Multi-level spatial modeling for stochastic distributed robotic systems

doi:10.1177/0278364910399521

Multi-level spatial modeling for stochastic distributed robotic systems

2011-04-13

51

OA

AI

A

Alcherio Martinoli

DOI：10.1177/0278364910399521

原文链接

原文求助

摘要

En 中文

We propose a combined spatial and non-spatial probabilistic modeling methodology motivated by an inspection task performed by a group of miniature robots. Our models explicitly consider spatiality and yield accurate predictions on system performance. An agent's spatial distribution over time is modeled by the Fokker-Planck diffusion model and complements current non-spatial microscopic and macroscopic models that model the discrete state distribution of a distributed robotic system. We validate our models on a microscopic level based on sub-microscopic, embodied robot simulations as well as real robot experiments. Subsequently, using the validated microscopic models as our template, abstraction is raised to the level of macroscopic difference equations. We discuss the dependency of the modeling performance on the distance from the robot origin (drop-off location) and temporal convergence of the team distribution. Also, using an asymmetric setup, we show the necessity of spatial modeling methodologies for environments where the robotic platform underlies drift phenomena.

Keyword:

Distributed intelligent systems

distributed robotics

probabilistic modeling

swarm intelligence

Multi-level spatial modeling for stochastic distributed robotic systems

摘要

期刊

机构

引用论文