Recently, with the development of life sciences, measurement of cell characteristics has been carried out to evaluate life phenomenon. The cell stiffness is one example of these characteristics and has much attentions. It is known that cell stiffness changes depending on the status of the cell. Thus, stiffness measurement is one of the important parameters to evaluate a cell. However, since measuring cell stiffness requires nano order manipulation and sensing, cell stiffness has not been extensively researched. Furthermore, it requires accurate operation skills and is difficult for humans. In this paper, we propose an automated stiffness measurement system. This system is combining force data and image data, and a multi-scale simultaneous observation. multi-scale simultaneous observation is realized with fusion of two different enlargement magnifications microscopes. By using this system, an automated and high-precision cell stiffness measurement is achieved.

In this study, we developed a polysaccharide hydrogel for 3D bioprinting. 3D bioprinting is printing hydrogel containing cells or biocompatible materials by using 3D printer and it can be applied to tissue regeneration. It is important to develop a new hydrogel for bioprinting. Polysaccharide based hydrogel has high biocompatibility and it may have a great potential for tissue regeneration in bioprinting. Chitosan, that is one kind of polysaccharide, was reported as wound healing and cellular adhesive material. We chemically modified chitosan and achieved visible light gelation of chitosan. Visible light is not harmful gelation way for cells compared to ultraviolet gelation. Indeed, we showed that this bioink has good printability and thus it may have a great potential for bioprinting.

Recent research has revealed that cancer cells far from blood vessels present in hypotrophic in tumor and they deeply involve malignant transformation such as metastasis and recurrence. In this study, we develop a device to make it possible to observe such cells present in the hypotrophic region in vitro. The device supplies nutrients from the inside of the cell aggregate using alginate fibers and can form regions with different nutritional states inside the three-dimensional cell aggregate.

In this paper, a tendon structures made by 3D printer were added to cultured muscle made from mouse skeletal muscle. This tendon structure was made from PDMS, but can be exchanged to various materials. In fact, the tendon structure was initially made from PDMS, but was later changed to a light-cured resin. By connecting the modular bioactuators made using the tendon structure in parallel, the contract force of the cultured muscle was improved.

This paper proposes a 3D assembly method of aligned fiber-structure such as a human skeletal muscle tissue. The innovative point is assembly method of fiber-structure consists of artificial modular micro-fibers. The magnetite particles are included inside of both ends of micro-fibers for “Assembly”, “Positioning” and “Fixation” of micro-fibers, while the capillary force is also used to bundle several fibers tightly. Several different kinds of fibers are fabricated and assembled to form the tissue like structure. In this paper, the muscle fiber and sensor fiber, which has spherical microsensors inside, are used for demonstration of the proof of our concept. After the assembly, the micro-fibers were successfully fused to form a tissue like structure with highly arranged cells, where the microsensors could be used to monitor the inner state of the tissue

This work describes a reference fluorescence microsensor for pH measurement in a cell culture environment without initial pH information. The reference pH microsensor is constituted of a fluorescence pH microsensor made of hydrogel, and polydimethylsiloxane as a barrier layer for hydrogen ion. Reference pH microsensors do not be affected by surrounding pH since PDMS does not pass hydrogen ions. Therefore, pH in the cell culture environment is measured by comparing the fluorescence intensities between fluorescence pH microsensor and reference pH microsensor. In this paper, the fabrication of reference pH microsensor and evaluation of pH dependency of reference pH microsensor were demonstrated.

For the donor scarcity problem, there are many Tissue Engineering studies. We fabricated toroidal cell structures inside a microflPA000354941-310uidic device for tissue engineering applications. A photo-crosslinkable biodegradable hydrogel gelatin methacrylate (GelMA) was employed to encapsulate biological cells for assembling cell structures. stabilizeing the unstable GelMA manufacturing method,we review the manufacturing method and searched for a method that can be stavly produced. So changing the diakysis method, It was possible to prepare GelMA which is suitable as a cell culture medium. RLC-18 cells grow up on the micro structure. When RLC-18 cells grow up, GelMA is gradually reduced.

For practical use and commercialization of nursing care robots, it is important to achieve “Practical use”, “Safety for elderly people” and “Reasonable Cost”. In this presentation, we introduce examples of our efforts to realize them.

