ICRA 2022 论文列表

Mean field approximation methodology has laid the foundation of modern Continuous Random Field (CRF) based solutions for the refinement of semantic segmentation. In this paper, we propose to relax the hard constraint of mean field approximation - minimizing the energy term of each node from probabilistic graphical model, by a global optimization with the proposed dilated sparse convolution module (DSConv). In addition, adaptive global average-pooling and adaptive global max-pooling are implemented as replacements of fully connected layers. In order to integrate DSConv, we design an end-to-end, time-efficient DilatedCRF pipeline. The unary energy term is derived either from pre-softmax and post-softmax features, or the predicted affordance map using a conventional classifier, making it easier to implement DilatedCRF for varieties of classifiers. We also present superior experimental results of proposed approach on the suction dataset comparing to other CRF-based approaches.

To achieve real-time dynamic simulation analysis and optimization design, a dynamic digital twin of a nonholonomic mobile manipulator (one UR5e mounted on an industrial mobile robot MIR 200) has been developed in this paper. First, the digital twin integrated with dynamics of a mobile manipulator is established. The framework of the dynamic digital twin is presented in detail. Then, the dynamic model of the system has been established with the consideration of the physical interaction between the robot and humans/environments using Lagrange formulation. Finally, the experimental testing has been conducted to validate the dynamic model and evaluate the performances (such as real-time property, accuracy, etc.) of the dynamic digital twin that is integrated with the physical human/environment-robot interaction.

Up-to-date High-Definition (HD) maps are essential for self-driving cars. To achieve constantly updated HD maps, we present a deep neural network (DNN), Diff-Net, to detect changes in them. Compared to traditional methods based on object detectors, the essential design in our work is a parallel feature difference calculation structure that infers map changes by comparing features extracted from the camera and rasterized images. To generate these rasterized images, we project map elements onto images in the camera view, yielding meaningful map representations that can be consumed by a DNN accordingly. As we formulate the change detection task as an object detection problem, we leverage the anchor-based structure that predicts bounding boxes with different change status categories. To the best of our knowledge, the proposed method is the first end-to-end network that tackles the high-definition map change detection task, yielding a single stage solution. Furthermore, rather than relying on single frame input, we introduce a spatio-temporal fusion module that fuses features from history frames into the current, thus improving the overall performance. Finally, we comprehensively validate our method's effectiveness using freshly collected datasets. Results demonstrate that our Diff-Net achieves better performance than the baseline methods and is ready to be integrated into a map production pipeline maintaining an up-to-date HD map.

In this paper, we present a robotic device for mouse tail vein injection. We propose a mouse holding mechanism to realize vein injection without anesthetizing the mouse, which consists of a tourniquet, vacuum port, and adaptive tail-end fixture. The position of the target vein in 3D space is reconstructed from a high-resolution stereo vision. The vein is detected by a simple but robust vein line detector. Thanks to the proposed two-staged calibration process, the total time for the injection process is limited to 1.5 minutes, despite that the position of needle and tail vein varies for each trial. We performed an injection experiment targeting 40 mice and succeeded to inject saline to 37 of them, resulting 92.5% success ratio.

Nowadays, wheeled mobile robots constitute a considerable portion of robots in industrial applications. Generally, regardless of their purpose, these systems are not designed to physically interact with humans, other robots, or the environment. In this study, we present a novel safe autonomous mobile - SAM - robot, which is a torque-controlled compliant robot that is conceived for safe human-robot interaction. This work provides an overview of the development philosophy of the system, its mechanical and mechatronics structure along with control and navigation architecture. Preliminary results show the advantages of the proposed mobile robot while interacting with its surroundings. We believe that this study will bring the wheeled mobile robots one step closer to the proactive interaction with their environment and humans surrounding them.

