A two-layered subgoal based mobile robot navigation algorithm with vision system and IR sensors

The present work describes the real-life implementation of a mobile robot navigation scheme where vision sensing is employed as primary sensor for path planning and IR sensors are employed as secondary sensors for actual navigation of the mobile robot with obstacle avoidance capability in a static or dynamic indoor environment. This two-layer based, goal-driven architecture utilizes a wireless camera in the first layer to acquire image and perform image processing, online, to determine subgoal, employing a shortest path algorithm, online. The subgoal information is then utilized in the second layer to navigate the robot utilizing IR sensors. Once the subgoal is reached, vision based path planning and IR guided navigation is reactivated. This sequential process is continued in an iterative fashion until the robot reaches the goal. The algorithm has been effectively tested for several real-life environments created in our laboratory and the results are found to be satisfactory.     

基于两轮机器人运动模型的Kalman滤波视觉导航

在视觉算法对目标物识别的基础上,对两轮移动机器人运动模型进行分析,提出将其作为扩展卡尔曼滤波的系统状态方程,并应用到视觉导航系统中。仿真实验及实际测试表明,方法能较好地对两轮移动机器人的运动实现预测和估计,减轻环境干扰影响、降低处理计算量并提高系统实时性,满足了机器人在复杂环境下视觉导航的要求。     

Visual feedback control of a robot in an unknown environment (learning control using neural networks)

In this paper, a visual feedback control approach based on neural networks is presented for a robot with a camera installed on its end-effector to trace an object in an unknown environment. First, the one-to-one mapping relations between the image feature domain of the object to the joint angle domain of the robot are derived. Second, a method is proposed to generate a desired trajectory of the robot by measuring the image feature parameters of the object. Third, a multilayer neural network is used for off-line learning of the mapping relations so as to produce on-line the reference inputs for the robot. Fourth, a learning controller based on a multilayer neural network is designed for realizing the visual feedback control of the robot. Last, the effectiveness of the present approach is verified by tracing a curved line using a 6-degrees-of-freedom robot with a CCD camera installed on its end-effector. The present approach does not necessitate the tedious calibration of the CCD camera and the complicated coordinate transformations. This revised version was published online in October 2004 with a correction to the issue number.     

New hybrid vision-based control approach for automated guided vehicles

Automated guided vehicles (AGVs) are a common choice made by many companies for material handling (MH) in manufacturing systems. AGV-based internal transport of raw materials, goods, and parts is becoming improved with advances in technology. Demands for fast, efficient, and reliable transport imply the usage of the flexible AGVs with onboard sensing and special kinds of algorithms needed for daily operations. So far, the majority of these transport solutions have not considered the modern techniques for visual servoing, monocular SLAM, and consequently, the usage of camera as onboard sensor for AGVs. In this research, a new hybrid control of AGV is proposed. The main control algorithm consists of two independent control loops: position-based control (PBC) for global navigation and image based visual seroving (IBVS) for fine motions needed for accurate steering towards loading/unloading point. By separating the initial transportation task into two parts (global navigation towards the goal pose near the loading/unloading point and fine motion from the goal pose to the loading/unloading point), the proposed hybrid control bypasses the need for artificial landmarks or accurate map of the environment. The state estimation of the robot pose is determined in terms of monocular SLAM, via extended Kalman filter coupled with feedforward neural network—the neural extended Kalman filter (NEKF). NEKF is used to model unknown disturbances and to improve the robot state transition model. The integration of the new hybrid control and NEKF has been tested in laboratory with the mobile robot and simple camera. Experimental results present the effectiveness of the proposed hybrid control approach.     

