This paper focuses on soft actuators that utilize fluid power to drive soft robots and describes their features and applications. First, it discusses how soft actuators function as elemental technology in robots. This is followed by an introduction to the driving principle and features of fluid-driven soft actuators. It also classifies these soft actuators based on the fluid power source and the active mode of operation. Furthermore, an overview is provided on the materials employed in soft actuators and the control and evaluation methods for them. Finally, currently reported applications of these soft actuators, such as wearable devices, grippers, and bio-inspired robots, are presented.

In recent years, the labor shortage caused by the decline in the working-age population has become increasingly severe across various industries. The construction industry is also facing an urgent situation. Improving productivity is critical in response to this situation. In addition, domestic infrastructure is aging, and climate-change-related incidents such as heavy rainfall and landslides are becoming more common on a global scale. To address these challenges, digital transformation (DX) deployment must be accelerated, and robotic technologies must be efficiently utilized.

Although various construction robot technologies have been developed to date, the robotization of construction and inspection work cannot be achieved using standalone technologies, regardless of their technological superiority. Today’s construction robots must function not only as standalone tools but also as part of a system integrated into construction and inspection methods, or even as a robotic system that operates throughout the construction site.

From this perspective, we published a special issue titled “Advanced Robotic Technology and System for DX in Construction Industry.” It includes papers and development reports on robotics, mechatronics, and information system technologies relevant to the construction industry.

We hope that this special issue will arouse interest in construction robots among researchers and engineers and accelerate the development of construction robots, related technologies, and DX in the construction industry.

Finally, we express our sincere gratitude to the editorial board of the Journal of Robotics and Mechatronics, the editorial team at Fuji Technology Press Ltd., and all of the reviewers.

The objective of this study is to achieve high efficiency in autonomous hydraulic excavators by imitating the bucket trajectory operated by an expert. For this purpose, bucket trajectories of experts were collected, and a trajectory was planned using machine learning of a model that relates measured soil shapes to the bucket trajectories of the experts. In this study, we proposed a hierarchical model consisting of a model for estimating movement and a trajectory, with a focus on the fact that different trajectories are generated for the same soil shape as a result of the analysis of the skilled persons’ movements. The trajectory output from the model was replanned to have a smooth trajectory using numerical optimization. For the backfilling task, the error from the target shape and the amount of soil transported per movement were compared with those of an expert. The proposed method increased the error from the target shape by approximately 66%, while the amount of soil transported was approximately 58% of that of the experts.

Wheel loaders are construction machines that are mainly used for excavating and loading sedimented ground into dump trucks. The objects to be excavated range from large materials, such as blast rock and crushed stone, to small materials, such as gravel, slag, and coal ash. Therefore, the excavation operation of wheel loaders requires a high skill level to cope with various terrains and soil types. As worker numbers at quarry sites decline, developing highly automated technology to replace operators is crucial. In particular, the geometry of the ground to be excavated by the wheel loader changes with each excavation, so the control parameters must be adapted sequentially during automated excavation. In this study, we proposed an online learning method using Bayesian optimization to search for control parameters from multiple trials and modify them sequentially. In particular, we formulate a multi-objective optimization problem maximizing a weighted linear combination of the payload and workload as an objective function. To validate the proposed method, we constructed an environment in which repeated digging tests can be performed using a robot manipulator with a bucket attached. Through comparative tests between feed-forward control, in which the robot moves along a fixed trajectory independent of the digging reaction force, and off-line control, in which the robot modifies the digging trajectory in response to the digging reaction force, we compared the ability of these methods to cope with terrain volume that is different from that of the optimization trial.

Automatic operation is one of the main research topics in the field of construction for solving labor shortages. However, workers still perform manual work at construction sites and there is an urgent need to automate this. To automate manual work at construction sites, construction machinery requires force control that absorbs external shock impact forces and provides appropriate forces along with environmental forces. In this study, the boom/arm/bucket joints of a 0.06 m³ class excavator were electrically driven, and a series elastic actuator (SEA) was applied to satisfy these two requirements. There are few examples of SEA studies on its application in large machines with high outputs, such as excavators. We designed the structure of the SEA, conducted the control design, and fabricated an actual bench simulating a 0.06 m³ class excavator to verify its performance. The results of the bench tests show that our SEA system achieves a control accuracy and responsiveness suitable for use in manual work.

