无人采煤机器人的发展和远程作业系统外文文献翻译、中英文翻译

英文原文Development of an Unmanned Coal Mining Robot and a Tele-Operation SystemSungsik Huh , Unghui Lee, Hyunchul Shim, Jong-Beom Park and Jong-Ho NohAbstract: The typical underground coal mine in Korea is an extremely harsh environment for workers due to high temperature, humidity, and harmful dusts. In such environment, a remotely operated robot can be of great help to alleviate the work load of miners. As a joint effort, KAIST, Hydraumatics Co., and Korea Coal Corporation have developed a mining robot and a tele-operation system to operate the robot from a safe remote place. In this paper, we present the design of the robot mechanism and the sensing algorithms for localization and elevation mapping. The robot is hydraulically powered to drive the track and the cylinders in the excavation mechanism. The robot is equipped with two cameras, two laser scanners, and many other sensors so that the human operator can perform various tasks such as shoveling and breaking at a remotely located console. The data from various sensors are fused together to provide the human operators with enhanced visual cues as well as the pose information of the robot and its boom/arm/bucket linkage. The robot has been recently deployed in an active coal mine to evaluate its effectiveness in the real environment and demonstrated its feasibility as a viable aid to assist mining workers.Keywords: Field Robot, Unmanned Coal Mining, Monte Carlo Localization, Elevation mapping, Tele-operation1 IntroductionAs the world is faced with the depletion of petroleum, the energy from coal has received a renewed attention. Coal is more abundant than petroleum throughout the world and, among the few natural resources found in Korea, coal is the only energy source that is readily available in our nation for electricity generation or heating. However, coal mining in Korea is heavily relying on manual labors and therefore it is a typical example of difficult and dangerous industries. Coal mine workers are exposed to harsh working environment with high temperature (30 degrees Celsius), high humidity (90100%), noise, and dust. The workers may get injured by explosion or machineries such as chain conveyors, railroad cars, and powered breakers or drills. They can be also trapped in collapsed mine shafts at hundreds of meters deep down in the ground. As a result, the total number of coal mine workers is decreasing while the average age of the mining workers is increasing every year. Therefore, it has been suggested to delegate some of the hardest works to unmanned coal mining robots to alleviate the work load of mine workers while increasing the productivity.The primary objective of the robot developed in our research is to assist human workers by reducing their workload. Our mining robot is conceived to be operated both manually and autonomously under the supervision of remote operators in real-time. Also, in order to enhance the operators perception of the work area, three dimensional synthetic visual cues are generated by combining the information from the laser range finders and attitude sensors.In the following, we present the overall system of the coal mining robot being developed for impending deployment in active mining shafts various locations in Korea.2 ROBOT MECHANICAL SYSTEM2.1 Robot Mechanism Design The coal mining robot system consists of a mechanical part, a sensing part, and a remote control station. Firstly, the mechanical part is described in this section. The exterior view of the robot is shown in Fig. 1 and 2. The robots width is narrower than other excavators with similar ratings since it has to move around in narrow mine shafts in Korea (3.0 meters wide). The steel tracks are actuated individually to move on rough, muddy and rocky ground. The robot manipulator has 5 degree of freedom from the tracks to the bucket, so that the robot can perform various tasks such as digging and sweeping using its bucket attached at the end of the arm and boom assembly.The hydraulic system consists of two hydraulic motors, cylinders, a pump, a filter, a reservoir, an electric motor, solenoid valves, and a valve controller. The hydraulic motor and cylinder transform the hydraulic pressure into angular and linear motion for locomotion and manipulation through tracks, boom, arm, bucket, and dozer. The hydraulic pump maintains hydraulic pressure, and the solenoid valves are actuated by the valve controller. The control inputs of joysticks in remote place are sent to valve controller either directly or through the computer for more autonomous operations. Table 1 shows the mechanical specification of the robotFig. 