在复杂连续环境下,强化学习系统的状态空间面临维数灾难问题,需要采取量化的方法,降低输入空间的复杂度。径向基神经网络(RBFNN:Radial Basis Function Neural Networks)具有较强的函数逼近能力及泛化能力,由此提出了基于径向基神经网络...在复杂连续环境下,强化学习系统的状态空间面临维数灾难问题,需要采取量化的方法,降低输入空间的复杂度。径向基神经网络(RBFNN:Radial Basis Function Neural Networks)具有较强的函数逼近能力及泛化能力,由此提出了基于径向基神经网络的Q学习方法,并将其应用于单机器人的自主导航。在基于径向基神经网络的强化学习系统中,用径向基神经网络逼近状态空间和Q函数,使学习系统具有良好的泛化能力。仿真结果表明,该导航方法具有较强的避碰能力,提高了机器人对环境的适应能力。展开更多
在室内同时定位与建图(SLAM)的实际应用中,对称单一结构环境易造成激光SLAM错误建图,低质量光照或低纹理环境易造成视觉SLAM失效。针对上述室内退化环境,提出一种将激光、视觉、惯性测量单元(IMU)进行紧耦合的LVI-SLAM方法。在该方法前...在室内同时定位与建图(SLAM)的实际应用中,对称单一结构环境易造成激光SLAM错误建图,低质量光照或低纹理环境易造成视觉SLAM失效。针对上述室内退化环境,提出一种将激光、视觉、惯性测量单元(IMU)进行紧耦合的LVI-SLAM方法。在该方法前端,设计视觉评价环节对视觉信息置信度进行自适应调整;在该方法后端,进行位姿图优化以及多传感器回环抑制累积误差。视觉评价实验、单走廊实验以及大场景建图实验的结果证明了该方法的鲁棒性和精确性。在面积为1050 m 2的复杂室内环境下,采用该方法建图误差为0.9%。展开更多
Mobile robot navigation in unknown environment is an advanced research hotspot.Simultaneous localization and mapping(SLAM)is the key requirement for mobile robot to accomplish navigation.Recently,many researchers stud...Mobile robot navigation in unknown environment is an advanced research hotspot.Simultaneous localization and mapping(SLAM)is the key requirement for mobile robot to accomplish navigation.Recently,many researchers study SLAM by using laser scanners,sonar,camera,etc.This paper proposes a method that consists of a Kinect sensor along with a normal laptop to control a small mobile robot for collecting information and building a global map of an unknown environment on a remote workstation.The information(depth data)is communicated wirelessly.Gmapping(a highly efficient Rao-Blackwellized particle filer to learn grid maps from laser range data)parameters have been optimized to improve the accuracy of the map generation and the laser scan.Experiment is performed on Turtlebot to verify the effectiveness of the proposed method.展开更多
A novel approach was presented to solve the navigation problem of autonomous mobile robots in unknown environments with dense obstacles based on a univector field method. In an obstacle-free environment, a robot is en...A novel approach was presented to solve the navigation problem of autonomous mobile robots in unknown environments with dense obstacles based on a univector field method. In an obstacle-free environment, a robot is ensured to reach the goal position with the desired posture by following the univector field. Contrariwise, the univector field cannot guarantee that the robot will avoid obstacles in environments. In order to create an intelligent mobile robot being able to perform the obstacle avoidance task while following the univector field, Dyna-Q algorithm is developed to train the robot in learning moving directions to attain a collision-free path for its navigation. Simulations on the computer as well as experiments on the real world prove that the proposed algorithm is efficient for training the robot in reaching the goal position with the desired final orientation.展开更多
In order to overcome the inherent oscillation problem of potential field methods(PFMs) for autonomous mobile robots in the presence of obstacles and in narrow passages,an enhanced potential field method that integrate...In order to overcome the inherent oscillation problem of potential field methods(PFMs) for autonomous mobile robots in the presence of obstacles and in narrow passages,an enhanced potential field method that integrates Levenberg-Marquardt(L-M) algorithm and k-trajectory algorithm into the basic PFMs is proposed and simulated.At first,the mobile robot navigation function based on the basic PFMs is established by choosing Gaussian model.Then,the oscillation problem of the navigation function is investigated when a mobile robot nears obstacles and passes through a long and narrow passage,which can cause large computation cost and system instability.At last,the L-M algorithm is adopted to modify the search direction of the navigation function for alleviating the oscillation,while the k-trajectory algorithm is applied to further smooth trajectories.By a series of comparative experiments,the use of the L-M algorithm and k-trajectory algorithm can greatly improve the system performance with the advantages of reducing task completion time and achieving smooth trajectories.展开更多
Autonomous navigation is a complex challenge that involves the interpretation and analysis of information about the scenario to facilitate the cognitive processes of a robot to perform free trajectories in dynamic env...Autonomous navigation is a complex challenge that involves the interpretation and analysis of information about the scenario to facilitate the cognitive processes of a robot to perform free trajectories in dynamic environments. To solve this, the paper introduces a Case-Based Reasoning methodology to endow robots with an efficient decision structure aiming of selecting the best maneuver to avoid collisions. In particular, Manhattan Distance was implemented to perform the retrieval process in CBR method. Four scenarios were depicted to run a set of experiments in order to validate the functionality of the implemented work. Finally, conclusions emphasize the advantages of CBR methodology to perform autonomous navigation in unknown and uncertain environments.展开更多