Japan is a country with many earthquakes. So there is a risk that the petroleum complex will fire. The tank fires in petrochemical complexes are special disasters that are large and difficult to extinguish. The major issues are that the radiant heat from the flame is large and there is a risk of explosion, so that people cannot easily approach the disaster site. The Fire and Disaster Management Agency has submitted a public invitation for "R & D on a firefighting robot system for disaster response" (a five-year project starting in fiscal 2014), and has completed the world's highest-level firefighting robot system that can operate near the fire site. The challenge was to make the robot resistant to radiant heat and handle the water supply hose without human assistance. This document describes the outline of the project, and the outlines of the water cannon robot and the hose extension robot development that I developed in this project.

Autonomous driving technology is attracting social attention. However, there are still a lot of challenges for practical use in building. We have developed the automatic operation platform for building. The automatic operation platform communicates with mobile robots strolling in a building or a facility to accomplish their tasks. Especially, our platform treats multiple robots from different manufacturers. This paper shows the structure of the platform and an experiment in the actual building. In the experiment we demonstrate that our platform navigates visitors through a service robot.

The decline of the working population is a social problem in Japan, but the security industry is particularly prominent. To increase the efficiency of the security work, we developed an autonomous patrol robot ”SQ-2“. We are still working to improve the feature of the robot. In order to deploy the developed software to robots being used by customers, it was necessary to improve the quality of the software. We constructed a productive development environment to deploy those software and operate our robot continuously. The environment has made it possible to develop and operate a robust robot system continuously.

To automate pick and place task by robots in warehouses is processed because of labor shortage. In order to pick and place items, the robots need to recognize the items. Some researches propose deep learning method for object recognition. They are trained by using simulated data. However, they require many 3D models of the items. Making 3D models in warehouses spends time and needs special equipment. Thus, to reduce the making cost is important for utilizing the deep learning method in real warehouses. In this study, we developed a method that re-uses primitive 3D model for building a 3D model of a new item. The method estimates a primitive class of the new item. Then the method builds 3D model corresponding the primitive class. We confirmed that the proposed method can make a 3D model of a box item and a cylinder item respectively.

In recent years, there has been a demand for the construction of a robot system that takes safety into consideration. We considered versatility in hardware and software and developed a wireless emergency stop device that conforms to functional safety standards. In this report, we introduce the key technologies for the device. Then, we show the results of evaluating the performance when incorporated into several systems.

Recently, every retail store has been trying to manage inventory automatically. POS (Point of sale) cash registers implemented at many stores enabled easy management of inventory at each store by automatically managing the exchange of money with customers and sales. However, in current convenience stores, employees do all the replenishing of beverages on display shelves in showcases manually. If employees can replenish their inventory only when needed, it will lead to reduction of labor cost. The purpose of this study is to construct a system that monitors the beverages on the display shelf of a convenience store using a camera and automatically measures the number of beverages. The measurement target is a PET bottle beverage, and here we experiment by using image processing such as feature point detection. This paper reports on the method and evaluation results.

Patients with respiratory diseases need to improve their thorax mobility. Thoracic excursion is one of the basic evaluation indexes for evaluating respiration function. Chest expansion training is often conducted to increase the thoracic excursion. However, it is difficult for the patients to keep the motivation for this rehabilitation. In this research, we developed a simple thoracic excursion measurement system with real-time visual bio-feedback in the chest expansion test. In addition, this system was produced by Takei Scientific Instruments Co., Ltd from 2013. In this paper, we illustrated a part of the experimental results in the development phase and introduced commercialized product. In the experiment by 33 healthy subjects, we suggested that the system was effective for measuring thoracic excursion from comparing the measurement result by tape. And we also continued to measure by physical therapist and to analyze the effect of biofeedback in the chest expansion test.

In this research, safety-related control system for robotic mobility, which is used to support travelling in a coexisting environment such as an airport, has been developed to prevent collision by using signal information of laser range sensors. The developed safety-related control system has safety performance compatible with the international standard on functional safety, IEC62061 SIL2, and is used in many demonstrations and actual operations.

In order to speed up the user evaluation of robot applications, this paper proposes a new method called “Ningen prototype”, with which humans act as a prototype robot and the user evaluation is conducted using it. The effectiveness and issues of this method are discussed based on a preliminary experiment.

A motion generation method using deep learning has been proposed that is robust against various environmental changes. In this paper, we describe a reflection motion method that responds immediately to environmental changes. The reflection motion can be developed using a simple feed-forward neural network. The tasks of verifying the reflex movement were ”grip force control” that grips various objects without crushing them, and ”obstacle avoidance” that avoids unknown objects during autonomous movement. Results of an experiment using a real robot confirmed a reflection motion could be generated.

