PI selva's lab, including former PHD student and nstrf fellow nozomi hitomi and current PHD students prachi dutta and PAU garcia buzzi, have done substantial work to improve the computational…

This $159,804.27 research grant was awarded by the National Aeronautics and Space Administration (NASA) Shared Services Center to the Texas A&M Engineering Experiment Station (Tees), a research organization within the Texas A&M University System. The grant supports work by Principal Investigator Dr. Selva's lab, including former and current graduate students, to develop computational tools for optimizing the design and architecting of distributed satellite systems. The research focuses on improving multiobjective evolutionary algorithms, which are used to solve complex problems with multiple, conflicting objectives. The work aims to better incorporate expert knowledge and adapt exploration-exploitation tradeoffs to enhance the optimization process. There is no set-aside designation for this grant, which was awarded in August 2018 with an August 2020 completion date.

PI selva's lab, including former PHD student and nstrf fellow nozomi hitomi and current PHD students prachi dutta and PAU garcia buzzi, have done substantial work to improve the computational efficiency of trade-space exploration tools that assist in the design and architecting of distributed satellite systems [1], [2], [11] [16], [3] [10]. One line of work that is relevant to this proposal has focused on multiobjective evolutionary algorithms (moea), which are typically used to optimize problems with multiple, incommensurable objectives and non-linear, non-convex, and non-differentiable properties. A challenge in moea and global optimization in general is the trade-off between exploration and exploitation: how to optimally assign computing resources to the conflicting objectives of discovering new promising areas where the global optimum might be (exploration), and searching in those areas to find the local optima (exploitation). Another related challenge is how to incorporate expert knowledge into the optimization. Expert mission designers may for example know that packing several high-energy instruments into a single spacecraft is generally not a good idea, or that certain orbits (e.g., afternoon sun-synchronous orbit) are better for certain instruments and applications (e.g., atmospheric chemistry spectrometers) than others. However, using this knowledge by encoding IT, for example, in hard constraints, may actually decrease performance if knowledge is not accurate or general enough, due to a lack of exploration. Our algorithms outperform the state-of-the-art because they can use available information including both expert knowledge and knowledge learned during the optimization to improve exploitation in a way that doesn t reduce exploration as much as other approaches (e.g., constraints) do. Specifically, advancements have been made on three fronts: 1) how to best incorporate expert knowledge to effectively guide the search towards promising regions of the trade-space, 2) how to make optimal exploration-exploitation decisions through adaptive operator selection strategies (AOS), and 3) how to extract new knowledge by applying data mining during the optimization process (knowledgedriven optimization or KDO). Figure 1 shows a flowchart of the framework developed by selva's group.