• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文学位 >Proactive replanning for multi-robot teams.
【24h】

Proactive replanning for multi-robot teams.

机译：主动对多机器人团队进行重新计划。

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Humans are adept at cooperating and coordinating with one another in dynamic, uncertain environments while adapting to unexpected events and delays. Human workers are able to predict the likely outcome of their current tasks, communicate with one another to schedule and reschedule cooperative tasks, and fluidly join and leave teams as the situation warrants.;This thesis investigates how to bring the fluidity, responsiveness, and reliability of human teams to a centrally coordinated, multi-robot team operating in domains requiring tight inter-robot coordination and either the maximization of overall reward or the minimization of makespan. Proactive replanning endeavors to create flexible teams, then exploit their adaptability by dynamically adjusting the teams' composition in response to predicted problems and opportunities. We have explored three components of proactive replanning: live duration prediction, mutable teams, and live task modification.;We present algorithms to predict a distribution across the remaining duration of a task, given a measurement of the task's current state and a relatively sparse set of training data. By predicting the duration of tasks throughout their execution, we are able to execute schedules that achieve 45% of the improvement possible with an omniscient planner.;Mutable teams are those that agents may join or leave throughout the cooperative task's execution. Mutable teams leverage the possibilities presented by optional roles: roles in a team that produce salutary effects, but are not strictly required for the task to proceed. By dramatically increasing the options available to the planner, the use of mutable teams allows the creation of significantly shorter schedules with 57% fewer planning iterations.;Live task modification is the adjustment of tasks already underway by the planner, requiring close coordination between planner and executive. It is informed by duration prediction and uses mutable teams to adjust executing tasks in response to predicted problems and opportunities.;We evaluate the effects of the three components of proactive replanning in isolation and in concert within a complex, stochastic, multi-agent simulated domain. We find that proactive replanning significantly reduces the makespan of executed schedules and increases the flexibility of the planner, bringing multi-robot systems one step closer to the fluid teamwork so effortlessly accomplished by humans.
机译：人类善于在动态，不确定的环境中相互合作和协调，同时适应突发事件和延迟。人类工作者能够预测他们当前任务的可能结果，彼此沟通以安排和重新安排合作任务，并根据情况需要流畅地加入和离开团队。本论文研究了如何带来流动性，响应能力和可靠性。将人员团队转变为一个集中协作的多机器人团队，这些团队在需要严格的机器人间协调以及最大化整体报酬或最小化生产期的领域中运营。积极的重新计划旨在创建灵活的团队，然后通过动态调整团队的组成以响应预期的问题和机会来利用其适应性。我们探索了主动重新计划的三个组成部分：实时持续时间预测，可变团队和实时任务修改。;在给出任务当前状态的度量和相对稀疏集的情况下，我们提出了算法来预测任务剩余持续时间的分布训练数据。通过预测任务在整个执行过程中的持续时间，我们能够执行计划，以无所不知的计划人员实现可能实现的改进的45％。可更改的团队是代理可以在整个协作任务执行期间加入或离开的团队。可变团队利用可选角色提供的可能性：产生有益效果但并非严格执行任务的团队角色。通过显着增加计划者可用的选项，使用可变团队可以创建明显更短的计划，使计划迭代次数减少57％。实时任务修改是计划者已经在进行的任务的调整，需要计划者与行政人员。它以持续时间预测为基础，并使用可变的团队来调整执行任务以响应预测的问题和机会。;我们在复杂，随机，多代理程序模拟域中单独或协同评估主动重新计划的三个组成部分的效果。我们发现，主动重新计划显着减少了已执行计划的有效期，并增加了计划者的灵活性，使多机器人系统更接近由人轻松完成的灵活团队。

著录项

  • 作者

    Sellner, Brennan.;

  • 作者单位

    Carnegie Mellon University.;

  • 授予单位 Carnegie Mellon University.;
  • 学科 Engineering Robotics.;Computer Science.
  • 学位 Ph.D.
  • 年度 2009
  • 页码 233 p.
  • 总页数 233
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类 自动化技术、计算机技术;
  • 关键词

  • 入库时间 2022-08-17 11:38:29

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号