Publications

Sorted by DateClassified by Publication TypeClassified by Research Category

Multi-agent active perception with prediction rewards

Mikko, Lauri and Frans A. Oliehoek. Multi-agent active perception with prediction rewards. In Advances in Neural Information Processing Systems 33, December 2020.

Download

pdf [352.5kB]  

Abstract

How can we plan efficiently in real time to control an agent in a complex environ- ment that may involve many other agents? While existing sample-based planners have enjoyed empirical success in large POMDPs, their performance heavily relies on a fast simulator. However, real-world scenarios are complex in nature and their simulators are often computationally demanding, which severely limits the performance of online planners. In this work, we propose influence-augmented online planning, a principled method to transform a factored simulator of the entire environment into a local simulator that samples only the state variables that are most relevant to the observation and reward of the planning agent and captures the incoming influence from the rest of the environment using machine learning methods. Our main experimental results show that planning on this less accurate but much faster local simulator with POMCP leads to higher real-time planning performance than planning on the simulator that models the entire environment.

BibTeX Entry

@inproceedings{Lauri20NeurIPS,
    author =    {Lauri, Mikko, and Oliehoek, Frans A.},
    title =     {Multi-agent active perception with prediction rewards},
    booktitle = NIPS33,
    year =      2020,
    month =     dec,
    keywords =   {refereed},
    abstract = {
    How can we plan efficiently in real time to control an agent in a complex environ-
    ment that may involve many other agents? While existing sample-based planners
    have enjoyed empirical success in large POMDPs, their performance heavily relies
    on a fast simulator. However, real-world scenarios are complex in nature and
    their simulators are often computationally demanding, which severely limits the
    performance of online planners. In this work, we propose influence-augmented
    online planning, a principled method to transform a factored simulator of the entire
    environment into a local simulator that samples only the state variables that are
    most relevant to the observation and reward of the planning agent and captures
    the incoming influence from the rest of the environment using machine learning
    methods. Our main experimental results show that planning on this less accurate
    but much faster local simulator with POMCP leads to higher real-time planning
    performance than planning on the simulator that models the entire environment.
    }
}

Generated by bib2html.pl (written by Patrick Riley) on Wed Dec 02, 2020 15:44:26 UTC