Post-stroke gait rehabilitation is necessary to aid social re-integration. An active knee orthosis (AKO) can aid gait training through the provision of bodyweight support and assistive knee torque. However, its use may cause instability and it does not ensure improved gait symmetry and speed. Use of a speed regulation device such as a robotic cane in conjunction with and AKO may overcome these limitations. Therefore, to combine the beneficial effects of an AKO and speed regulation, we have devised a gait assistance system (GAS) that combines our developed AKO with our Haptic cane (HC) to provide combined knee assistance and speed regulation. The system can provide constant speed regulation and proprioceptive input to improve gait speed, symmetry and balance, while providing assistive torque to improve knee range of motion. The system is evaluated through tests with nine healthy subjects who wore ankle weights on one leg to simulate hemiparesis. The results show that for majority of the outcome measures (gait and balance parameters), use of the GAS and HC generated significantly better results than use of only the AKO. However, for nearly all the measures there were no significant differences between GAS and HC. Thus, the results indicate that the HC and GAS may be used according to patient's condition, where more severe patients who require assistive torque may use the GAS and less severe patients may use only the HC.

Previous studies suggest that bipedal robots that have prismatic compliance in the legs can achieve efficient walking. However, how this efficiency can be achieved still remains an open research problem. In this study, we developed a 2-degree-of-freedom planar bipedal robot comprising Neidhart springs and five-bar parallel mechanisms. The five-bar parallel mechanism allows the robot to achieve prismatic compliance. In addition, a lightweight torque limiter function is achieved by Neidhart springs with an original design. We implemented a simple controller and conducted walking experiments. Experimental results showed that the robot walked effectively using the regenerative energy stored in the prismatic springs.

This paper describes the development of a 3- stage signalling framework to trigger a social robot's bottom- up reactive behavior inspired by a biological model. In the first stage, low-level firing of stimuli due to external sources is constructed through perception grounding. This is followed by a saliency classifier which fires-up high level salient signals that require attention and are used to trigger the robot's reactive behavior. The whole framework evolves primarily on the knowledge ontology that defines the characteristics of the social robot and the querying mechanism that correlates the perceived stimuli with the ontology to trigger the reactive behavior. We evaluated the performance of our system with timing metrics and we achieved good results for our application.

Detecting the onset/ongoing of slip, i.e. if a grasped object is slipping or will slip from the gripper while being lifted, is crucial. Conventionally, it is regarded as a tactile sensing related problem. However, recently multi-modal robotic learning has become popular and is expected to boost the performance. In this paper we propose a novel CNN-TCN model to fuse tactile and visual information for detecting the onset/ongoing of slip. In our experiments, two uSkin tactile sensors and one Realsense435i camera are used. Data is collected by randomly grasping and lifting 35 daily objects 1050 times in total. Furthermore, we compare our CNN-TCN model with the widely used CNN-LSTM model. As a result, our proposed model achieves a 88.75% detection accuracy and outperforms the CNN-LSTM model combined with different pretrained vision networks.

Lower limb exoskeletons have been successfully used in robotic-assisted rehabilitation. However, the design limitations of exoskeletons mechanics, such as weight and the lack of kinematic compatibility relative to the user's joints, limit the outcomes of treatment. To address these shortcomings, this work presents the design of a magneto-rheological fluid-based actuator for a knee exoskeleton, namely MRKE. The system was designed to ensure better mobility of the user, presenting high backdrivability and kinematic compatibility with the knee joint. The power train of system is a BLDC 70 W motor integrated to a harmonic drive gearbox. To improve kinematic compatibility relative to the user's knee, a four-bar crossed linkage mechanism (FBLM) was designed to follow the trajectory of the knee center of motion. A customized MR clutch was projected to decouple the motor-reducer from the FBLM, thus enabling high backdrivability. Preliminary results showed a small error (< 3 mm) between the FBLM and the knee center of rotation. Moreover, the MR clutch allowed for low backdrive torque (1.0 N.m) compared to the torque to backdrive the motor-reducer (16.6 N.m).