1 The front and top view of the robot CAD model Fig. 2 The coal mining robotTable 1 Mechanical specification of the robotDimension(Length/Width/Height)1.78x0.75x1.36 mWeight2500 kgMax. Driving Torque300 kgf.mHydraulic Pressure200 barElectric Motor Power20 HPHydraulic Pump Capacity45 l/minMax. Velocity/Climbing Capability1.2 km/h / 20 deg2.2 Sensors and CommunicationThe sensor suite of the robot consists of two laser scanners, an AHRS(Attitude and Heading Reference System), a rotary encoder, two inclinometers, two infrared cameras, a computer and a wireless LAN adaptor. The electronic components are installed on the upper body of the robot to acquire the sensor data and communicate with the operator. The details of those sensors and a computer are listed in Table 2.Fig. 3 shows the overall sensors and communication structure of the robot. Control signals are transmitted from the operators joysticks to valve controller and robot actuators. It is transmitted on a wireless CAN channel separated from the telemetry system in order to provide its independent operation.Sensor data are sent back through wireless LAN port(IEEE 802.11g/n) so that the operator can receive the high-bandwidth information including two onboard video channels. Two video streaming channels from the onboard cameras deliver the real-time views of left- and right-side of the bucket.Table 2 Sensor specificationComputerNational Instrument cRio-9022、533MHz Processor、256MB Memory、 2GB StorageLaser ScannerSICK LMS291-S05AHRSMicrostrain 3DM-GX3IR CameraSamsung Techwin SCO-2080RRotary EncoderAutonics EPM50InclinometerDAS DM1-70Fig. 3 Sensors and communication structure of the robot and the remote control station2.3 Tele-Operation SystemThe tele-operation is implemented by the remote control station as depicted in a diagram at Fig. 3. The operator can control the robot by either two distinct joystick modules, fixed or portable. These joystick modules are instantly switchable between each other by a takeover button on the robot. The operator can choose which joystick module to use. Each joystick module has four 2-axes joysticks to control various parts on the robot and an emergency stop button. They are linked to robot manipulatorstilting motion of bucket, arm, boom and dozer, boom rolling motion, body rotating and track traveling. The portable joystick module allows the operator to directly access the robot in a visual range. It can be used for system checkup or in a situation when the operator needs to be present at the work area to perform more delicate tasks.The joystick module on the remote console is the primary area to operate the robot in a more comfortable environment away from the mining area. After switching to fixed joystick module, the operator can control robot manipulators while looking at the video displays. The remote control station deployed in a recent field test is shown as Fig. 4. It has two monitors that show the onboard video(Fig. 5) and the status of the robot. The status display includes three-dimensional graphic rendering of the robot and obstacles in the work area to show the robot pose and the linkages with respect to the environment since it is hard to sense the depth information from the camera view only. The environment map around the robot is constructed by the laser scan data and attitude sensors. The mapping technique is explained in the following section.Fig. 4 The remote control station Fig. 5 An onboard camera image3 ROBOT POSE ESTIMATION AND ENVIRONMENT MAPPING3.1 Robot LocalizationIn order to estimate position and attitude of the robot, a Monte Carlo Localization(MCL) algorithm using a particle filter is implemented. We assume that the ground is sufficiently flat and therefore the MCL can be applied to the case of 2-D space, and the robot