In this paper, a quadrotor helicopter with microphones is considered for the sound source localization from the sky. Since the noise emitted by the helicopter itself is loud, it is difficult to capture the target sound signal clearly and in order to obtain accurate sound source position information, it is necessary to take the uncertainty in the measurement into account. Recursive Bayesian Estimation is utilized to handle the uncertainty and this paper proposes to introduce predictive control method to search the sound source so that the helicopter is possible to localize the target accurately.

Robots and unmanned aerial vehicles, or drones, are the hot research topics aiming at application to disaster response, infrastructure inspection, transportation, etc. 2.4GHz band wireless LAN is used for most of the control link of those robots because of cost and device availability. However, the wireless LAN protocol is not suitable for robot remote control because of fluctuation of transmission delay time and interference problem. Therefore, in this research, we have developed a multi-hop wireless network specialized for remote control link of robots and drones out of visual line-of-sight for future robot society. This paper presents the design and performance of prototype system obtained in field test.

近年，災害対応ロボットやドローンの研究開発が盛んに行われている．ロボットやドローンの安全運行管理に貢献することをめざし，ドローン間及び中空～低空を飛行する可能性のあるヘリや小型有人飛行機等と位置情報を共有するためのシステムとして．我々は920MHz帯を用いて各局のGPS位置情報を相互に共有する放送型システム「ドローンマッパー」を開発した．本システムは既存規格である920MHz帯特定小電力無線を活用したシステムであり，見通し外の飛しょう体位置を検知するマルチホップ機能も実装する．タブレット型ディスプレイに周囲約の1～2kmの範囲で飛行するドローンや有人ヘリの位置をリアルタイムで表示できる．本稿では，ドローン間位置情報共有ネットワークの概要やシステム機能・性能及び測定実験を報告する．

This paper describes an extrinsic parameter calibration of light detection and ranging (LiDAR) for a 3D measurement system using an unmanned aerial vehicle (UAV). Recently, the use of small UAVs becomes widespread in 3D measurements because small UAVs are highly useful. We target 10 cm accuracy of 3D measurement aiming for use in disaster environments. 3D measurements using small UAV need estimating not only accurate position and attitude of UAV but also accurate extrinsic parameters of the LiDAR. We propose a calibration method of extrinsic parameters of the LiDAR using reflectors. The proposed method estimates the angle of the LiDAR that minimizes the RMS error of the distance between the reference points of the reflectors and the measured 3D points. As the result of the calibration, vertical RMS error of 3D measurements was 0.0408 m and horizontal RMS error was 0.0472 m.

This paper describes a precise attitude estimation method for a 3D measurement system using LiDAR on a small unmanned aerial vehicle (UAV). Recently, the use of small UAVs becomes widespread in 3D measurements because small UAVs are highly useful. In 3D measurement using LiDAR, the precise attitude of the UAV is necessary because attitude estimation errors influence the 3D measurement accuracy. However, it is difficult to estimate the attitude of small UAVs because they are too small to be equipped with large and precise inertial measurement unit (IMU). Therefore, we proposed a method using six global navigation satellite systems (GNSS) receivers and a low-cost and light-weight IMU to estimate the precise attitude. The method combines Euler angles calculated statistically using six GNSS receivers with angular velocities measured by the IMU. As a result of evaluation, the proposed method is effective on improvement of the attitude estimation accuracy.

Due to the frequent occurrence of large-scale disasters such as earthquakes, tsunamis, volcanic eruptions, and floods of rivers, the demand for emergency response, restoration, and preventive disaster using robot technology has been increasing. One of the robot technologies is grasp of the damage situation using a flying robot which has been developed rapidly in recent years. Therefore, in this research, using images of the disaster site obtained from the flying robot and the terrain database which stores 3-D data before the disaster occurs, we aim to efficiently detect the collapse of electric poles, which are artificial objects, and the water level difference before and after the damage of rivers which are natural objects. By detecting these two disaster situations, we show the validy of our methods to grasp the damage situation using the terrain database.

UAV has been attracting attention for the purpose of collecting information in disaster area. Therefore, we are required to create a fast and accurate 3D map using the UAV. During autonomous flight, we create a 3D map and estimate position of the vehicle using SLAM (Simultaneous Localization and Mapping) because it is impossible to use the GPS in indoor environment. It is possible to navigate while using obstacle position by 3D SLAM. However, in a typical UAV configuration, since it is mainly observed by the front camera, detection of an object directly under the UAV is not always fast and accurate. Therefore, we propose a method to detect obstacles under the UAV quickly by matching the information of the bottom cameras with the point cloud information obtained by SLAM. This aims for fast and robust navigation with cheap and lightweight equipment.