Corridor navigation of the mobile robot using image based control

In this paper, the wall following navigation algorithm of the mobile robot using a mono vision system is described. The key points of the mobile robot navigation system are effective acquisition of the environmental information and fast recognition of the robot position. Also, from this information, the mobile robot should be appropriately controlled to follow a desired path. For the recognition of the relative position and orientation of the robot to the wall, the features of the corridor structure are extracted using the mono vision system, then the relative position, the offset distance and steering angle of the robot from the wall, is derived for a simple corridor geometry. For the alleviation of the computation burden of the image processing, the Kalman filter is used to reduce search region in the image space for line detection. Next, the robot is controlled by this information to follow the desired path. The wall following control scheme by the PD control scheme is composed of two control parts, the approaching control and the orientation control, and each control is performed by steering and forward-driving motion of the robot. To verify the effectiveness of the proposed algorithm, the real time navigation experiments are performed. Through the result of the experiments, the effectiveness and flexibility of the suggested algorithm are verified in comparison with a pure encoder-guided mobile robot navigation system.     

基于BP神经网络的小型组足球机器人视觉子系统的图像分割技术

为解决小型足球机器人视觉子系统图像分割的实时性和光照适应性问题,将BP神经网络技术应用到图像分割中.在图像分割技术和BP神经网络的理论分析基础上,建立了两者之间的关系,并建立了相应的BP神经网络模型.图像像素离散化并将其H、Cb、Cr分量值作为神经网络的输入,将目标像素点分类类别作为神经网络的输出.通过改进神经网络学习...     

基于多帧差分与自适应技术的运动目标检测方法研究

提出了动态环境下适用于移动机器人视觉导航的运动目标检测方法,将多帧差分与自适应背景剪除技术相结合,考虑了背景的运动变化和对移动机器人的运动进行了补偿,优化了图像采集频率和摄像机运动速度之间的参数,并选取相应的最优帧数进行差分。该方法获得了准确、完整的检测结果,提高了检测的实时性,可用于视觉导航的智能移动机器人系统。     

Joining different metallic sheets without protrusion by flat hole clinching process

In this paper, we propose an architecture based on an artificial neural network (ANN), to learn welding skills automatically in industrial robots. With the aid of an optic camera and a laser-based sensor, the bead geometry (width and height) is measured. We propose a real-time computer vision algorithm to extract training patterns in order to acquire knowledge to later predict specific geometries. The proposal is implemented and tested in an industrial KUKA KR16 robot and a GMAW type machine within a manufacturing cell. Several data analysis are described as well as off-line and on-line training, learning strategies, and testing experimentation. It is demonstrated during our experiments that, after learning the skill, the robot is able to produce the requested bead geometry even without any knowledge about the welding parameters such as arc voltage and current. We implemented an on-line learning test, where the whole experiments and learning process take only about 4 min. Using this knowledge later, we obtained up to 95 % accuracy in prediction.     

基于置信传播的多运动平台随机组网协同导航

在协同网络中,随着导航个体数目的增加,如何确定参与组网的导航平台并获取其状态与观测量的数据关联性是实现 小型规模的多运动平台协同导航的关键。 本文提出一种基于置信传播的多运动平台随机组网协同导航方法,采用随机有限集 对状态和量测量进行建模并附加标签,构造带有标签的多伯努利粒子滤波器,采用基于置信传播的多运动平台随机组网协同导 航方法,并利用运动目标平台与基站的绝对观测、运动目标平台与临近平台相对观测融合,通过粒子滤波器和置信传播对系统 中的量测信息进行处理,相比较于非参数置信传播算法扩展因子结构作为置信度算法的逼近,该算法实现概率数据关联进行导 航系统状态估计。 仿真和物理实验结果表明:非参数置信传播算法在不同基站和不同粒子数的解算结果性能较差;本文提出的 算法受基站个数和粒子个数的影响较小鲁棒性好,收敛性好,均方根误差不高于 0. 3 cm,精度高于前者一个数量级,该算法能 够有效获取导航平台的位置信息。     

基于神经网络的立体视觉摄像机标定

摄像机标定是立体视觉研究的重要组成部分。根据神经网络与摄像机标定的特点,提出了基于神经网络的双目立体视觉摄像机标定方法,构建了三种不同的神经网络,并在试验中利用BP算法和实际数据对各种方法进行了分析和比较。     