states to be estimated are 2-D position xt , yt and heading t .The probabilistic model of localization is written as below: （1） （2）Where is pose of the robot and m is a local 2-D map data measured in advance. Measured distances and bearing from the vehicle to objects and motion increment can be measured by using a laser scanner and an odometer, respectively. However, odometry of track vehicles like this robot is not accurate due to excessive slip of the track. Therefore, the localization algorithm ignores the motion update term by odometry but uses only the random sampling term. The robot pose posterior can be derived as a Bayesian filter.5 （3）where the likelihood of each laser measurement is （4）An occupancy grid map for localization is built by the horizontal laser scanner in advance since a local 2-D map of mine shafts for daily work is not complicated or large to use such a large-scale mapping technique. Once the map is built, the MCL algorithm is started to estimate robots position and heading. The MCL result using about 80 particles provides real-time 2-D pose data for 3-D elevation mapping process in Section 3.3.The two-dimensional map used for localization is not much affected by mining work. In a typical mine shafts which has the width of 3.0m and the height of 2.4m, the horizontal laser scanner is located at 1.4m high above the ground, and mining works such as digging or sweeping-out is usually operated under the height of horizontal laser scanner. In addition, the localization algorithm does not count obstructed region by the robot manipulator since the obstructed region can be calculated using pose of manipulators sensed by inclinometers.3.2 Robot Manipulator Pose EstimationIn order to estimate pose of manipulators, several sensors are installed at boom, arm and bucket joint as shown in Fig. 6. A rotary encoder sensor is attached at a rotating joint between the arm and the bucket. The rotary encoder is an absolute-type so that the bucket does not have to be initialized after turning on the robot. Two inclinometer sensors that can sense two-axes rotation(pitch and roll) are attached at the middle of the boom and the arm parts. Therefore, we can estimate the pose of the whole manipulators using forward kinematics of robot, and the pose data are integrated with the robot position to show the 3-D graphics of the robot.Fig. 6 Sensors on robot: Two IR cameras, two laser scanners(left), a rotary encoder and two inclinometers to estimate pose of the robot and its manipulators(right)3.3 Duplex Elevation MappingIn order to scan the surrounding three-dimensional environment, we use a laser scanner vertically installed on the robot. The laser scanner scans 180 degrees vertically and generates 361 points per scan from the ground to the front, and finally to the ceiling of mine shafts as depicted in Fig. 7. These scanned points including bearing r and distance r are transformed to height information of the environment with regard to the robot position and attitude from MCL algorithm.After transformed to height data, they are divided into two elevation maps which are an upper ceiling map mU and a lower ground map mL . The two maps have same length, width and cell size, and they are parallel to the ground plane initially. The height data update cells in both elevation maps every time step, and the maximum number of updated cells in an elevation map at a time step is smaller than 181 cells since vertical obstacles (e.g., walls) reflect more than one scan point. We call this technique updating both elevation maps using a vertical scanning laser measurement as “Duplex elevation mapping”, and the procedure is listed in Algorithm 1. In this algorithm, height of the robot is fixed at ht as assumed that the ground is sufficiently flat to be applied to 2-D localization.Fig. 7 Scan range diagram of the vertical laser scanner(top), side view of the upper and the lower elevation map built by the laser scanner(bottom)Obviously, these elevation maps have a limitation that cannot express horizontally protruded objects in the 3-D space. However, there are few or no horizontally protruded objects in the work area which is a blind end of mine shafts, so that building two elevation