Learning Actions with Symbolic Literals and Continuous Effects for a Waypoint Navigation Simulation.

Morgan Fine-Morris,Bryan Auslander,Hector Muñoz-Avila, Kalyan Moy Gupta

sai intelligent systems conference(2020)

Cited 0|Views4
No score
Abstract
We present an algorithm for learning planning actions for waypoint simulations, a crucial subtask for robotics, gaming, and transportation agents that must perform locomotion behavior. Our algorithm is capable of learning operator’s symbolic literals and continuous effects even under noisy training data. It accepts as input a set of preprocessed positive and negative simulation-generated examples. It identifies symbolic preconditions using a MAX-SAT constraint solver and learns numeric preconditions and effects as continuous functions of numeric state variables by fitting a logistic regression model. We test the correctness of the learned operators by solving test problems and running the resulting plans on the simulator.
More
Translated text
Key words
waypoint navigation simulation,symbolic literals,actions
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined