【24h】

Human Supervisory Perception Concept for Use In Multiple Semi-Autonomous Vehicle Control

机译:用于多个半自动车辆控制的人类监督感知概念

获取原文
获取原文并翻译 | 示例

摘要

This paper will present a conceptual intelligent methodology for controlling multiplernunmanned ground vehicle systems (UGVs) in a vision-based road following scheme byrnusing “human” perception to update behavioral algorithms as needed. First, baselinernautonomous behavioral control algorithms are developed using artificial neural networkrnstructures trained using datasets based on recorded human behaviors collected in arnteleoperated mode. These baseline algorithms are implemented into a group of UGVs.rnTheir operational performance is under the direct supervision of a human operator. If it isrnnecessary for the operator to take direct control of a UGV due to sub-optimalrnperformance or a hazardous condition, a new dataset is collected and augmented withrnprevious data to serve as a new training set to re-train the neural networks for therndevelopment of “improved” behavior control algorithms. Finally, the newly updatedrnalgorithms are implemented into the group of UGVs using a novel dataflow scheme.
机译:本文将提出一种概念性的智能方法,通过使用“人类”感知来根据需要更新行为算法,从而在基于视觉的道路跟踪方案中控制多人地面车辆系统(UGV)。首先,使用人工神经网络结构开发基线自主行为控制算法,该结构使用数据集训练,该数据集是基于在自动操作模式下收集的已记录人类行为的。这些基线算法被实施到一组无人飞行器中。它们的操作性能在操作员的直接监督下进行。如果由于次优性能或危险情况而导致操作员不需要直接控制UGV,则将收集一个新的数据集并使用先前的数据进行扩充,以用作新的训练集来重新训练神经网络,以开发“改进的”行为控制算法。最后,使用新颖的数据流方案将新更新的算法实现到UGV组中。

著录项

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