In 2015, Field Robotics Laboratory at Tohoku University had joined a competition named “ARGOS” which stands for Autonomous Robot for Gas and Oil Sites. It is a challenge to develop a robot that can autonomously patrol and inspect a gas and oil plant site. During the second competition at the competition site, the robot “AIRK” had a problem that it could not detect a suspended obstacle. It represented off-limit region, and it located higher than the field of view of the laser scanner. To detect such obstacle with minimum remodeling, the authors decided to use the onboard forward-facing camera and a help of computer vision to detect an obstacle that was invisible to the laser scanner. In this paper, the basic ideas and evaluation results of the obstacle detection method are introduced.

In teleoperation of heavy machines, work efficiency is lower than in ordinary manned operations. One of the reasons is that it is difficult for operators to imagine the situation of work sites. Thus, we proposed views to acquire spatial knowledge before work based on the mechanism of spatial cognition. Humans have cognitive maps which can be divided into survey and route perspective. In this study, we focus on the route perspective because operators usually watch a cab view. It is knowledge acquired from subjective viewpoint and can be easily acquired from active movement. Therefore, we decided views which are from operator's viewpoint and can be changed by operator's intention as views to acquire route perspective. We conducted experiments using a virtual reality simulator to evaluate the proposed views. The results indicated that the proposed views could decrease the number of stops and increase speed in grasp.

In order to maneuver a human-operated work machine efficiently and safely, the operator is required to have not only operational skill to control the machine as intended, but also planning skill to determine operational strategy such as safe operation and efficient procedure. However, there are few studies on quantifying planning skill. In this paper, planning skill for machine operators is tried to quantify by focusing on operational procedure and attention. We create two experiments to derive feature value for operational procedure and attention. The results showed that shorter travel distance of the end-point of the manipulator for procedure and larger number of cooperated joints for attention leads to improving work efficiency.

This paper presents a virtual simulator that supports the design of a lane for cyclist and pedestrian (LPB). For the background of increasing the number of accidental contact on LPB, the authors assumed that LPB have never been designed from the view point of pedestrians. Hence, this study suggests a design support system that optimizes the environment factors so that the psychological stress is reduces and traffic conditions become smooth. This paper developed the virtual simulator that realizes flow of pedestrians as basic study. In this paper, firstly constructs of virtual environment of LPB and constructs of pedestrian model are described. Finally, the simulation results of the flow of pedestrian models are discussed.

Tumbling operation is useful to transport a large or heavy object. We develop twin arm robot for the purpose of tumbling operation on massive rock. In order to operate the object by the robot, it is necessary to determine the position of the action points. In this paper, a method of determining the action points is proposed. First, equations and constraints on tumbling force and moment are derived. Then, an objective function related to the force is set, and determination of the action points is solved as a nonlinear optimization problem. Then, we explain the composition of the constructed algorithm. Finally the usefulness of the algorithm is shown by a numerical example.

We develop a control device for teleoperation and a small robot for social infrastructure inspection. The control device is implemented force display function with force projection type bilateral control. For force projection type bilateral control, the robot is controlled by torque control method. The robot's arm cannot be controlled with torque control method in the small robot using position controlled servo motors. In this paper, a control device for operating the small robot which cannot be controlled by torque control method is implemented force display function with force projection type bilateral control. First, we explain a method which removes the drift of a force sensor without using a temperature sensor. A usefulness of the method is validated by experiment. Then, we suggest a method which implements force display function in the control device.

This paper proposes the improvement of the robotic gripper developed in our previous study. In the previous study, we developed the gripper with soft surfaces and underactuated joints. The soft surface was fabricated from a rubber bag filled with viscoelastic fluid and a microgripper inside the fluid. The underactuated joint was installed a ratchet. By utilizing contact with an environment such as a supporting surface, three grasping modes, parallel gripper, pinching and enveloping, were realized. The gripper was able to grasp various objects such as heavy, rigid, deformable, fragile and small object. However, the gripper cannot the objects, which is more than 30 mm in width. Therefore, the boundary length is increased by the mechanism everts the fingertip.