maps is more efficient than building a full 3-D space map in real-time requiring heavy computational loads. The duplex elevation mapping result that merged by two elevation maps is shown in Fig. 8. Fig. 8 Synthetic 3-D graphic rendering by duplex elevation mapping4 CONCLUSIONIn this paper, we presented an unmanned coal mining robot in the subterranean environment by a remote control system. The developed mechanical and hydraulic system can power the robot to actuate tracks, manipulators, and the sensor system. The robot can estimate its pose and that of manipulators by sensor data and implemented localization algorithm. Measured data acquired from the sensors on the robot is transmitted back to the operator at the remote place so that the operator can monitor and control the robot easily through the synthetic 3-D graphic rendering merged by two elevation maps. The mining robot will be upgraded further to perform in a more autonomous manner with less direct control of the remote operator.REFERENCES1 D. W. Hainsworth, “Teleoperation User Interfaces for Mining Robotics”, Autonomous Robots, Vol. 11, pp. 1928, 2001. 2 J. N. Bakambu, V. Polotski, “Autonomous System for Navigation and Surveying in Underground Mines”, Journal of Field Robotics, Vol. 24, No. 10, pp. 829-847, 2007. 3 S. Thrun, S. Thayer, W. Whittaker, C. Baker, W. Burgard, D. Ferguson, D. Hahnel, M. Montemerlo, A. Morris, Z. Omohundro, C. Reverte, W. Whittaker, “Autonomous Exploration and Mapping of Abandoned Mines”, IEEE Robotics and Automation Magazine, Dec. 2004.4 A. Nuchter, H. Surmann, K. Lingemann, J. Hertzberg, S. Thrun, “6D SLAM with an Application in Autonomous Mine Mapping”, Proc. of the IEEE International Conference on Robotics & Automation, 2004. 5 S. Thrun, W. Burgard, D. Fox, Probabilistic Robotics, MIT Press, 2005. 中文译文无人采煤机器人的发展和远程作业系统Sungsik Huh , Unghui Lee, Hyunchul Shim, Jong-Beom Park and Jong-Ho Noh摘要：在韩国，对于煤矿工人来讲，煤矿井下是一个典型的高温，湿度大和有害粉尘多的极度恶劣的环境。在这样的环境中，远程操作机器人，可以有很大的帮助，以减轻矿工的工作负荷。作为科学技术院，经过Hydraumatics有限公司和韩国煤炭公司的共同努力，发明了采矿机器人，可以在一个安全偏僻的地方运用远程操作系统操作机器人。在本文中，我们提出了机器人的机制、定位和高程测绘遥感算法的设计。机器人以液压动力驱动在轨道上行走，并在挖掘机制的气瓶。该机器人配备了两个摄像头，两个激光扫描仪，以及其他许多传感器，使人类操作员可以执行各种任务，如在远程控制台进行铲和破碎。从各种传感器收集到的数据融合在一起，增强视觉线索提供人的运营商以及姿态信息的机器人臂/手臂/桶联动。最近煤矿机器人已被部署在真实的环境，以评估其有效性，并证明其可行性，作为一种可行的援助，以协助采矿工人。 关键词：场机器人，无人采煤，蒙特卡罗定位，高程映射，远程操作1 引言作为世界面临着枯竭的石油，煤能源得到重新重视。在世界各地煤炭是较丰富的能源相对于石油，其中在韩国发现的一些自然资源，煤炭是唯一很容易在韩国用于发电或供热的能量来源。然而，韩国的煤炭开采在很大程度上依靠手工劳动，因此它是一个典型的伴随着高温（ 30摄氏度），高湿度（90100），噪声，粉尘，煤矿工人暴露于恶劣的工作环境的困难和危险的行业。工人可能受伤因爆炸或机械设备损坏，如链式输送机，火车车厢，供电断路器或演习。他们也可能被困在倒塌的矿井几百米深处的井下。结果，煤矿工人的总数正在减少，而采矿工人的平均年龄正在逐年增加。因此，它已被委派做一些最难的工作，以减轻对煤矿工人的工作负荷，同时提高生产力。在我们的研究开发机器人的主要目的是协助减少其人力工人工作量。我们的采矿机器人设想手动和自主实时远程操作的情况下运作。此外，为了提高工作区操作者的洞察力，通过激光测距仪和姿态传感器的信息合成三维视觉。接下来，我们目前正在为将积极部署在不同的挖掘地点在韩国开发的煤炭开采机器人系统的整体。2 机器人机械系统2.1 机器人机构设计煤矿机器人系统由机械部分，传感部分，以及一个遥控站。首先，机械部分是本节中描述。机器人的外观，如图1和2。机器人的宽度比其他类似等级的挖掘机窄，因为它有走动在韩国狭窄的矿井（3.0米宽）。钢轨道分别被驱动在粗糙，泥泞和岩石地面上移动。机器人可以从5个不同角度进行工作从链轮到铲斗，这样可以使机器人执行各种任务，如用它附着在斗臂和液压马达一端的铲斗进行挖掘和清扫。液压系统由两个液压马达，油缸，泵，过滤器，一个水库，一个电动马达，电磁阀，阀门控制器。液压马达和气缸转换成角和直线运动和操纵，通过轨道，发动机，斗臂，铲斗，和推土机液压装置。液压泵控制液压和电磁阀，阀门控制器控制驱动。在偏远的地方用操纵杆控制输入，直接发送到阀门控制器或通过电脑更多的自主行动。表1显示了机器人的机械规格。 图1前面和顶视图机器人CAD模型 图2煤炭采矿机器人表1 机器人的机械规格尺寸（长/宽/高）1.78x0.75x1.36m重量2500kg最大驱动扭矩300kgf.m液压200bar电机功率20HP液压泵容量45升/分钟最大(速度/爬坡角度)1.2km/h/ 202.2 传感器和通信机器人的传感器套件由两个激光扫描仪，航姿系统（姿态和航向参考系统），旋转编码器，两个倾角罗盘，两个红外摄像机，一台电脑和一个无线局域网适配器。电子元件被安装在机器人上半身来获得传感器数据。表2列出了这些传感器和一台电脑的细节。图3显示了整体的传感器和通信结构的机器人。从运营商的操纵杆来控制信号的传输阀门控制器和机器人驱动器。它是无线传输通道从遥测系统中分离出来，以提供其独立运作。传感器数据被送回通过无线局域网端口（IEEE的802.11g / n）的，使操作者可以接收高带宽的信息，包括两个板载视频通道。内建相机的两个视频流渠道，提供实时的左侧和右侧铲斗的信息。 表2 传感器规格计算机美国仪器crio-9022、533MHz处理器、256 MB内存、2GB存储激光扫描仪SICK LMS291-S05 航姿系统微应变3DM-GX3红外摄像机三星Techwin SCO-2080R旋转编码器奥托尼克斯EPM50 倾角罗盘DAS DM1-70图3传感器和通信结构的机器人和遥控站2.3 远程作业系统描绘在图.3遥控站实施远程操作。操作员可以控制任何两个不同的操纵杆模块，固定或便携式机器人。这些操纵杆模块是彼此之间瞬间切换，通过对机器人的收购按钮。操作者可以选择使用操纵杆模块。每个操纵杆模块有4个2轴操纵杆来控制机器人和一个急停按钮上的各个部分。它们与机器人倾斜桶，手臂，繁荣和推土机，繁荣滚动运动，身体旋转和轨道行驶的议案。便携式的操纵杆模块允许操作员直接访问在可视范围内的机器人。它可以用于系统的检查或情况时，运营商需要在目前的工作领域进行更细腻的任务。在远程控制台上的操纵杆模块是操作机器人在一个更舒适的环境，远离矿区的主要领域。切换到固定操纵杆模块后，操作员可以控制机器人，而在视频显示。遥控站部署在最近的实地测试，如图.4所示。它有两个显示器显示板载视频（图5）和机器人的状态。状态显示，包括机器人三维图形渲染和障碍，在工作区中显示机器人的构成和环境方面的联系，因为它是很难感受到的深度信息，只能从相机视图。由激光扫描数据和姿态传感器的机器人周围的环境地图构建。映射技术在下一节解释。图4遥控站图 图5内建摄像头图像3 机器人姿态估计和环境贴图3.1 机器人定位以估计的立场和态度的机器人，蒙特卡罗本地化（MCL）使用粒子滤波算