The digitization of the construction site environment has progressed rapidly. In this study, the operations of hydraulic excavators—which are machines widely used in the construction industry—were advanced to enable automation and unmanned operation. To achieve this, it is necessary to determine the environment of the machines at the sites, and a real-time measurement and visualization methodology that can be installed at common construction sites is required. In this study, we propose a measurement system for reconstructing a wide range of surrounding environments using machine-borne sensors mounted on a hydraulic excavator. The proposed system measures the entire surrounding environment using a sensor unit composed of a laser imaging detection and ranging (LiDAR) and a wide-angle camera. Furthermore, methods of time-series integration for wide-range and dynamic measurements during work for occlusion-robust visualization are proposed. In an experiment using actual machines on an earth-moving site, we validated the performance of our proposed system by quantitative evaluation and confirmed that the system provides an effective solution for the digitization of construction sites.

In the future, a situation in which operators switch between different machine classes in teleoperated hydraulic excavator is envisioned. In such a case, because the classes have different dynamic characteristics, the operator is expected to acquire an internal model for each class and switch models each time the operator switches between machines. However, in the case of teleoperated hydraulic excavator, the operator cannot obtain information such as the size and dynamic characteristics of the machine to be switched to; thus, the operator may not be able to switch internal models properly, which may affect the operation efficiency. Therefore, this study proposes a method in which images and videos are used to present the dynamic characteristics of the next machine to be operated during machine changeover in teleoperated hydraulic excavator. To verify the effectiveness of the proposed method, a simulator that imitates teleoperated hydraulic excavator was built and tested on test subjects. The swing operation time significantly increased when the machine was switched without presentation, compared with the case without switching. Meanwhile, the proposed method did not increase the swing operation time associated with machine switching, suggesting its effectiveness. The video presentation method was more effective than the image presentation method for suppressing the increase in swing operation time, indicating that the operator can immediately switch to an appropriate internal model with the presentation of the dynamic characteristics of the machine in advance using video.

Research and development efforts have been undertaken to develop a method for accurately localizing construction vehicles in various environments using multiple 3D-LiDARs installed in the work environment. In this approach, it is important to calibrate the installed positions and orientations of the multiple LiDARs as accurately as possible to achieve high-accuracy localization. Currently, calibration is performed manually, which results in accuracy variance depending on the operator. Furthermore, manual calibration becomes more time consuming as the number of installed LiDARs increases. Conventional automatic calibration methods require the use of dedicated land markers because stable features are difficult to acquire in civil engineering sites in which the environment is altered by work. This paper proposes an automatic calibration method that calibrates the positions and orientations of 3D-LiDARs installed in the field using multiple construction vehicles on the construction site as land markers. To validate the proposed method, we conducted calibration experiments on a group of 3D-LiDARs installed on uneven ground using actual construction vehicles, and verified the calibration accuracy using a newly proposed accuracy evaluation formula. The results showed that the proposed method can perform sufficiently accurate calibration without the use of dedicated land markers in civil engineering sites, which increase costs and make features difficult to acquire.

The exploration and utilization of water resources on the Moon are of substantial global interest. To utilize lunar resources and construct bases, the construction machinery should travel over the lunar surface (which is mainly covered with powdery regolith) and excavate the regolith. However, various technical issues should be resolved to achieve this efficiently. In this study, a new platform rover was developed, and its motion behavior was analyzed to better understand the traveling and excavation behaviors of construction machinery on the Moon. The rover is a four-track vehicle equipped with a robotic arm consisting of a boom, arm, and bucket. To analyze the rover’s motion behavior in sandy terrain, we first developed a simulator based on terramechanics and performed a numerical analysis. Subsequently, various experiments were conducted using the rover in the JAXA Space Exploration Field, which simulates the lunar environment. In the experiments, the rover traveled over level and sloped terrains and excavated the ground. The simulation and experimental results revealed similar trends in the traveling and excavation behaviors of the rover. These results can serve as basic guidelines for the design and operation of construction machinery on the Moon.

