Off-campus UMass Amherst users: To download dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users, please click the view more button below to purchase a copy of this dissertation from Proquest.
(Some titles may also be available free of charge in our Open Access Dissertation Collection, so please check there first.)
Sophisticated control for interpretation: Planning to resolve sources of uncertainty
Abstract
Interpretation is the process of determining a high-level, abstract view of sensor or other observational data. Interpretation can be a complex and uncertain process which requires sophisticated evidential reasoning and control schemes. We have developed an interpretation framework which models interpretation as a process of gathering evidence to resolve particular sources of uncertainty in the interpretation hypotheses. This allows us to directly resolve uncertainty through the use of differential diagnosis techniques instead of being limited to incremental hypothesize and test. The key components of the approach are an evidential representation which includes explicit, symbolic statements of the sources of uncertainty in the evidence for the hypotheses and a script-based, incremental control planner. The control plan schemas which define the interpretation methods contain explicit information gathering actions which examine the symbolic sources of uncertainty associated with particular hypotheses. This information is used to post goals for resolving specific uncertainties in the current interpretations. These goals then direct the system to expand methods which are appropriate for resolving the uncertainties represented in the goals. Strategy knowledge is defined in focusing heuristics which are applied during the planning process to select the best methods and method instances to be pursued. Sometimes focusing decisions may not be able to be made definitively at the appropriate point during planning because there is insufficient information to select the most appropriate alternative. In these cases, the decisions may be postponed and multiple alternative methods partially expanded in order to accumulate sufficient information to select between them. This is accomplished with a refocusing mechanism which allows focusing decisions to be reconsidered once the alternatives have been expanded appropriately. Thus, the control process can be viewed as a search for the best methods to pursue as well as a search for the correct interpretations. The refocusing mechanism has also been extended to provide the goal-directed planner with a data-directed, opportunistic control capability.
Subject Area
Computer science|Artificial intelligence
Recommended Citation
Carver, Norman F., "Sophisticated control for interpretation: Planning to resolve sources of uncertainty" (1990). Doctoral Dissertations Available from Proquest. AAI9110115.
https://scholarworks.umass.edu/dissertations/AAI9110115