Using probabilistic models to handle the messy, quantitative "noise" of the real world, like unreliable sensors or imperfect physical actions. Key Findings
In experiments with physical robots, this combined architecture reduced the time it took to finish tasks by 39% compared to traditional methods. Kr (3) mp4
The system automatically picks out the most important details for each step, which prevents the robot from being overwhelmed by too much data. Where to Find It Using probabilistic models to handle the messy, quantitative
: An Architecture for Knowledge Representation and Reasoning in Robotics" by Shiqi Zhang and colleagues. Where to Find It : An Architecture for
Using a declarative "action language" to help the robot understand qualitative knowledge and prioritized rules.
This paper introduces a system designed to improve how robots handle complex tasks by combining two different ways of "thinking":