Learning Situation-Dependent Costs:
Karen Zita Haigh and Manuela M. Veloso,
from Probabilistic Robot Execution
School of Computer Science, Carnegie Mellon University
Invited submission to the journal Robotics and Autonomous Systems
Physical domains are notoriously hard to model completely and correctly,
especially to capture the dynamics of the environment. In this article, we
a robot that learns from its execution experiences.
Since actions may have different costs under different conditions, we
introduce the concept of situation-dependent rule, in which situational
features are attached to the costs or probabilities, reflecting patterns and
dynamics encountered in the environment.
extracts learning opportunities from massive, continual,
probabilistic execution traces. It then correlates these learning
opportunities with environmental features, creating situation-dependent
costs for its actions. We present the development and use of these rules
for a robotic path planner.
Our empirical results show that situation-dependent rules effectively
improve the planner's model of the environment, thus allowing the planner to
predict and avoid failures, to create plans that are tailored to the real
world, and to respond to a changing environment.