In current robot manipulation, the ability to manipulate rigid objects is sufficient, but it is still not enough to manipulate deformable objects such as rope and cloth. In this paper, we targets ”origami ”as an example of a soft object. In order to realize complicated folding, it is necessary for a robot to estimate a state of paper using a paper deformation model. In our previous model, creases were not considered adequately in the paper deformation model. Therefore, we aim at a recognition and motion control method to enable paper folding operation even when creases are added to paper during operation, and a method that deals with unexpected paper changes by reflecting creases in the 3D model is proposed. In the method, recognition of creases and addition of them to the 3D model is achieved. The proposed method is verified experimentally.

This paper presents a framework to extract the grasping points of objects being assembled. With this framework the objects are recognized and tracked using Augmented Reality (AR) markers. The data of the user's hand manipulating the objects are acquired with a motion capture device. The grasping points are determined with the motion capture data, the model of the objects, and point cloud based algorithms. The different modules of the framework are integrated with the Robot Operating System (ROS). The functionality of the framework is demonstrated with an experiment where the user manipulates an object while the grasping points are determined.

Our research group has been focusing on the automation of string tying-untying operations by a robot. In this paper, shape abstraction data of string and an operation planning method using it are proposed. In our planning method, string deformation is represented as state transition of known topology representation and proposed shape abstraction data. Proposed shape abstraction data enables a robot to decide shape operations that can correctly control string deformation toward target shape. In addition, an optimal operation for changing the string toward the target shape is calculated by scoring the difficulty of total shape operations. The results of experiments using proposed method are reported at final.

This paper describes the reinforcement of the grasping force in a robot hand. The robot hand with high degree-of-freedom and strong grasping force is necessary for performing various works. Robot hands have grasped an object by contacting each finger to the object. When it is desired to grasp strongly, human grasps an object by putting a finger on another finger. We designed the robot hand which grasps an object strongly by contacting a finger to another finger. The effectiveness is verified in the experiments of rotating the cap of a plastic bottle.

This paper describes a manipulation method for Flat-Flexible-Cable(FFC). To construct electric produces, A FFC should insert on hooks shaped on a manufacturing component. We focus on this manipulation and propose novel end-effectors and inserting motion sequence. A dual-armed robot was used to proof them and we confirmed that the cabling task was achieved with acceptable success rate.

Wheel loaders used in quarries usually repeat scooping and loading actions as face operations. So it is relatively easy to automate the operation and it also contributes reduction of cost and workers. The authors proposed the efficient trajectories for scooping piles with homogeneous rocks. On the other hand, there are often massive rocks in piles at the faces of quarries. Such a massive rock is usually picked up by inserting the loader bucket into the pile just below the rock. In our previous study, it was found that the consumption energy for picking up a massive rock becomes less by choosing the pushing operation. However, it is not clear why the reaction forces become constant and how the value is determined. In this research, the relationship between the physical phenomenon and the reaction forces are analyzed and the reaction force model is developed.

In recent years, the power assist suits have been proposed in several fields. In this work, we propose the novel assist suit for climbing the wall and moving along ceiling. With this support, the works at high places and rescue activities can be performed efficiently. In order to achieve these supports, we developed adsorption type assist suit with two functions. The first function is the adsorption mechanism on the wall and ceiling. For this purpose, we used Universal Vacuum Gripper, in short UVG, which is a vacuum suction cup that can be adsorbed even on uneven surfaces. The second function is a mechanism that assists the movement to lift the body when climbing and moving the ceiling surface. So, we use Makibben type artificial muscle and assist the wearer. In this paper, we report on development status and performance test of adsorption type assist suit.

Binding of reinforcing bars is a very important task in making reinforced concrete structures. However, since the work consists of crouching or bending over for long periods of time in harsh conditions, it is burdensome to the workers' bodies. In recent years, prolonging of the construction period is becoming a problem due to a decrease in the number of workers. Therefore, we propose an automatic system using a robot for reinforcing bar binding. In this research, we designed and developed a robot control system for an automatic reinforcing binding system. We also conducted experiments on experimental fields, and confirmed that the robot can successfully perform automatic binding work.