One of the problems in a super-aging society is to secure transportation for the elderly. To solve the issues, the authors focused on electric mobility scooters that is one of the personal mobility. The purpose of this research is to develop the autonomous driving system of electric mobility scooters for realizing safe transportation and new mobility services. This paper describes the introduction of traversable area judge methods in the walking environment the authors proposed. And, demonstration experiment field that simulated the walking environment prepared is also described.

This paper gives an experimental study on nonlinear guaranteed cost control approach to automatic flight and landing of a small flying-wing unmanned aerial vehicle (UAV). The nonlinear guaranteed cost control is achieved by fuzzy model-based control for a longitudinal nonlinear dynamic model considering the complicated aerodynamics. A set of guaranteed cost controller design conditions represented in terms of linear matrix inequalities is numerically solved to obtain the feasible optimal feedback gains by minimizing the upper bound of a given cost function. The experimental results in both the automatic flight and landing demonstrate the utility of the nonlinear guaranteed cost control approach to the small flying-wing UAV.

The goal of this research is to design a robust stabilized control of a drone in severe environments, especially in strong wind areas or package delivery. We adopted the active disturbance rejection control (ADRC) in this project due to its ability to estimate the disturbance without knowing an accurate model of disturbance. Considering the limitation of drone hardware, instead of third-order nonlinear extended state observer in ADRC, we used a second-order linear extended state observer to estimate disturbance. The performance of the ADRC based control law is experimentally validated on a real quadcopter. According to the experiment data, ADRC shows significant improvements in stabilized control of a drone compared to the conventional PID control.

UAV is used in various fields, and it is possible to reduce physical burden by using it for transportation. Existing UAVs are limited since there’s a limit to the weight they can carry. Therefore, we propose a transportation support system using multiple coaxial-rotor UAVs. In this paper, the attitude of coaxial-rotor UAV is controlled by STA architecture. It allows considering the state of the environment when controlling multiple coaxial-rotor UAVs. For that purpose, we construct a NN considering past time series using LSTM. By performing NN learning and STA in parallel, the posture angle is predicted and changed to match the target angle. For the experiment, the test bench is constrained to 1 DoF and we compare the performance when the STA architecture uses a conventional NN or LSTM. As a result, it was suggested that adapting the STA architecture using LSTM tends to be better than conventional NN.

In recent year, rapid development of UAV’s flight technology and its practicality are attracting attention, and the UAVs are applied to various business. However, risk of accidents due to crashes becomes apparent. For this problem, we aim to develop a UAV which has a paraglider and makes soft landing to safety area after gliding. In this paper, a procedure to land softly to safety area using the paraglider and a method to acquire various information necessary for the procedure through crash experiment using a mock-up are proposed.

In recent years, hornets appear in urban areas and it is regarded as a problem since the hornets attack human. It is also dangerous to destroy hornet nests since hornets have poison and they attack the people who approach to the nest. In this research, finding hornet`s nests by tracking hornets by combing computer vision and aerial drone is proposed.

A quad tilt rotor UAV is known as one of advanced systems of the conventional multirotor UAV. The quad tilt rotor UAV enables independent control of position and attitude, and flight in a 90-degree attitude state. However, the quad tilt rotor UAV has the drawbacks of thrust loss and difficult flight at certain attitude. As a solution to these issues, this paper proposes a quadrotor UAV with dual-axis tilting mechanisms which rotate thrust direction in two axes. This paper describes the developed airframe and its control law, and verifies the performance of the developed UAV by numerical simulation and flight experiments.

Recently, the obstacle avoidance and localization of UAV in places without GPS signal has received widespread attention. This paper presents a visual feedback control of UAV with two monocular cameras. In the system, visual odometry and visual servoing are switched in realtime.

In this paper, we focus on an aerial surveillance system composed of aircrafts such as sentinels and searchers. The mission of the sentinels and searchers is to patrol unidentified flying targets. For this purpose, we use single sentinel and multiple searchers. These aircrafts have different roles as follows: the sentinel is able to monitor broader area and detect incoming targets, and the searchers approach and identify the target detected by the sentinel. In our previous surveillance system, the sentinel was assumed to be fixed. In this system, the sentinel is also allowed to move according to the target. Therefore, in this paper, we optimize the cooperative surveillance behavior of the moving sentinel and searchers by designing reward functions of a value iteration method. Through simulation experiments, we show that the cooperative surveillance enables the searchers to identify incoming targets as many as possible.

