Robotics Research Lab
CRES
USC Computer Science
USC Engineering
USC
/ Research / Projects / Scalable Behavior-Based Multi-Robot Systems

Introduction Project Goals Experimental Domains Relevant Publications

Introduction

The original formulation of the behavior-based approach to robot control can be characterized by what we refer to as PAB, or Port-Arbitrated Behaviors. PAB techniques provide clean, consistent abstractions for behavioral interaction, and facilitate the bottom-up design methodology that has led to the "New AI" revolution in robot control.

Most researchers, however, tend to look towards more "Traditional AI" approaches when scaling robot behavior to higher levels of competence. This is particularly apparent in the case of multi-robot coordination, where such standard abstractions do not exist for inter-robot behavioral interaction.

The Ayllu distributed behavior-based control system extends the well-defined techniques of PAB control across IP networks, and adds to the standard PAB repertoire a number of abstractions that allow systems to scale arbitrarily.

This project focuses primarily on the Broadcast of Local Eligibility (BLE), a standard technique that allows fully-distributed, flexible team coordination which is adaptive to environmental changes and robust to robot failure. The combination of BLE for inter-robot coordination and local subsumption, or Cross-Subsumption, is used as the basis for system implementation in various experimental domains.

Project Goals

This project seeks to answer the following questions about such networked PAB systems:

Experimental Domains

blerobots

Relevant Publications