Utilizing building information modeling (BIM) for the analysis of existing pipelines necessitates the development of a swift and precise recognition method. Deep learning-based object recognition through imagery has emerged as a potent solution for tackling various recognition tasks. However, the direct application of these models is unfeasible due to their substantial computational requirements. In this research, we introduce a lightweight encoder explicitly for pipe recognition. By optimizing the network architecture using attention mechanisms, it ensures high-precision recognition while maintaining computational efficiency. The experimental results showcased in this study underscore the efficacy of the proposed lightweight encoder and its associated networks.

This study proposes an automatic findings generation method that performs in-context few-shot learning of a visual language model. The automatic generation of findings can reduce the burden of creating inspection records for infrastructure facilities. However, the findings must include the opinions and judgments of engineers, in addition to what is recognized from the image; therefore, the direct generation of findings is still challenging. With this background, we introduce in-context few-short learning that focuses on image similarity and text diversity in the visual language model, which enables text output with a highly accurate understanding of both vision and language. Based on a novel in-context few-shot learning strategy, the proposed method comprehensively considers the characteristics of the distress image and diverse findings and can achieve high accuracy in generating findings. In the experiments, the proposed method outperformed the comparative methods in generating findings for distress images captured during bridge inspections.

At unmanned construction sites, machinery generally stops functioning or tele-operation is interrupted owing to the problems related to wireless communication. However, the entry of human workers into such sites is often restricted, making it extremely difficult for workers to resolve problems or restore connections in wireless communication. In this study, we interviewed personnel from relevant industries to identify common issues in wireless communication. To address these issues, we focused on improving the availability of wireless communication and proposed a wireless communication status monitoring function. The results from a prototype of the proposed function are also presented.

Semantic information has started to be used in localization methods to introduce a non-geometric distinction in the environment. However, efficient ways to integrate this information remain a question. We propose an approach for fusing data from different object classes by analyzing the posterior for each object class to improve robustness and accuracy for self-localization. Our system uses the bearing angle to the objects’ center and objects’ class names as sensor model input to localize the user on a 2D annotated map consisting of objects’ class names and center coordinates. Sensor model input is obtained by an object detector on equirectangular images of a 360° field of view camera. As object detection performance varies based on location and object class, different object classes generate different likelihoods. We account for this by using appropriate weights generated by a Gaussian process model trained by using our posterior analysis. Our approach follows a systematic way to fuse data from different object classes and use them as a likelihood function of a Monte Carlo localization (MCL) algorithm.

Liquid waste is a type of dirt that is often found in toilets. Detection of liquid waste such as water or urine in the restroom is challenging due to their limited physical appearances, e.g., transparency and small size. This paper proposes a new method to detect water droplets, including water splashes, on the toilet floor by using the heat absorption capacity of liquid. Water, air, and floor have different heat capacity characteristics. Increasing temperature difference between water droplets and surroundings is done using blowing air on the surface of the detection area. A thermal camera is used to observe the detection area and an adaptive threshold is implemented to localize water droplets. This study also proposed a low-cost calibration chessboard method for thermal images that can produce good contrast images for calibrating wide-angle thermal camera modules. The results obtained from the experiment were promising, the system was able to detect single water drop up to 2 mm in diameter on a floor of 90 × 170 cm, and detection rate was above 95% for water droplets with a minimal size of 5 mm in diameter.

We developed a two-finger robot that can grasp, rotate in-hand, and place an uneven object. To improve the success rate of in-hand rotation, three pressure sensors were mounted on each finger to detect the center of pressure (CoP) on each finger. Two approaches, called object on-the-fly adjustment and re-grasp object movement (ROM), were proposed to use the obtained CoP information to adjust the position of each finger before performing in-hand rotation and were compared with a general in-hand rotation movement. Evaluation experiments were conducted on several objects, and the effectiveness of the proposed approaches was demonstrated with the highest success rate of 93% for ROM compared with 63% for general in-hand rotation.

