Engineering topics
Publications and source records attributed to Pang, Wanlin.
The project goals are: Make data analysis faster and cheaper. Increase use of NASA data by removing barriers to data access. Cope with data heterogeneity. Support code reuse and rapid application development. Support multiple applications, users. Including fire and health domains. Improve QOS. Always provide an answer. Tell user how good it is, where it come from.
The latest generation of NASA Earth Observing System (EOS) satellites has brought a new dimension to continuous monitoring of the living part of the Earth System, the biosphere. EOS data can now provide weekly global measures of vegetation productivity and ocean chlorophyll, and many related biophysical factors such as land cover changes or snowmelt rates. However, the highest economic value would come from forecasting impending conditions of the biosphere, to allow decision makers to mitigate dangers or exploit positive trends. NASA's strategic plan for the Earth Science Enterprise i d e n a s ecological forecasting as a focus for research. Ecological forecasting predicts the effects of changes in the physical, chemical and biological environment on ecosystem activity. Possible applications of such a system include predicting shortfalls or bumper crops of agricultural production, populations of threatened or invasive species or wildfire danger in time to allow improves preparation and logistical efficiency. Petabytes of remote sensing data are now available to help measure, understand and forecast changes in the Earth system, but using these data effectively can be surprisingly hard. The volume and variety of data files and formats are daunting. Simple data management activities, such as locating and transferring files, changing file formats, gridding point data, and scaling and reprojecting gridded data, can consume far more personnel time and resources than the actual data analysis. Some scientists commit to a particular data source or resolution just because using anything different would be more effort that it's worth. Better tools can help, but most of the tools developed to date are little more than shell scripts; they lack the flexibility to meet the diverse needs of users and are difficult to extend to handle changes in available data sources.
This paper presents a graph-based backtracking algorithm designed to support constrain-tbased planning in data production domains. This algorithm performs backtracking at two nested levels: the outer- backtracking following the structure of the planning graph to select planner subgoals and actions to achieve them and the inner-backtracking inside a subproblem associated with a selected action to find action parameter values. We show this algorithm works well in a planner applied to automating data production in an ecological forecasting system. We also discuss how the idea of multi-level backtracking may improve efficiency of solving semi-structured constraint problems.
This paper discusses a planner-based approach to automating data production tasks, such as producing fire forecasts from satellite imagery and weather station data. Since the set of available data products is large, dynamic and mostly unknown, planning techniques developed for closed worlds are unsuitable. We discuss a number of techniques we have developed to cope with data production domains, including a novel constraint propagation algorithm based on planning graphs and a constraint-based approach to interleaved planning, sensing and execution.
This paper discusses the data production problem, which consists of transforming a set of (initial) input data into a set of (goal) output data. There are typically many choices among input data and processing algorithms, each leading to significantly different end products. To discriminate among these choices, the planner supports an input language that provides a number of constructs for specifying user preferences over data (and plan) properties. We discuss these preference constructs, how we handle them to guide search, and additional challenges in the area of preference management that this important application domain offers.
In this paper, we introduce JNET, a novel constraint representation and reasoning framework that supports procedural constraints and constraint attachments, providing a flexible way of integrating the constraint system with a runtime software environment and improving its applicability. We describe how JNET is applied to a real-world problem - NASA's Earth-science data processing domain, and demonstrate how JNET can be extended, without any knowledge of how it is implemented, to meet the growing demands of real-world applications.
This paper discusses an approach to representing and reasoning about constraints over strings. We discuss how many string domains can often be concisely represented using regular languages, and how constraints over strings, and domain operations on sets of strings, can be carried out using this representation.
NASA s vision for Earth Science is to build a "sensor web"; an adaptive array of heterogeneous satellites and other sensors that will track important events, such as storms, and provide real-time information about the state of the Earth to a wide variety of customers. Achieving his vision will require automation not only in the scheduling of the observations but also in the processing af tee resulting data. Ta address this need, we have developed a planner-based agent to automatically generate and execute data-flow programs to produce the requested data products. Data processing domains are substantially different from other planning domains that have been explored, and this has led us to substantially different choices in terms of representation and algorithms. We discuss some of these differences and discuss the approach we have adopted.
Many AI tasks can be formalized as constraint satisfaction problems (CSPs), which involve finding values for variables subject to constraints. While solving a CSP is an NP-complete task in general, tractable classes of CSPs have been identified based on the structure of the underlying constraint graphs. Much effort has been spent on exploiting structural properties of the constraint graph to improve the efficiency of finding a solution. These efforts contributed to development of a class of CSP solving algorithms called decomposition algorithms. The strength of CSP decomposition is that its worst-case complexity depends on the structural properties of the constraint graph and is usually better than the worst-case complexity of search methods. Its practical application is limited, however, since it cannot be applied if the CSP is not decomposable. In this paper, we propose a graph based backtracking algorithm called omega-CDBT, which shares merits and overcomes the weaknesses of both decomposition and search approaches.
Image process planning, lifted plan graph and induced CSP splitting states and actions in lifted plan graph enables improved arc consistency in underlying CSP.