动态异构云机器人实时感知的资源分配算法的研究

For traditional robotic systems, robots have to carry adequate physical processing power and various sensors among other resources to facilitate the completion of various tasks such as simultaneous localization and mapping, object recognition and grasping, especially in real-time operation. With the cloud, robots and automation system would not be limited by onboard computation, memory and learning strategies. However, cloud robotics technology cannot be well developed and applied in various applications without running the intelligent sensor data retrieval and resource allocation strategies in the system. Regarding the problems, the proposed project focuses on resource allocation algorithms for real-time perception in cloud robotic systems. The main objectives are as follows: (1) Regarding the heterogeneous resources and high dynamic demands, this project proposes a dynamic resource allocation model based on graph theory and hidden Markov process. (2) To propose the bidding game optimization method based dynamic resource allocation algorithms for real-time perception in cloud robotic systems. (3) To prove the Nash Equilibrium existence and Pareto efficiency of the proposed resource allocation algorithms and to evaluate the algorithms by real-time perception scenarios. The analysis and evaluation results can be used as a reference to the algorithm improvement.

传统机器人在执行任务如即时定位与地图构建、物品抓取等任务时，大量环境感知数据的存储与计算对机器人硬件配置的要求越来越高，实时执行尤其困难。云机器人不仅能够卸载复杂的计算任务到云端，而且可以共享云感知信息和技能。然而，由于缺乏建立智能的信息检索和资源配置决策的基础支撑平台，目前云机器人的研发还处于萌芽阶段。针对上述问题，本项目围绕动态异构的云机器人实时感知开展资源分配算法的研究工作, 主要目标包括: (1) 针对云机器人的动态、异构的特性，基于图论和隐马尔可夫过程，提出实时感知场景下的动态资源供需模型；(2) 研究面向云机器人实时感知系统的动态资源供需架构，基于竞价博弈的最优化方法，提出能够适应具有动态异构特性的资源配置算法；(3) 基于纳什均衡点存在的证明和帕累托效率分析进行资源配置算法的理论验证，并建立云机器人实时感知场景实验来验证算法实用性能，为算法改进提供参考。

传统机器人在执行任务如即时定位与地图构建、物品抓取等任务时，大量环境感知数据的存储与计算对机器人硬件配置的要求越来越高，实时执行尤其困难。云机器人不仅能够卸载复杂的计算任务到云端，而且可以共享云感知信息和技能。然而，由于缺乏建立智能的信息检索和资源配置决策的基础支撑平台，目前云机器人的研发还处于萌芽阶段。针对上述问题，本项目围绕动态异构的云机器人实时感知开展资源分配算法的研究工作, 主要目标包括: (1) 针对云机器人的动态、异构的特性，基于图论和隐马尔可夫过程，提出实时感知场景下的动态资源供需模型；(2) 研究面向云机器人实时感知系统的动态资源供需架构，基于竞价博弈的最优化方法，提出能够适应具有动态异构特性的资源配置算法；(3) 基于纳什均衡点存在的证明和帕累托效率分析进行资源配置算法的理论验证，并建立云机器人实时感知场景实验来验证算法实用性能，为算法改进提供参考。