This paper describes automatic adjustment of operation input based on excavator motion prediction for tip-over prevention of a teleoperated excavator. This method predicts a future motion of excavator when an operation input is given. This allows the determination of the risk of tip-over by the operation input before an excavator moves. When the risk is detected, the operation input to the actuator is modified to keep the excavator from tip over. Future motion fo excavator is predicted by computing the movement of a teleoperated excavator based on an approximate excavator movement model. The proposed method successfully demonstrates tip-over prevention via simulation model experiments.

We are developing a robot for gas pipe inspection, which obtain trust forces by pneumatic actuators. The air pressure to drive them is provided by a compressor, which is set out of the pipe, via tubes. Fictions between these tubes and pipe walls sometimes constrain motions of the robot hardly. Therefore, we developed a tube-sending mechanism, which stays between the compressor and the robot, to support motions of the robot.

A Hydraulic excavator is commonly used in construction site. However, its user interfaces by operation levers is difficult to operate skillfully. In this study, we propose a virtual operating experience system of a hydraulic excavator using an RC excavator and augmented reality technique for skill training and evaluation of the operation. The proposed system realizes a virtual cockpit view by using an omni-directional camera and a head mounted display. It enables the similar operating feeling as that of a real excavator. We compare an expert and beginners by using the proposed system and discuss the difference.

In the previous research, a method to correct on grasping position/posture error on a vision sensor was proposed in grasping tasks with a robotic hand using a proximity sensor mounted on each of the fingertips. However, depending on color and shape of an object and surrounding environment, a position/posture suitable for grasping may not be achieved in some cases. Therefore, in this research, we develop a hierarchical information processing sensor system integrating a processing system that extracts only limited information from the proximity sensor output at high speed and a processing system capable of analyzing a lot of information although it takes time. It is expected that a feedback action itself and the subsequent grasping action can be appropriately made according to an object and surrounding environment from the information obtained in a process while performing the conventional feedback.

Collaborative robots that are able to coexist with workers are in demand, hence many companies have developed these robots recently. In our laboratory, we have developed assembly task partner robots beforehand. These robots need to avoid collisions with workers moving in the shared workspace. We have carried out collision avoidance motion planning using the RRT algorithm within the space where time axis is added to the C-space of robot. However, there was an issue that computation cost was unstable due to the random nature of RRT algorism. Therefore, we tried stabilizing the computation cost by choosing sampling points from probabilistically collision-free area that is modeled using GMM.

This paper presents the under actuated robot which have adaptive motion and hard stiffness by wire drive and jamming transition phenomenon. This robot hand have the fingers which can change the posture passively by wire drive. The fingers are connected to an actuator through elasticity elements. Therefore all fingers can change theirs posture passively along object shape by driving the actuator. After fingers motion, the density of granular body in the fingers increases by sucking air in the fingers. Jamming transition caused, and the fingers have hard stiffness. The robot hand can grasp wide range objects by this process.

This paper presents the recognition of object categories and object arrangements using You Only Look Once framework. Daily life objects are usually surrounded by other objects. The nearby objects around a target are obstacles to object picking. Therefore, different grasp strategies are needed depending on the object arrangements as well as the object categories. The recognition method of two arrangement categories of books (horizontal stacking and vertical stacking) is discussed in this paper. The recognition results show that the method is useful for recognizing object categories and object arrangements.

This paper presents the measurements of a hand posture in picking up an object from vertical or horizontal stack. Many objects which stored in a warehouse or home are place in a specific arrangement pattern. The object arrangement pattern affects the grasp strategy. We show some results of principal component analysis using the measured hand posture data in object picking.

In this study, passive dynamic control of a manipulator is designed and realized. The control strategy is based on the idea that the shift in position and posture of an end effector attributable to an external force is regarded as deformation of the robot. The standard linear solid (SLS) model is used to generate the deformable behavior, which describes the combination of plastic and elastic deformation. In addition, the relation between the SLS model and Maxwell/Voigt models are analyzed. Physical simulations with a robotic arm are executed to validate and analyze the proposed control law.

To increase the speed of transporting vehicle used in underground narrow roads, our research group is developing a differential four-wheel vehicle that allows contacts between a wall and roller bumpers. While wall contacting driving, large concavity and convexity of a wall have a large impact on the safety. Therefore, it's important to improve the detecting and avoiding performance. In a previous study, wall contacting driving method has been developed using one-tenth scale model, but it could not realize the avoidance of wall convexity and concavity in high-speed and sharp curved course. To realize the target capability, this study conducts an experiment using a modified scale model that has another distance sensor. The experimental results demonstrate the basic feasibility on the target capability.

