IROS 2018 论文列表

For autonomous multi-robot teams, the individual team members are tasked with completing their assigned tasks as defined by a team plan provided by a centralized team planner. However in complex dynamic domains, the team plans are generated by the team planner with assumptions due to the complexity of modeling the domain. Failures in execution are therefore inevitable for the team members, and as such, replanning will occur for the team. In this paper, we introduce a rationale-driven team plan representation that provides rationales on why actions were chosen by the team planner. During a failure, the individual team members autonomously use our described intra-robot replanning algorithm to select all applicable replan policies for a given rationale. We then describe a method to learn the predicted cost of each replan policy, given a state of the environment, in order for the individual robots to select the lowest costing replan policy to improve team performance.

Problem benchmarks are used in experimental science as a reference against which results of experiments using distinct approaches to solve the problem are compared and evaluated in relative terms. In Robotics, just formulating a general performance assessment problem is difficult per se, as robot systems are composed of very diverse subsystems (e.g., localisation, human-robot interaction, task planning, motion planning). This paper introduces a probabilistic approach to benchmarking and evaluating performance of robot systems, which uses probability theory as the common language to quantify the performance of distinct functionalities of a robot system and their impact on the performance of a task carried out by that system. The approach can be used to analyse the performance of a task plan from the performances if its composing functionalities, or to (re)plan when a performance degradation in functionality is predicted to cause performance degradation of the task plan beyond acceptable limits.

This paper proposes a practical approach for fastening the lengthy experimentational processes that may occur with iterative learning control (ILC) upto a certain level using simple low-order identified models. The traditional practice in ILC experiments is to update the ILC signal by directly using the experimental data after each run of the process which corresponds to one ILC update per one run. When considered from the point of experimental time, even conducting a moderate number of ILC updates can take quite long with this procedure. Since an accurate linear model can adequately represent the actual system upto a certain amplitude and/or frequency of the desired reference, we propose that the total experimental time can be reduced by updating the ILC signal via predicted system data until the limits of the linear model. This approach allows one to carry out large number of ILC updates while not needing to carry out the same amount of real experiments. Consequently, a significant number of experiments that would be needed for achieving the same results can be skipped with a simulation approach. The efficiency of the proposed method was tested through experimentation with three different UAV reference trajectories and the results demonstrated that it is possible to attain significant amount of tracking precision in several flight experiments.

In this paper, the problem of suppressing sloshing dynamics in liquid handling robotic systems has been faced by designing a dynamic filter that starting from the desired motion of the liquid container calculates the complete position/orientation trajectory for the robot end-effector. Specifically, a design philosophy mixing a filtering technique that suppresses the frequency contributions of the reference motion that may cause liquid oscillations and an active compensation of lateral accelerations by a proper container re-orientation has been adopted. In principle, the latter contribution requires the knowledge of acceleration of the reference trajectory, but because of the use of an harmonic smoother that performs a shaping of the original motion, it is possible to obtain the value of the acceleration in runtime. In this way, the proposed methods can be applied also to reference motions that are not known in advance, e.g. commands directly provided by a human operator. This possibility has been demonstrated by means of a number of experimental tests in which the user teleoperates the robot carrying the container with the liquid by simply moving in the free space its hand, whose 3D position is detected by a motion capture system.

Although there has been recent progress in control of multi-joint prosthetic legs for periodic tasks such as walking, volitional control of these systems for non-periodic maneuvers is still an open problem. In this paper, we develop a new controller that is capable of both periodic walking and common volitional leg motions based on a piecewise holonomic phase variable through a finite state machine. The phase variable is constructed by measuring the thigh angle, and the transitions in the finite state machine are formulated through sensing foot contact together with attributes of a nominal reference gait trajectory. The controller was implemented on a powered knee-ankle prosthesis and tested with a transfemoral amputee subject, who successfully performed a wide range of periodic and non-periodic tasks, including low- and high-speed walking, quick start and stop, backward walking, walking over obstacles, and kicking a soccer ball. The proposed approach is expected to provide better understanding of volitional motions and lead to more reliable control of multi-joint prostheses for a wider range of tasks.