状态延迟输入神经网络及其在机器人定位监督控制中的应用

对复合输入动态递归网络作了改进 ,提出一种新的动态递归神经网络结构 ,称为状态延迟输入动态递归神经网络 (State Delay Input Dynamical Recurrent Neural Networks)。这种具有新的拓扑结构和学习规则的动态递归网络 ,不仅明确了各权值矩阵的含义 ,而且使权值的训练过程更为简洁 ,意义更为明确。网络增加了输入输出层前一步的状态信息 ,使其收敛速度和泛化能力与其他常用网络结构相比 ,均有明显提高 ,增强了系统实时控制的可能性。本文将该网络用于机器人定位监督控制系统中 ,通过利用神经网络建立起被控对象的逆模型 ,与传统 PD控制器结合 ,确保了控制系统的稳定性 ,有效地提高系统的精度和自适应能力。仿真结果表明了这种改进的有效性和优越性     

移动相机下基于三维背景估计的运动目标检测

室内环境中的运动目标检测是计算机视觉领域的研究热点,而移动相机造成的动态背景是运动目标检测的难点。本文提出一种基于同步定位与地图创建(ORB-SLAM)三维背景估计的运动目标检测算法,首先使用移动相机遍历整个室内环境,采用ORB-SLAM技术建立当前全局环境的三维背景特征点云模型;然后基于局部视频建立局部三维特征点云,根据定位信息将当前局部三维特征点云与环境三维背景特征点云进行嵌入,基于环境背景信息,采用三维均值漂移(3DMS)算法,对局部三维特征点云进行前景特征点提取;运用深度卷积神经网络,对提取的前景特征点所在候选区域进行运动目标确认。通过多个室内场景的实际实验进行验证,结果表明本文方法具有较高的运动目标检测准确率和召回率,提出的运动目标检测算法充分利用了三维背景信息,采用深度卷积神经网络进行确认,有效地改善了检测的准确性和鲁棒性。     

Development of autonomous mobile robot with manipulator for manufacturing environment

This paper describes the developmental effort involved in prototyping the first indigenous autonomous mobile robot, AMR, with a manipulator for carrying out tasks related to manufacturing. The objective is to design and develop a vehicle that can navigate autonomously and transport jobs and tools in a manufacturing environment. Proprioceptive and exteroceptive sensors are mounted on AMR for navigation. Among the exteroceptive sensors, a stereovision camera is mounted in front of AMR for mobile robot perception of the environment. Using the widely supported JPEG image file format, full high-resolution color images are transmitted frame by frame from the mobile robot to multiple viewers located within the robot work area, where fast reconstruction of these images enables remote viewing. A CMOS camera mounted on the manipulator identifies jobs for pick-and-place operation. A variation of correlation based adaptive predictive search (CAPS) method, a fast search algorithm in template matching, is used for job identification. The CAPS method justifiably selects a set of search steps rather than consecutive point-to-point search for faster job identification. Search steps, i.e., either coarse search or fine search, are selected by calculating the correlation coefficient between template and the image. Adaptive thresholding is used for image segmentation for parametric calculations required for proper gripping of the object. Communication with the external world allowing remote operation is maintained through wireless connectivity. It is shown that autonomous navigation requires synchronization of different processes in a distributed architecture, while concurrently maintaining the integrity of the network.     

一种基于视觉跟踪特征点的室内导航方法

提出一种基于视觉的跟踪特征点的室内导航方法,即在地面上布置具有几何位置关系的特征点序列,利用单目摄像头采集图像,通过图像处理提取跟踪特征点的坐标值,再利用计算机图像坐标系和世界坐标系之间的转换关系,计算出移动机器人在世界坐标系中相对特征点的偏移距离和偏移角作为反馈,来控制移动机器人,以达到室内导航的目的。     