A lot of disasters occur in japan. So, we have to restore the damaged area as soon as possible. However, we can't approach the area because of dangerous. Therefore, we use remote controlled construction machines to restore the damaged area. But now, the operator of these machines needs to be helped by a person who operates camera. So, I aim to develop automatic camera control system for camera operator. In this paper, we constructed a system to control the camera using the results of image processing. In addition, I compare multiple image processing methods to recognize the position of the construction machines in the screen. Also I experimented a new method combining multiple image processing methods.

This paper describes generating panoramic images with spherical projection using images acquired from cameras on a small mobile robot. We develop a small mobile robot on teleoperation for social infrastructure inspection. Multiple small cameras are equipped on this robot. Panoramic images around the robot which is generated using a multiple images acquired from the cameras are provided to an operator with a head-mounted display. First, we explain how to project images acquired from the cameras to spherical surfaces. Then, making an experiment which generating the panoramic images. Providing the panoramic images to the operator and revealing problems of generating the images.

A stereo camera method is used for the measurement of the image. This technique uses the principle of the triangle by positioning two cameras to measure the distance of the targeted object. In a previous research, position measurement of the camera`s image by using stereo camera method based on the Movement polyopia point method MMS(Multi Mobile Sight) has been developed. MMS can change the measurement range by enlarging the measurement distance and changing the distance and posture between cameras. However, in order to measure, it is necessary to measure the distance between the cameras and its posture, thus observation of the targeted object and the cameras are needed. Therefore, we aim to acquire position and orientation information for 3D measurement by mounting a projection marker on the camera.

Reinforcing bar binding is essential work for reinforced concrete construction. This work requires a lot of manpower and causes workers physical burden. However, due to the decrease in workers in recent years, the labor force is short and workers' burden is becoming increasingly high. As one solution to these various problems, we proposed and developed an automatic reinforcing bar binding robot. Specifically, we have developed various mechanisms necessary for the robot to perform binding work, and repeated experiments and improvements at the work site. As a result, we confirmed that the reinforcing bars can be binded automatically.

This article proposes a control system design that takes into account the control input dimension to be use in damping control of suspended load for crane systems. Many researches on transporting and anti-swaying control of cranes have been reported. However, most of those researches deals with overhead cranes. Only few studies deal with truck cranes since in the truck cranes, the suspended load should be controlled by using booms that have physical nonlinearity. Also, when introducing a new control system, the users want to use the same control input dimension as the previous controller. For this purpose, a crane model is extended with a hydraulic control valve model, and a controller is designed based on this extended model. This allows to generate control inputs in the same dimension as the existing valve controller. To show the validity of the proposed method, experiments using a truck crane has been conducted. The experimental results showed that the proposed controller has been successfully worked and the sway of the suspended load has been damped well.

This paper discusses how to implement an impedance control system using the joint velocity servo driver of a PA10 manipulator. Specifically, this paper addresses technical difficulties in implementing the impedance control system. First, the impedance control system, which was designed previously, is briefly described. Then, how to overcome the technical difficulties is discussed to develop new practical methods that are necessary for realizing the impedance control system.

Natural disasters and terrorism occur frequently around the world in recent years, and the damage site may be in a situation where it is difficult for people to enter. When people investigate, explore and rescue such environments, it is very dangerous to consider the occurrence of secondary disasters due to building collapse and fire. It is considered effective to use a robot for dangerous work such as a disaster. Many robots with flipper arms have been developed as rough terrain mobile robots, but few researches have been done on the shape of flipper arms. Our laboratory has developed a special shaped multi-function flipper arm and a rough terrain mobile robot ALACRAN-3.

Inside the building of the first nuclear power plant in Fukushima is contaminated with high levels of radioactivity.So, the remote control running device is needed to decontamination activity.Since decontamination activity uses the water, it requires a mechanism for moving the hose and cables.We developed” Inline Mecanum Wheel ”to solve this problem.But, we need ”Inline Mecanum Wheel unit 1” to insert from the tip of the hose, we assume take time to attach to the hose.In order to solve this problem, I developed” Inline Mecanum Wheel unit 2” which can be divided and attached at the center of the hose.

There are some situations not to be able to move on gravel ground by crawler vehicle. That is caused by some sane or stones to stick between sprocket tooth and chains. At this time, surveying experiments using actual crawler vehicle have been executed to find the reason to be stuck vehicle. Experimental results have shown some clues of stuck crawler.