In this paper we demonstrate a universal jamming gripper with a partially filled membrane that can pick up submerged objects resting on soft substrates. Jamming grippers take advantage of the phenomenon of particle jamming to control the compliance of an end effector membrane. Changes in internal membrane pressure are used to transition the membrane between hard and soft states. The effort was motivated by the need for tools to sample artifacts on deep sea shipwrecks, which are often found resting on waterlogged timbers, or partially buried in fine, loose sediment. Limiting downward force protects the target, and reduces the likelihood that it will be pushed down in to the substrate, which could lead to a failed grasp. In benchtop tests, the downward force, and the ratio of maximum lifting force to downward force, are shown to be strongly dependent on the initial volume of particles and fluid in the gripper membrane. The gripper achieves lifting forces 6.7 times the downward force on targets with high aspect ratios. Experiments in a fresh water tank demonstrate the ability to grasp objects resting on soft sediment, and compliant foam. Finally, experiments at sea demonstrate that the end effector functions at depths of more than 1000m seawater, successfully grasping a range of irregular objects.

This paper presents a parallel robotic mechanism for endoscope tip stabilization and guidance for a robot-assisted minimally invasive laser osteotome. The mechanism attaches to the bone of the patient, providing a stable and robust platform for the laser integrated in the endoscope tip which has to be moved precisely in the sub-millimeter range along a preoperatively planned path. This method is only possible because cutting bone with laser instead of using conventional bone drills and saws involves considerably lower interaction forces. The design, kinematics, control, and motion performance of the concept are presented for an upscaled prototype. The obtained deviation of the endoscope tip motion from the reference path lies in the sub-millimeter range. This result allows us to conclude that the concept is more than promising. Furthermore, we expect that the herein presented principle will influence the way osteotomies will be performed in the future.

This paper investigates the design of a soft elbow exosuit capable of providing supplemental lifting assistance by reducing muscle activity of the bicep muscle. The aim is to improve the efficiency and endurance of workers who are tasked with repetitive lifting. The design consists of an array of pneumatically pressurized soft actuators, which are encased in nylon fabric that allows for a high force-to-weight ratio of 211.SN/g. An analytical model governing the bending behavior of two consecutive actuators and torque generated by the exosuit is developed, with test results showing less than 10% error from the theoretical model. An elbow joint torque value of 27.6N.m is achieved at 300kPa, which is comparable to the 30N.m maximum set by OSHA requirements in the USA. Further testing with a healthy participant is performed using surface electromyography (sEMG) sensors and a motion capture system to assess the capabilities of the exosuit to provide active assistance to the bicep during isometric and concentric contractions. Measurable assistance to lifting is observed with minimal obstruction to the user's range of motion for all experiments.

In this paper, we analyze systems with low throughput human-machine interfaces (such as a brain-computer interface, single switch interface) from the controls perspective. We develop some principles for performance improvement in such systems based on the parallelization of inference and robot motion. The proposed principles are used to design a novel shared position control to navigate a circular massless holonomic robot in a known environment. The system is implemented in simulation and integrated with a real robotic wheelchair. Robot experiments demonstrated the viability of the proposed navigation method in various modes of operation.

Autonomous elevator operation is considered an intelligent solution in handling the inter-floor navigation problem of service robots. As one of the most fundamental steps, elevator button recognition starts to receive more and more attention. However, due to the challenging image conditions and severe class imbalance problem, the performance of existing results is unsatisfying. In this paper, we propose to combine an optical character recognition (OCR) network and the Faster RCNN architecture into a single neural network, called OCR-RCNN to facilitate an end-to-end training and elevator button recognition procedure. To verify our method, we collect a large dataset of elevator panels and carry out extensive comparative experiments. The experiment results show that our method can greatly outperform the traditional recognition pipelines, yielding an accurate and robust performance on recognizing untrained elevator buttons.