A prototype single rotor UAV with vertical fins under the main rotor was produced. The vertical wings have stator wings and movable fins. This UAV performs anti-torque correction on stator wings and controls attitude with movable fins. Exercise analysis considering the bias of momentum by the main rotor was performed. Based on the results of the motion analysis, a simulator based on detailed equations of motion was prototyped. A prototype of the flight control method was created in consideration of the momentum bias, and the effectiveness of the control method was confirmed by simulation. In addition, a prototype of the software was created and confirmed using the prototype.

In this study, we propose and develop a drone with a tilting frame whose two cameras are fixed to achieve a surveillance under non-GPS environment. It can tilt its frame in the pitch direction by actuating only one servo motor using the parallel link mechanism. In a flight experiment, we confirm that the drone can fly stably by using two cameras under non-GPS environment.

Robotic manipulators that use thrusters for weight compensation are an active research topic because they can exceed the limits of elongation. However, existing manipulators that use thrusters have limited elongation because the hardware design is not sophisticated enough. This paper focuses on overcoming these limitations and realizing longer articulated manipulators. As a means of realizing such a manipulator, we adopted the mechanism of Hiryu-II, which was previously developed by our research group. In order to cancel the moment of each link, the thruster position was derived in consideration of the balance between the static load of the manipulator and the thrust force. Consequently, we demonstrated the feasibility of the proposed manipulator with a prototype of length 6.6 m.

Low-speed-descent objects such as an object descending with a parachute are widely used in the aerospace field. They are used for ensuring safety or making the success of their missions but have some disadvantages, especially in safety. The authors’ previous studies discussed a solution by mid-air retrieval of low-speed-descent objects by using fixed-wing UAVs to overcome the disadvantages. It adopts the updating final state control in the process from before to the retrieval point, and the frequency-shaped LQR control in the process to horizontal steady flight after retrieval. The proposed method verifies its effectiveness by numerical simulations. This paper describes the construction of its experimental system to verify its effectiveness experimentally. The aircraft which has advantage for the experiment is selected, the sensing and control systems that meet the requirements are constructed, and their usefulness is verified by a preliminary experiment.

To realize an aerial manipulation by using a multicopter with a robot arm, a control of its center of gravity is required. For safe flight, it is necessary to obtain its parameters utilized for multicopter control before take-off. However, the accuracy of parameter estimation depends on the posture of the robot arm used for calibration, and its appropriate posture has not been discussed yet. Therefore, in this research, we propose a method to calculate the optimal posture using a simulation focusing on measurement errors. Experiment result show that the multicopter can control the arm without losing its posture by using estimating the parameters using the proposed optimal posture.

In recent years, research for automating the inspection of concrete structures using multi-copters has been actively conducted. In particular, the development of tethered multi-copters has attracted widespread attention. However, since the feeding cable is connected to the power supply, the tension of the cable affects the multi-copter as a disturbance. Therefore, it became difficult to approach the inspection location for tethered multi-copters. In this paper, we apply the catenary theory, which has been widely used for adjusting the tension or length of a multi-copter feeding cable in a two-dimensional space, to a tethered collision-tolerant multi-copter in a three-dimensional space. In addition, we report a method of minimizing the disturbance of the feeding cable by adjusting the length of the feeding cable as needed in a three-dimensional space.

We have developed a reinforcement learning approach for controlling a flapping robot toward autonomous flight control. The flapping robot is equipped with a center-of-gravity (COG) shift mechanism, thereby enabling seamless changes in the flight attitude between hovering and level flights. The yaw angle can be controlled by differentiating flapping frequencies of left and right wings. We demonstrates that reinforcement learning is effecttively utilized to stabilize the hovering flight.

We have developed a flapping robot inspired a butterfly that has law flapping frequency and a few degrees of freedom of the wings. In this paper, we investigate relation between CoG position of a manufactured robot and its attitude control to glide. As a result, the CoG of robot was located 2mm ahead of geometric center of wings to glide. Furthermore, the more forward the CoG was located, the more stable a robot glided. On the other hand, the more backward the CoG was located, the earlier a robot began gliding.

The method attaching a color 2D code to an object which it is difficult to find features even by using color and point cloud data with CSHOT descriptor is proposed. If the object composed of simple planes, there are no remarkable characteristic on the SHOT descriptor. Furthermore, there are products that have different characteristic color information even though they are the same product like food packages. The proposed method is powerful tool to find such kind of objects.