文摘在复杂连续环境下,强化学习系统的状态空间面临维数灾难问题,需要采取量化的方法,降低输入空间的复杂度。径向基神经网络(RBFNN:Radial Basis Function Neural Networks)具有较强的函数逼近能力及泛化能力,由此提出了基于径向基神经网络的Q学习方法,并将其应用于单机器人的自主导航。在基于径向基神经网络的强化学习系统中,用径向基神经网络逼近状态空间和Q函数,使学习系统具有良好的泛化能力。仿真结果表明,该导航方法具有较强的避碰能力,提高了机器人对环境的适应能力。
文摘在室内同时定位与建图(SLAM)的实际应用中,对称单一结构环境易造成激光SLAM错误建图,低质量光照或低纹理环境易造成视觉SLAM失效。针对上述室内退化环境,提出一种将激光、视觉、惯性测量单元(IMU)进行紧耦合的LVI-SLAM方法。在该方法前端,设计视觉评价环节对视觉信息置信度进行自适应调整;在该方法后端,进行位姿图优化以及多传感器回环抑制累积误差。视觉评价实验、单走廊实验以及大场景建图实验的结果证明了该方法的鲁棒性和精确性。在面积为1050 m 2的复杂室内环境下,采用该方法建图误差为0.9%。
基金National Natural Science Foundation of China(Nos.51475328,61372143,61671321)
文摘Mobile robot navigation in unknown environment is an advanced research hotspot.Simultaneous localization and mapping(SLAM)is the key requirement for mobile robot to accomplish navigation.Recently,many researchers study SLAM by using laser scanners,sonar,camera,etc.This paper proposes a method that consists of a Kinect sensor along with a normal laptop to control a small mobile robot for collecting information and building a global map of an unknown environment on a remote workstation.The information(depth data)is communicated wirelessly.Gmapping(a highly efficient Rao-Blackwellized particle filer to learn grid maps from laser range data)parameters have been optimized to improve the accuracy of the map generation and the laser scan.Experiment is performed on Turtlebot to verify the effectiveness of the proposed method.
基金Project(2010-0012609) supported by the Basic Science Research Program,Korea
文摘A novel approach was presented to solve the navigation problem of autonomous mobile robots in unknown environments with dense obstacles based on a univector field method. In an obstacle-free environment, a robot is ensured to reach the goal position with the desired posture by following the univector field. Contrariwise, the univector field cannot guarantee that the robot will avoid obstacles in environments. In order to create an intelligent mobile robot being able to perform the obstacle avoidance task while following the univector field, Dyna-Q algorithm is developed to train the robot in learning moving directions to attain a collision-free path for its navigation. Simulations on the computer as well as experiments on the real world prove that the proposed algorithm is efficient for training the robot in reaching the goal position with the desired final orientation.
基金Supported by the National Key Basic Research Program of China(973 Project)(No.2013CB035503)
文摘In order to overcome the inherent oscillation problem of potential field methods(PFMs) for autonomous mobile robots in the presence of obstacles and in narrow passages,an enhanced potential field method that integrates Levenberg-Marquardt(L-M) algorithm and k-trajectory algorithm into the basic PFMs is proposed and simulated.At first,the mobile robot navigation function based on the basic PFMs is established by choosing Gaussian model.Then,the oscillation problem of the navigation function is investigated when a mobile robot nears obstacles and passes through a long and narrow passage,which can cause large computation cost and system instability.At last,the L-M algorithm is adopted to modify the search direction of the navigation function for alleviating the oscillation,while the k-trajectory algorithm is applied to further smooth trajectories.By a series of comparative experiments,the use of the L-M algorithm and k-trajectory algorithm can greatly improve the system performance with the advantages of reducing task completion time and achieving smooth trajectories.
文摘Autonomous navigation is a complex challenge that involves the interpretation and analysis of information about the scenario to facilitate the cognitive processes of a robot to perform free trajectories in dynamic environments. To solve this, the paper introduces a Case-Based Reasoning methodology to endow robots with an efficient decision structure aiming of selecting the best maneuver to avoid collisions. In particular, Manhattan Distance was implemented to perform the retrieval process in CBR method. Four scenarios were depicted to run a set of experiments in order to validate the functionality of the implemented work. Finally, conclusions emphasize the advantages of CBR methodology to perform autonomous navigation in unknown and uncertain environments.
