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.)
Interweaving reason, action and perception
Abstract
In their attempt to understand and emulate intelligent behavior Artificial Intelligence researchers have taken a reductionist approach and divided their investigation into separate studies of reason, perception and action. As a consequence, intelligent robots have been constructed using a coarse grained architecture; reasoning, perception and action have been implemented as separate modules that interact infrequently. This dissertation describes an investigation into the effect of reducing this architecture granularity on the efficiency of the overall system. The thesis takes the position that significant computational efficiencies can be gained by introducing a fine grained integration or "interweaving" of these functions and demonstrates these savings for an intelligent navigation system. The dissertation develops a paradigm referred to as the "reason a little, move a little, look a little", or RML paradigm, and describes a specific RML implementation used for experimentation. The results from this work show significant complexity reduction for planning and vision due to the RML paradigm. It is argued that the replanning triggered by the realities of carrying out actions in a real environment cause the planner in a coarse grained system to exhibit a complexity of O($\rho\sp2)$ ($\rho$ = path length) whereas the fine grained RML system exhibits a complexity between O($\rho\ \log(\rho))$ and O($\rho)$. It is then argued that the complexity of vision in the RML system is $$\rm O\left\lbrack max\left\lbrack n{*}m, {n\over s\sp2}\right\rbrack\right\rbrack$$where n$*$m is a measure of the complexity of the local environment and s is the number of vision invocations per foot. Finally, it is shown that the RML system results in accurate action execution. In addition to this global result, the design details of the experimental system reveal novel approaches to representation, planning and vision. The environment is represented as a network which organizes space into a "contained-by" hierarchy which is used as a mechanism to cope with positional uncertainty and to focus planning computations. Planning is done in three dimensions and plans are developed using a hierarchical decomposition induced by the geometry of the environment. The approach to vision is analogous to the way a blind man uses his cane: to verity that reason is consistent with reality.
Subject Area
Computer science|Artificial intelligence
Recommended Citation
Fennema, Claude Lester, "Interweaving reason, action and perception" (1991). Doctoral Dissertations Available from Proquest. AAI9207395.
https://scholarworks.umass.edu/dissertations/AAI9207395