全天候移动车间巡检机器人目标定位算法

全天候移动车间巡检机器人移动轨迹复杂,为获取高精度的巡检机器人目标定位结果,提出一种全天候移动车间巡检机器人目标定位算法。优先标定得到移动车间环境的相机,获取相机参数,通过高低纹理匹配完成移动车间环境重建。然后通过相机内外参数将匹配点的图像坐标和世界坐标相关联,以此为依据估计巡检机器人的位姿。最终将得到的移动车间环境地图和周围数据相结合,采用粒子滤波算法对全天候移动车间巡检机器人位置组建的粒子群集合优化处理,通过不断迭代更新,输出目标定位结果。结果表明,所提算法可以有效降低巡检机器人目标定位时间以及联合定位误差,获取准确率更高的目标定位结果。     

模糊先验引导的高效强化学习移动机器人导航

针对移动机器人仅使用 RGB-D 相机在室内场景下执行坐标点主动视觉导航任务,提出了一种基于模糊先验引导的高效强化学习控制算法,设计了一种先验与强化学习的融合控制方式.结果表明,所提出的融合控制方法能够在确保较好的导航准确率基础上,避免训练中无意义的探索并缩短强化学习训练时长,实现强化学习策略的高效学习.     

基于单目视觉直线跟踪的SLAM实现

同时定位和地图生成问题(SLAM)是移动机器人的一个基本问题,有着广阔的应用背景.由于视觉传感器具有成本低、信息丰富的特点,是当前SLAM问题研究的一个热点.针对视觉传感器视野小、远距离观测精度低的问题,提出了一种基于直线跟踪的单目视觉SLAM算法.该算法通过对直线特征的近距离观测和跟踪,保证了定位的精度和地图生成的效率.在Pioneer3dx机器人平台上进行的实验研究表明,该算法尤其适合于结构化环境几何地图的构建.     

Robust and efficient vision system for group of cooperating mobile robots with application to soccer robots

In this paper a global vision scheme for estimation of positions and orientations of mobile robots is presented. It is applied to robot soccer application which is a fast dynamic game and therefore needs an efficient and robust vision system implemented. General applicability of the vision system can be found in other robot applications such as mobile transport robots in production, warehouses, attendant robots, fast vision tracking of targets of interest and entertainment robotics. Basic operation of the vision system is divided into two steps. In the first, the incoming image is scanned and pixels are classified into a finite number of classes. At the same time, a segmentation algorithm is used to find corresponding regions belonging to one of the classes. In the second step, all the regions are examined. Selection of the ones that are a part of the observed object is made by means of simple logic procedures. The novelty is focused on optimization of the processing time needed to finish the estimation of possible object positions. Better results of the vision system are achieved by implementing camera calibration and shading correction algorithm. The former corrects camera lens distortion, while the latter increases robustness to irregular illumination conditions.     

A QP Artificial Neural Network inverse kinematic solution for accurate robot path control

In recent decades, Artificial Neural Networks (ANNs) have become the focus of considerable attention in many disciplines, including robot control, where they can be used to solve nonlinear control problems. One of these ANNs applications is that of the inverse kinematic problem, which is important in robot path planning. In this paper, a neural network is employed to analyse of inverse kinematics of PUMA 560 type robot. The neural network is designed to find exact kinematics of the robot. The neural network is a feedforward neural network (FNN). The FNN is trained with different types of learning algorithm for designing exact inverse model of the robot. The Unimation PUMA 560 is a robot with six degrees of freedom and rotational joints. Inverse neural network model of the robot is trained with different learning algorithms for finding exact model of the robot. From the simulation results, the proposed neural network has superior performance for modelling complex robot’s kinematics.     

一种基于ARM嵌入式系统的低成本小型移动机器人平台

文章介绍一种应用常见经济型器件构建的移动机器人平台,它满足低成本、装配简单、可扩展性好等要求。我们选择了高速低功耗的ARM芯片作为处理器,为机器人设计了丰富的功能,并与上位机视觉定位和控制系统结合,使其适用于导航与定位、运动控制策略、多机器人系统体系结构与协作机制等领域研究。