Planning Algorithms for Multi-Robot Active Perception
Planning Algorithms for Multi-Robot Active Perception
复制标题
多机器人主动感知规划算法
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Graeme Best
中科院分区：
文献类型：
--
作者：
Graeme Best
多机器人主动感知的规划算法机器人系统的一项基本任务是使用机载传感器和感知算法来理解环境的高级语义属性。这些语义属性可能包括环境的地图、对象的存在、姿态和类别、其他代理的行为或动态领域的参数。观察值高度依赖于视点，因此，通过规划机器人的运动来获得高价值的观察值可以大大提高感知算法的性能。这激发了主动感知的问题，其目标是在考虑运动约束和手头任务的感知目标的同时，为机器人团队规划观察视点。这个基本问题是许多机器人应用的核心，包括环境监测、搜索和救援、行星探索和精准农业。本文的核心贡献是一套针对多机器人主动感知的规划算法。这些规划算法旨在提高多机器人系统在许多方面的系统级性能。我们的目标是解决现有工作中尚未充分解决的多机器人主动感知的各种挑战：在线和随时规划，长期优化，分散协调，对不可靠通信的鲁棒性，预测其他代理的计划，以及利用特定感知模型的特征。我们首先提出分散式蒙特卡罗树搜索（Dec-MCTS）算法，作为一种普遍适用的分散式多机器人主动感知算法。Dec-MCTS是广泛使用的蒙特卡罗树搜索算法的一种新颖的分散变体，并且还利用变分方法的思想来规划动作序列的概率分布。Dec-MCTS是随时随地的，对通信损失具有鲁棒性，在扩展决策树时平衡了探索-开发的权衡，并收敛于最小化kl -散度到最优联合计划的计划。
Graeme Best Doctor of Philosophy The University of Sydney May 2018 Planning Algorithms for Multi-Robot Active Perception A fundamental task of robotic systems is to use on-board sensors and perception algorithms to understand high-level semantic properties of an environment. These semantic properties may include a map of the environment, the presence, pose and class of objects, the behaviour of other agents, or the parameters of a dynamic field. Observations are highly viewpoint dependent and, thus, the performance of perception algorithms can be greatly improved by planning the motion of the robots to obtain high-value observations. This motivates the problem of active perception, where the goal is to plan the observation viewpoints for a team of robots while considering both the motion constraints and the perception objectives of the task at hand. This fundamental problem is central to many robotics applications, including environmental monitoring, search and rescue, planetary exploration, and precision agriculture. The core contribution of this thesis is a suite of planning algorithms for multi-robot active perception. These planning algorithms are designed to improve the systemlevel performance of multi-robot systems on many fronts. We aim to address various challenges of multi-robot active perception that have not been adequately addressed in existing work: online and anytime planning, optimising over a long time horizon, decentralised coordination, being robust to unreliable communication, predicting plans of other agents, and exploiting characteristics of specific perception models. We first propose the decentralised Monte Carlo tree search (Dec-MCTS) algorithm as a generally-applicable, decentralised algorithm for multi-robot active perception. Dec-MCTS is a novel, decentralised variant of the widely-used Monte Carlo tree search algorithm, and also leverages ideas from variational methods to plan over probability distributions of action sequences. Dec-MCTS is anytime, is robust to communication loss, balances the exploration-exploitation trade-off when expanding the decision tree, and converges towards a plan that minimises KL-divergence to the optimal joint plan.
多机器人在线构建通讯地图
DOI：
10.1109/icra.2017.7989300
发表时间：
2017
期刊：
IEEE International Conference on Robotics and Automation (ICRA
影响因子：
--
作者：
Banfi, Jacopo;Li, Alberto Quattrini;Basilico, Nicola;Rekleitis, Ioannis;Amigoni, Francesco
通讯作者：
Amigoni, Francesco
DOI：
10.1109/icra.2015.7139866
发表时间：
2015
期刊：
--
影响因子：
--
作者：
Ondruska P
通讯作者：
Ondruska P
使用桥接分布进行对象跟踪中的贝叶斯意图预测
DOI：
10.48550/arxiv.1508.06115
发表时间：
2015
期刊：
--
影响因子：
--
作者：
Ahmad B
通讯作者：
Ahmad B