There have been many vehicle type robots proposed for indoor applications. On the other hand, there are relatively small number of outdoor vehicles that are compact and affordable. We expect that more and more outdoor service robots will arise used in various industrial fields, such as agriculture, construction and so on. We propose a tracked vehicle for research on driving control and application verification in such fields.

As a part of our researches on several types of Giacometti robots, a 20-m-long, very light, and very simple robot arm called Giacometti arm with balloon body is discussed in this paper. Although Giacometti arm is not suitable for precise positioning, rapid motion, and high load capacity, which are the aspects most conventional robots focus on, it is designed for very specific purposes such as inspection using a small camera at its tip and is designed to be essentially safe even if it falls down or hits an object. It is realized using helium-filled balloon bodies and thin pneumatic muscles.
We created a 20-m-long robot arm. It weights 940g and has 20 degrees of freedom. We confirmed its usability at a mock-up facility.

Authors have developed an autonomous control system for wheel loaders to automate scooping and loading operation in quarries. In our previous research it is assumed that rock piles consist of homogeneous sized grains, and reaction forces from the piles during scooping are formulated and they are verified by fundamental experiments with our mechanical simulator. However, various sized rocks exist around the faces of quarries in practice. Therefore the reaction force model of such piles is necessary to automate the scooping in quarries. In this paper, we investigate the influence of the relationship between thickness of penetrating plate and sediment particle size, and tip shape of plate in penetration resistance.

Many robotics technologies, autonomous machines (incl. intelligent construction system) or remote control machines or others, are introduced into a construction machine today. Some studies on efficiency and improvement of these technologies has been performed in PWRI. In this paper, results on the studies are shown.

Inspections by mobile robots are required in nuclear power and steel plants. The robots need to climb and descend stairs because equipments are installed on difference levels. This paper proposes a turning motion on the stairs for tracked vehicles. A characteristic of the proposed turning motion is by using the reaction force from the safety wall. Experimental result shows that the movement direction can be changed by the proposed turning motion. In addition, the proposed turning motion enables the vehicle to run with less slippage than other turning motions.

When an eruptive activity began in active volcanoes, it is difficult to enter the volcano area. Therefore, expectations for robotic exploration are increasing to get information for observing the status of volcanos. Tracked vehicles have high terrainability, so it is suitable for exploring volcanic environments. However, when such a vehicle travels on a loose weak slope, a large side slip occurs, particularly in rotating motion. Odometry becomes inaccurate owing to this slip, and it fails to follow a specified route. In this research, we propose odometry and path following control method to compensate side slip. We verified the effectiveness of the control and the improvement of the position estimation accuracy in indoor sandy slope experiments.

In recent year, the elderly and handicapped person who using electric wheelchair and electric cart are increasing. These mobile equipment has ability of energy efficiency mechanism. However, this mechanism is difficult to move in uneven terrain. When electric wheelchair climbs over a step, driver may be dangerous by exerting force between wheel and a step. So we should think about safety in climbing over a step. In this report, we propose a model of electric wheelchair in approaching a step. We use the model in order to prove safe driving by simulation. Therefore, we focus on triangle which established by the contact point of wheel and a step, actuator of rear wheel and sitting position of driver. And we consider the danger to the driver in the case thistriangle rotates.

Japan has problems with a declining birthrate and aging population and demand for life support robots has increased. We have been conducting research and development on a robot base and we have managed to develop robot base capable of carrying a 100kg load and overcoming steps, and slopes. And we have demonstrated the safety enhancement of the robot base through experiments to reduce the impact of collisions using a bumper sensor and control the deceleration using laser range finders. In this study, we will develop six-wheel type robot base “Taurus” and present the some experimental results.

Recent investigations have been demonstrated autonomous mobile robots that move on a various real environment. In our laboratory has demonstrated autonomous mobile robots in universities and parks in Tsukuba city so far. However, we have never demonstrated in the metropolis environment. Here we study aim to be able to move the autonomous mobile robot while much pedestrian traffic in the metropolis environment such as Shinjuku station. Therefore, we considered characteristics of metropolis environment, expected questions, and those solutions. We activated the robot equipped with those key technologies around Shinjuku station. As a result, the robot couldn't reach the destination, but we got the robot which moves in the metropolis environment should solve the problems as knowledge.