Application #adaptability has a new dimension in cyber-physical systems. To navigating unexpected circumstances, autonomous systems need to possess significant #planning capabilities. Planning as a field of #AI receives very little attention comparing to machine learning or knowledge management. In many cases, we are still talking about simple applications like autonomous sanitizer spraying, which is a relatively simple task. #Autonomous construction or autonomous military systems need to operate in environments where the decision space is vast. It means that a sequence of actions necessary to accomplish the task is long and needs to be put together by the autonomous systems. Planning is fundamentally different from inductive learning, although a predictive component would be helpful. The number of degrees of freedom, i.e., possible actions of an autonomous articulated robot or an attack drone is significant. Any AI developer who tested his/her algorithm in a block world will know the NP-hardness of the problem.
#winograd #shrdlu #desisionspace #aiplanning #sussman
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.