Combining Planning and Learning of Behavior Trees for Robotic Assembly

TL;DR

This paper introduces a method that combines automated planning and machine learning to generate Behavior Trees for robotic assembly, improving flexibility and transferability to real systems.

Contribution

It presents a novel approach using Genetic Programming to integrate planning results into Behavior Trees, enhancing performance over individual methods.

Findings

Outperforms standalone planning and learning methods on assembly tasks.

Successfully transfers high-level Behavior Trees to real robots without additional training.

Demonstrates versatility across various robotic assembly problems.

Abstract

Industrial robots can solve very complex tasks in controlled environments, but modern applications require robots able to operate in unpredictable surroundings as well. An increasingly popular reactive policy architecture in robotics is Behavior Trees but as with other architectures, programming time still drives cost and limits flexibility. There are two main branches of algorithms to generate policies automatically, automated planning and machine learning, both with their own drawbacks. We propose a method for generating Behavior Trees using a Genetic Programming algorithm and combining the two branches by taking the result of an automated planner and inserting it into the population. Experimental results confirm that the proposed method of combining planning and learning performs well on a variety of robotic assembly problems and outperforms both of the base methods used separately. We also show that this type of high level learning of Behavior Trees can be transferred to a real system without further training.