PILOT: An Actor-oriented Learning and Optimization Toolkit for Robotic Swarm Applications

摘要

We present PILOT (Ptolemy Inference, Learning, and Optimization Toolkit), an actor-oriented machine learning and optimization toolkit that is designed for developing data intensive distributed applications for sensor networks. We dene an actor interface that bridges state-space models for robotic control problems and a collection of machine learning and optimization algorithms, then demonstrate how the framework leverages programmability of sophisticated distributed robotic applications on streaming data. As a case study, we consider a cooperative target tracking scenario and study how the framework enables adaptation and implementation of control policies and simulation within environmental constraints by presenting actor-oriented abstractions that enable application developers to build state-space aware machine learning and optimization actors.