In this study, the cam-like passive mechanism, known as the engage–disengage mechanism (EDM) of the intertarsal joint of ratites, and its design principles are investigated. This mechanism operates through the interplay of a muscle and three ligaments located on the medial and lateral sides of the intertarsal joint and the articular surface morphology of the tibiotarsus. The interplay of the musculoskeletal ligamentous elements creates two stable equilibrium points when they are almost fully extended and flexed. To elucidate the EDM in the intertarsal joints of ratites, we dissected the hindlimb of an emu (Dromaius novaehollandiae) and examined anatomical features around the joint. Subsequently, we replicated the intertarsal joint of ratites using a physical model. This model consists of three-dimensional-printed ostrich bones, coil springs, and nylon strings simulating the muscle and ligaments. This model successfully replicated the EDM and facilitated the analysis of the interplay of musculoskeletal ligamentous elements. We demonstrated that the medial ligaments and the morphology of the tibiotarsal articular surface play significant roles in facilitating the execution of EDM. Furthermore, we observed that the articular surface morphology resembles a well-known cam system in engineering and is responsible for the EDM.

In areas inaccessible to humans, such as the lunar surface and landslide sites, there is a need for multiple autonomous mobile robot systems that can replace human workers. Robots are required to remove water and sediment from landslide sites such as river channel blockages as soon as possible. Conventionally, several construction machines are deployed at civil engineering sites. However, owing to the large size and weight of conventional construction equipment, it is difficult to move multiple units of construction equipment to a site, which results in significant transportation costs and time. To solve such problems, this study proposes GREEMA: growing robot by eating environmental material, which is lightweight and compact during transportation and functions by eating environmental materials once it arrives at the site. GREEMA actively takes in environmental materials, such as water and sediment, uses them as its structure, and removes them by moving itself. In this study, two types of GREEMAs were developed and experimentally verified. First, we developed a fin-type swimming robot that passively takes in water into its body using a water-absorbing polymer and forms a body to express its swimming function. Second, we constructed an arm-type robot that eats soil to increase the rigidity of its body. We discuss the results of these two experiments from the viewpoint of explicit-implicit control and describe the design theory of GREEMA.

In this study, a simultaneous localization and mapping (SLAM) system for a two-wheeled mobile robot was developed in an indoor environment using RGB images, depth images, and wheel odometry. The proposed SLAM system applies planar motion constraints performed by a robot in a two-dimensional space to robot poses parameterized in a three-dimensional space. The formulation of these constraints is based on a conventional study. However, in this study, the information matrices that weigh the planar motion constraints are given dynamically based on the wheel odometry model and the number of feature matches. These constraints are implemented into the SLAM graph optimization framework. In addition, to effectively apply these constraints, the system estimates two of the rotation components between the robot and camera coordinates during SLAM initialization using a point cloud to construct a floor recovered from a depth image. The system implements feature-based Visual SLAM software. The experimental results show that the proposed system improves the localization accuracy and robustness in dynamic environments and changes the camera-mounted angle. In addition, we show that planar motion constraints enable the SLAM system to generate a consistent voxel map, even in an environment of several tens of meters.

Aiming at solving the problems such as long calibration time, low precision, and complex operation in hand-eye calibration of welding robot, an automatic hand-eye calibration algorithm based on linear structured light was proposed to solve the calibration matrix X by using AX=ZB calibration equation. Firstly, a square calibration plate is customized to effectively constrain the structured light. The α-shape algorithm was adopted to extract the contour of the 3D point cloud model of the calibration plate. Secondly, an improved random sampling consistency algorithm which could determine the optimal iterative number was proposed to fit the contour point cloud, the contour point cloud model fitted was obtained. Finally, the 3D coordinates of the target points were determined with the linear structured light to complete the hand-eye calibration. In order to prevent the calibration plate from deviating from the acquisition range of the vision sensor during the calibration process, the distance between the linear structural light and the inner circle in the calibration plate was set to limit the motion range of the robot. In order to eliminate the error transfer of the robot body, an optimal solution of the rotation matrix R and the translation vector t of the calibration data was calculated with the singular value decomposition (SVD) and the least square rigid transpose method. The experimental results show that the calibration accuracy reaches 0.3 mm without compensating the robot body error, and the calibration speed is improved by 36% than the existing automatic calibration method. Therefore, the algorithm proposed can automatically complete the calibration only by establishing the user coordinates in advance, which improves the working accuracy and efficiency of the welding robots greatly.