This paper proposes a unique and robust UAV autonomous navigation approach along one side of overhead transmission lines for inspection. To this end, we establish a perspective model and develop a novel Pan/Tilt monocular-based navigation scheme. Simultaneously, the following three key issues are addressed. First, to locate the effective landmark - transmission tower timely and reliably, we customize a neural network for tower detection and combine it with a fast and smooth tracking. Second, to provide UAV with a robust and precise heading, we detect the transmission lines and compute and optimize their vanishing point. Third, to keep a safe distance from transmission lines, we optimize a homography matrix to restore the parallel nature of transmission lines and perceive the distance variation by a point set registration model. Finally, by the designed UAV platform, we test the whole system in a real-world transmission-line inspection scenario under different weather condition and achieve an encouraging result. Our approach provides great flexibility for refined inspection and effectively improves inspection safety.

This work presents a novel method for estimating online the torques at the ankle, knee and hip of a user with the goal of generating reference signals for torque controlled lower limb exoskeletons. In particular, this approach attempts to address difficulties arising in real life scenarios when noncyclic locomotion activity, unexpected terrain or unpredicted interactions with the surroundings occur. An advantage of the proposed method is that it does not require any information on the user's upper body (i.e. pose, weight and center of mass location)or on any interaction of the user's upper body with the environment (i.e. payload handling or pushing and pulling task). By monitoring the interaction of the user's feet with the ground through a novel sensorized shoe sensing system, the method applies an inverse static analysis on the user's lower limbs to estimate in real time the torque at each leg joint. The system is fully wearable, ergonomic and portable and uses a reduced number of body posture sensors. The design of the sensorized shoes permits plantar flexion, while measuring the toe and heel orientation and the interaction loads. This allows walking on irregular terrains and natural feet postures in different tasks. Trials were performed to validate the proposed approach under different tasks and terrains. Finally, the knee torques estimated online by the proposed strategy were used as reference signals to drive the iT-Knee Bipedal System in an assistive task.

In this paper, we propose a novel input device for a robotic prosthetic hand based on capacitance change. The proposed device can sense the deformation of the skin due to the activity of the muscle by measuring the capacitance change between the skin and the electrode. Therefore, it can be used as a sensor for estimating a user's intention through medical electrodiagnostic techniques such as electromyogram (EMG)and force myography (FMG). The proposed device can acquire data in a non-invasive way and is advantageous with easier data processing than that for an EMG signal. Moreover, it is resistant to the impedance change of skin because the capacitance is measured in a non-contact manner, unlike the existing methods which work with direct contact with the skin such as EMG. Additionally, unlike FMG, the device is lightly attached to the skin without being strongly fixed with velcro, which offsets problems that occur while re-wearing the device. To demonstrate the feasibility of the proposed idea, three of the newly developed input devices were used to classify four hand movements (fist, scissors, paper, and rest)using a multilayer perceptron (MLP). As a result, the classification success rates for the fist, paper, scissor, and rest motions were obtained as 99.3%, 98.3%, 98.4%, and 99.1%, respectively.

This paper presents a new application of atmospheric pressure plasma jet (APPJ) aiming for fabricating a microfluidic system on a polydimethylsiloxane (PDMS) surface. While PDMS is widely used for microfluidic chips, the fabrication of a chip requires different instruments which are not easily accessible for small-scale companies or laboratories. Therefore, we are motivated to develop a simple and low-cost method for such a fluidic system fabrication. The idea of this work is to directly pattern a fluidic channel on a PDMS surface with a plasma jet, which is known for its capability of modifying the hydrophobicity on a surface. The feasibility test first showed that fluid only flows in plasma-treated regions as having physical walls. The plasma parameters were then optimized using Taguchi method based on experiments. The optimization significantly reduce the required plasma treating time from more than 30 treating rounds to only 3 treating rounds, over ten times improved. Methods for further improving the resolution to micrometer-scale have been discussed. In addition to the advantages of fast and low-cost of the proposed method, making microfluidic channels on the surfaces of PDMS chip is also convenient for recollecting cultured cells on a chip in the field of regenerative medicine.

