract
Designing Planetary Entry, Descent, and Landing Systems requires analyzing a wide range
of architectures and scenarios with high fidelity Monte Carlo simulations of performance under uncertainty. Given the complexity of these systems, datasets contain tens of thousands of
parameters describing the system and the environment. These datasets are generally manually analyzed by subject matter experts, trying to find interesting correlations and couplings
between parameters that explain the behaviors observed. Such analysis work is critical, given that it could lead, for example, to the discovery of major flaws in a design. While the subject matter experts can leverage their knowledge and expertise with past systems to identify issues and features of interest in the current dataset, the next generation of EDL systems will make use of new technologies to address the issue of landing larger payloads, and may present unprecedented challenges that may be missed by the human.
In this thesis, we present Daphne, a cognitive assistant, into the process of EDL architecture analysis to support EDL experts by identifying key factors that impact EDL system metrics. Specifically, this thesis describes the current capabilities of Daphne as a platform for EDL architecture analysis by means of a case study of a sample EDL architecture for an ongoing NASA mission, Mars 2020. Given that the work presented in this thesis is in its early development, the thesis focuses on the description of the expert knowledge base and historical database developed for the cognitive assistant, as well as on describing how experts can use it to obtain information relevant to their EDL analysis process by means of natural language or web visual interactions, thus reducing the effort of searching for relevant information from multiple sources.
A popular approach to automate the extraction of explanation rules of data is association rule mining, in which rules with high statistical strength are mined from a dataset. However, current rule mining algorithms (e.g., apriori, FP-growth) generate too many rules that are redundant or not useful because they are
too complex, too obvious, or don’t make sense to the user. In this thesis, we propose a new
approach to improve the comprehensibility, insightfulness, and usefulness of the association
rules generated during the analysis of an EDL dataset by leveraging a user-provided knowledge
graph. The knowledge graph captures the user knowledge about EDL and the specific problem
at hand. We then use a statistical relational learning framework based on probabilistic soft
logic to assess the degree of consistency of the rule with our knowledge of the system. We
hypothesize that rules that are considered more consistent with the knowledge graph will be
perceived by the user as being more comprehensible (making more sense) than rules that are
less consistent with the knowledge graph. We test this hypothesis – and more generally the
relation between our proposed metric and the perceived usefulness and insightfulness of a rule– in a small study with N=6 subject matter experts. Results support our primary hypothesis and
also show interesting relationships between comprehensibility, usefulness, and insightfulness of the extracted rules. These findings can enable a more personalized and adaptive approach
to intelligent data understanding, a key enabling technology to help aerospace organizations
make sense of the large and heterogeneous datasets that are becoming available in many areas
of science and engineering.
Citation
Santini De Leon, Samalis (2021). Intelligent Data Understanding for Entry, Descent, and Landing, Architecture Analysis. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /195156.