Although research has been conducted on the simultaneous operation of multiple autonomous underwater vehicles (AUVs) to improve the efficiency of oceanographic surveys, it has become clear that sonar interference occurs between AUVs that are in close proximity. In this study, a formation control system was developed for actual operation at sea, based on the assumption of an operational system with an autonomous surface vehicle (ASV) and multiple AUVs. A control algorithm was implemented to maintain the AUVs within the desired range of the ASV, ensure communication, and avoid sonar interference by sending commands to increase or decrease the speed of the AUVs. Simulations were performed to verify the effectiveness of the developed method for the operation of one ASV and 10 AUVs. Evaluation scores for the distance between vehicles were introduced to quantify the effectiveness of the algorithm. The proposed method obtained the highest scores for formation control in the assumed operational scenarios. This confirms the effectiveness of the proposed method in avoiding the side-by-side relationships that cause sonar interference.

Dynamic human movements are achieved by appropriate constraints on the degrees of freedom of the complex and flexible human body. The anatomy trains (ATs) theory explains such constraints with whole-body muscular connections called ATs. This paper proposes the design of a quasi-passive dynamic walker with whole-body viscoelastic connections inspired by the ATs theory and investigates the contributions of these long-distance connections to the achievement of gait. We designed a biped robot with a trunk and head, whose passive joints were supported by rubber fiber bands. The robot, named “PEARL III,” is equipped with an antagonistic pair of McKibben pneumatic actuators for each leg at the human hamstring and rectus femoris positions. The most important feature of this robot is that fabric wires mechanically connect its rubber bands and actuators on the back side from the head to the foot, modeled after one of the human ATs, the superficial back lines (SBLs). In an experiment, PEARL III achieved 2D quasi-passive dynamic walking on an inclined plane by contracting and relaxing its actuators using periodic feedforward control. This result suggests that in both the robot and human cases, when a controller contracts the SBL only in the stance phase during passive dynamic walking, the SBL can achieve whole-body posture control and weight support. In addition, the SBL appears to achieve this function depending on their mode of attachment to bones and the presence or absence of antagonistic muscles (or ATs). In the future, by introducing various ATs into robots while recognizing the importance of the appropriate attachment of ATs and the presence of their antagonistic muscles (or ATs), we can expect similar effects in various 3D movements.

We propose a methodology that uses the relative positional information of multiple antennas to estimate the Wrong Fix, which refers to an erroneous determination of the carrier-phase ambiguity utilized in GNSS satellites. The proposed approach is based on the fundamental notion that the mutual positional relationship of multiple antennas mounted on a mobile robot remains constant, and it uses machine-learning techniques based on the relative position information among the antennas to identify instances of Wrong Fixes. The relative distance between the antennas is derived from the real-time kinematic (RTK) position information of each antenna. The confidence level of the RTK positioning results was calculated using logistic regression, considering the measurement error with respect to the true value. To determine the Wrong Fixes, a labeled dataset was constructed, indicating that data were categorized as wrong fixes when the error from the true value exceeded 0.1 m. This dataset served as the training database for the logistic regression model. Experimental results demonstrate that the proposed methodology effectively reduced the root mean squared error between the measured location, classified as fixed by a trained discriminator, and the true value.

This paper proposes an admittance control scheme for robots equipped with joint-level position controllers involving deadtime. Its main feature is an elaborate discrete-time jerk limiter, which limits the third derivative of the position command sent to the controller. The jerk limiter is designed to suppress undesirable oscillation especially when the robot is in contact with a stiff environment. The controller is designed as a differential inclusion involving normal cones in the continuous-time domain, and its discrete-time algorithm is derived by the implicit Euler discretization. The presented controller was validated with experiments using a collaborative robot UR3e of Universal Robots, which has a deadtime of 6 ms in the velocity-command mode.