In this study, we propose an extended coaxial spherical joint module (E-CoSMo) with three to four degrees of freedom (DOFs) for a multi-DOF robot platform. The E-CoSMo consists of a coaxial spherical parallel mechanism (CSPM) with three DOFs and one extended DOF based on a universal joint mechanism (UJM) coaxially connected to the CSPM. This structure enables the application of serial link configuration (such as shoulder-elbow) with wide and universal ROMs while allowing all four actuators to be placed in the base. This makes the inertia of the moving link part to be dramatically reduced and thus contributes to decreasing the mechanical impedance of the multi-DOF robot system. In addition, through the effective design of the coaxial spherical joint module, the output rotational torque in a specific axial direction reaches approximately three times then the torque of a single actuator. To optimally implement this, we applied an optimal design approach that considers the mechanical performance and design constraints. The mechanical impedance reduction effect through the proposed module is discussed. The feasibility of the E-CoSMo is also verified through a dynamic simulation. Finally, the proposed mechanism is verified using a fabricated prototype.

Many surgical robots are remotely actuated by means of wire cables. In the past, the cables wound around circular pulleys at the robot joints did not constitute a problem of the cable driver structure. However, the pulleys inside the joints are removed recently in order to miniaturize the joints, so a specially designed cable driver suitable for the miniature joint structure is required for stable driving. In this paper, we propose a novel cable driver design for driving a pulleyless rolling joint and extend it to 2-DOF structure. Then, the proposed cable driver is manufactured using 3D printing with the 2-DOF bending joint, and an experiment is performed to evaluate them using the prototype. The cable driver proposed in this paper can drive pulleyless rolling joints stably with low cable tension. In addition, it can decouple yaw and pitch motion of the joints completely, therefore it can be applied to a variety of thin robots and instruments including steerable endoscopes and surgical robots.

This paper presents a novel agricultural robot for greenhouse applications. In many greenhouses, including the greenhouse used in this work, sets of pipes run along the floor between plant rows. These pipes are components of the greenhouse heating system, and doubles as rails for trolleys used by workers. A flat surface separates the start of each rail set at the greenhouse headland. If a robot is to autonomously drive along plant rows, and also be able to move from one set of rails to the next, it must be able to locomote both on rails and on flat surfaces. This puts requirements on mechanical design and navigation, as the robot must cope with two very different operational environments. The robot presented in this paper has been designed to overcome these challenges and allows for autonomous operation both in open environments and on rails by using only low-cost sensors. The robot is assembled using a modular system created by the authors and tested in a greenhouse during ordinary operation. Using the robot, we map the environment and automatically determine the starting point of each rail in the map. We also show how we are able to identify rails and estimate the robots pose relative to theses using only a low-cost 3D camera. When a rail is located, the robot makes the transition from floor to rail and travels along the row of plants before it moves to the next rail set which it has identified in the map. The robot is used for UV treatment of cucumber plants.

A sensing glove is an attractive application of wearable devices. Soft sensors are emerging to replace rigid sensing units, especially for wearable sensor systems, due to its inherent softness, flexibility, and stretchability. However, the fabrication process for the soft sensors is usually complex, time-consuming, labor-intensive, and has low production rate. To integrate a sensor system, an assembly process is essential, which may make the system bulky. Moreover, a solution for the electrode parts has rarely been suggested, although a bulky electrode part may obstruct the user's movement and degrade performance of the sensor. Thus, in this study, a novel fabrication process is suggested based on direct ink writing (DIW) of eutectic gallium-indium (EGaIn), which forms all the items in the sensor system from the sensing units, wiring, and the electrode part. A sensing glove for 2D finger motions was fabricated, and its performance was verified in terms of linearity, dynamic response, and accuracy. The sensing glove can be used as an easily-wearable and an intuitive interface to the virtual reality environment.

Efforts to engineer insect-sized (~100 mg) robots are motivated by their potential advantages relative to larger robots, such as greater deployment numbers at the same cost. Previous iterations have demonstrated controlled flight, but were limited in terms of locomotion capabilities outside of flight. They also consisted of many parts, making them difficult to fabricate. Here we present a re-design that lowers the center of mass, allowing the robot to additionally land without the need for long legs. Furthermore, we show that the new design allows for wing-driven ground locomotion. This is achieved by varying the speed of downstroke relative to the upstroke of the flapping wings, which also allows for steering. By landing and subsequently moving along the ground, the robot can negotiate extremely confined spaces and underneath obstacles, as well as navigate to precise locations for sensing operations. The new design also drastically reduces the number of parts, simplifying fabrication. We describe the new design in detail and present results demonstrating these capabilities, as well as feedback-stabilized flights.