In order to enable autonomous mobile robots to move in human-crowded environments, active inducement including using voice and light contact can be one of effective solutions. However, to ensure safety and effectiveness, the robots must predict how the crowds will behave according to the robot approach as well as among humans forming crowds. In this study, we thus develop a navigation system that predicts multiple human movement based on inducible social force model (i-SFM), which enables the robots to select proximal paths with active inducement including human contacts, and determines the path by considering the movement efficiency of both robot and humans. Experimental results indicated that the proposed method could predict the movement of the human crowds and improve movement efficiency, compared with conventional method.

Conventional remote operation of an excavator has low work efficiency compared with direct operation. This is because it is difficult for the operator to recognize the state of the excavator and its surrounding environment. To overcome this problem, we have developed a semi-autonomous control system which consists autonomy (dynamical system with attractor) and human action (admittance control). In this paper, we propose a method of deforming the trajectory of the attractor to change the excavation start position. In the proposed method, the deformation of the trajectory is realized so as to change only the excavation start position by coordinates transformation, and the parameters in the coordinates transformation matrix are changed depending on operational force. In addition, we present a newly designed attractor for a 4-DOF excavation trajectory with bifurcation. We implement the proposed method and the attractor on the prototype of an excavation robot and verify the operation.

This article proposes a variable impedance control to improve the stability by changing impedance parameters according to the strength of griping force of the operator. In the proposed method, based on a stability analysis of impedance control, appropriate impedance parameters are derived so as to satisfy both the light maneuverability and robust stability. In the proposed variable impedance control, the impedance parameters are adjusted to be relatively heavier as the gripping force is becoming stronger to keep the system stable. The validity of the proposed system has been confirmed by experiment using a one degree-of-freedom linear stage driven by a ball screw with an AC motor, having kept the stability regardless of the amount of the applied gripping force.

This article proposes a human-adaptive impedance control to achieve human-robot co-manipulation, in which a recurrent neural network (RNN) estimates a human state that is used in improving the contact stability of impedance control. The human sate here is defined as what indicates that the human-arm stiffness becomes harder and the stability of the impedance control is deteriorating. In the proposed method, the human state is estimated from rectified-and-integrated electromyogram (iEMG) signals while the person is manipulating an object with a robot. According to the degree of the estimated human state, the proposed method changes the impedance parameters to be heavier online so as to make the system more stable and then returns the mechanical parameters to be lighter once the stability is restored. The validity of the proposed method is verified by experiments using a commercial-of-the-shelf manipulator. The experimental results showed that the proposed human-adaptive impedance control, which adjusts the impedance parameters online according to the human state, is effective to prevent the human-robot cooperative system from coming unstable.

This study aims at realization of a multi-DOF load compensation manipulator that is capable to adjust the compensation force at multiple heights. Articulated manipulators are used for picking works and manipulators that can corporate and reduce energy consumption are required. By using the load compensation manipulator, the gravity applied to the manipulator and the object is completely compensated. Therefore, energy consumption can be greatly reduced and the operator can easily move the weightless manipulator. The load compensation manipulator has 3 required performances: the ability to adjust the compensating force, the ability to adjust the compensating force at multiple heights, and the ability to have multi-DOF. In this study, the compensation force can be adjusted at any height by using a constant torque spring and a transmission, and it is possible to easily increase the number of DOF by adding a differential gear to them.

In this paper, we develop an interface for personal mobility robots. The robots are able to carry passengers. The passengers/users are allowed to control the robot through the interface. For this purpose, we focus on brain signals of the users. The electroencephalography, EEG, is used as the input for the brain-mobility-interface, BMI. As the output from the BMI, control commands are sent to the robot. In order for the BMI to classify the input EEG into the commands, we apply a deep learning technique. Two types of networks, such as deep neural network and convolutional neural network, are used as the classifiers. Through the experiment with a personal mobility robot, we show that the BMI enables a passenger to control the robot toward a destination.

This study aims at multi-tasking support using a wearable robot arm “third arm” for daily chores execution. In everyday life, when performing actions such as opening an umbrella or stirring a pot, it becomes necessary to define a target space or point.Additionally, with wearable robotics, the operator must be able to communicate such localized instruction to the robotic limb independently of his/her own actions and without interrupting them. This research proposes an interface for polar coordinate spatial localization indication based on combining head direction and head back-and-forth movement, "Head Bobbing", to establish and point to the intended target. Using VR space, this method was evaluated by considering ease of use during simultaneous task execution, and impact on performance accuracy of the operator on his independent task.

This paper proposes a new speed control method for the electric smart wheelchair to avoid collision with other pedestrians safely and smoothly focusing on the awareness of the pedestrian. One of the main issues in robotic wheelchair research is to avoid pedestrians for safety. In order to realize smooth collision avoidance, how the wheelchair and surrounding pedestrians interact with each other must be considered. The pedestrians sometimes behave differently depending on their awareness of the other agents. Awareness, a pedestrian has recognized the ego wheelchair or not, is introduced in the multi-agent-based simulation to simulate the pedestrian motion. On the other hand, the prediction model of the pedestrian is embedded in Model Predictive Control and switched depending on the awareness. Finally, the usefulness of the proposed method is shown in simulations.

Nowadays, human-robot-collaboration (HRC) gains high interests among researchers. And considered to be introduced in the field of manufacturing. In such a system, new risks which have not been considered may will be estimated. One of them is that the robot and the worker may collide with each other. And to avoid collisions, definition of separation distance between the robot and the worker is needed. According to ISO10218-1, it is said that the safety of the worker needs to be maintained even in the case of a single fault of the safety system. So, when the separation distance is defined, a single fault should be considered. Let’s consider a single fault of robot. As the method to quantify running away, potential runaway motion in task space (PRAM-t) was proposed. In PRAM-t, dynamic manipulability ellipsoid is used to calculate acceleration when robot runs away. However, when we consider running away originating to single fault, it could be exaggeration. So, I propose new method to compute acceleration during the running away of robot using an algorithm of computation of minimum volume enclosing ellipsoids. And I compare my method and DME by case study.

When human interacts with robots or machines, many skills such as cognition, planning, and control are involved. However, it is difficult to perform in every single one of these fields. Therefore, assist regarding human-machine operation has been actively researched. For instance, a method used to feedback surrounding information using a camera is used in automobiles. However, it has been reported that this method reduces human concentration. In order to solve this problem, assist methods using tactile feedback and force feedback has been proposed and demonstrated their usefulness as a way for humans to obtain feedback more naturally. Among them, we focused on subliminal calibration. Subliminal calibration is an algorithm that focuses on the process of acquiring skills and determines the value of the calibration so that the process doesn’t disturb the operator. The purpose of this research to enhance the operator’s skills in order to help him operate machines at will.

In human-robot interaction, the passivity is important. With high dynamic range (HDR) six-axis force/torque sensors, which can measure about 0.1N force while maintaining the same rated values as comon force sensors for industrial purposes, impedance parameters of admittance controllers can be set to smaller values to improve the passivity. When a large external force is applied, however, speed command of admittance control abruptly increases because of small impedance parameters. In this research, this problem solves by applying an arctangent function to the speed command; by the saturation with the arctangent function, manipulators can achieve both enough passivity and safety. We comfirmed the effectiveness of the proposed methods in experiments.

This paper investigates various collision detection methods to implement safe collaborative work between humans and robots. The first method is to detect the collision from the change of the commanded torque at the time of collision by using the data of the filtered commanded torque. The second method is to detect a collision by comparing the commanded torque with the reference torque when the robot has passed a predetermined trajectory. The third method is a collision detection method using estimation. We conducted a collision experiment using a sponge and an experiment in which changes the target position on the way. The results show that the collision detection method using the estimation is more versatile than other methods.

This paper proposes a method to estimate external forces and moments without a force sensor when the robot is in the wrist singular configuration. External forces and moments are estimated with Jacobian matrix and joint torques by the equation of statics. When the robot is in a singular configuration, however, Jacobian matrix becomes rank deficient and it becomes difficult to obtain estimated values. In the proposed method, external forces and moments are estimated by using the singular value decomposition. Experiments show that the proposed method is valid for estimation of external forces and moments in the wrist singular configuration.

This paper describes a novel estimation method of human position for the collision avoidance system in practical use. This method uses RSSI-based position fingerprint to estimate human position at low cost and latency even if human is around obstructions such as walls. We conducted an experiment to verify the error between estimated and measured values of distance and direction. Its result shows about 6m maximum error for distance and about 5 degrees maximum error for direction. From this experiment, our proposed method is expected to make humans aware of the direction of others approaching to avoid the collision. As future work, the accuracy of human position estimation should be improved especially the distance error which is occurred by the reflection of a radio wave.