Autonomous latching is essential for autonomous surface vessels (ASV) to reach full independence from human intervention. As part of the ASV Roboat project, a new solution for self-latching maneuvers has been developed and is presented here. We propose a system that has the key requirements of full integration with the navigation control system and zero-gap connection with the dock, the latter being essential for wireless charging of the ASV. Dedicated markers are used to identify docking targets, relying on computer vision algorithms to determine distance and bearing to the target. In its idle state, the locking solution uses mechanical power-off brakes, minimizing energy consumption while ensuring the boat stays in position indefinitely once docked. A prototype of the proposed mechanism has been built and installed in Roboat. Experimental tests showing the mechanism performance and capability to autonomously approach the docking station are discussed in this work.

Tactile sensing typically involves active exploration of unknown surfaces and objects, making it especially effective at processing the characteristics of materials and textures. A key property extracted by human tactile perception in material classification is surface roughness, which relies on measuring vibratory signals using the multi-layered fingertip structure. Existing robotic systems lack tactile sensors that are able to provide high dynamic sensing ranges, perceive material properties, and maintain a low hardware cost. In this work, we introduce the reference design and fabrication procedure of a miniature and low-cost tactile sensor consisting of a biomimetic cutaneous structure, including the artificial fingerprint, dermis, epidermis, and an embedded magnet-sensor structure which serves as a mechanoreceptor for converting mechanical information to digital signals. The presented sensor is capable of detecting high-resolution magnetic field data through the Hall effect and creating high-dimensional time-frequency domain features for material texture classification. Additionally, we investigate the effects of different superficial sensor fingerprint patterns for classifying materials through both simulation and physical experimentation. After extracting time series and frequency domain features, we assess a k-nearest neighbors classifier for distinguishing between different materials. The results from our experiments show that our biomimetic tactile sensors with fingerprint ridges can classify materials with more than 7.7% higher accuracy and lower variability than ridge-less sensors. These results, along with the low cost and customizability of our sensor, demonstrate high potential for lowering the barrier to entry for a wide array of robotic applications, including modelless tactile sensing for texture classification, material inspection, and object recognition.

This paper introduces a design method for an efficient and agile quadruped robot. A mixed-integer optimization formulation including the number of gear teeth is derived to obtain the optimal gear ratio that minimizes cost for a running-trot with the target speed of 3 m/s. With the inclusion of integer constraints related to the number of gear teeth, detailed design considerations of gear trains can be included in the optimization process. Thermal dissipation of the motor controller is also taken into account in the optimization to consider heat generation during high-speed running. KAIST Hound, a 45 kg robot, designed with the obtained design parameters has successfully demonstrated a 3 m/s running-trot using a nonlinear model predictive controller (NMPC). Furthermore, the robot has proved its robustness by the demonstration of additional experiments such as 22° slope climbing, 3.2 km walking, and traversing a 35 cm obstacle.

In recent years, a rising trend of floor cleaning robots could be observed in the consumer electronic market. Area coverage performance is a crucial factor that determines the overall productivity of a floor cleaning robot. Nevertheless, the area coverage performance of commercially available floor cleaning robots is limited due to narrow spaces resulting from complex furniture arrangements. Traditionally, new robot designs (both hardware and algorithmic) are explored to over-come the coverage limitations. Developments of reconfiguration mechanisms and path planning algorithms for floor cleaning robots could be considered as examples. This paper proposes a novel concept called “design by robot,” enabling a floor cleaning robot to make suggestions on workspace modifications to maximize its area coverage performance in a given workspace. In this regard, the robot analyzes a workspace to be cleaned through internal simulations based on the metric map of the workspace. A metaheuristic optimization technique determines the optimum placings of objects. Particle Swarm Optimization (PSO), Surrogate Optimization (SO), and Generalized Pattern Search (GPS) are individually used in this regard. Experiments, including scenarios of robot deployments, have been considered for validation. The statistical outcomes of the experimental results validate that the area coverage performance of a floor cleaning robot could be significantly improved by considering the workspace modifications suggested by the robot. Moreover, the proposed concept “design by robot” enables users to gain significantly improved performance from floor cleaning robots through collaboration.

This paper presents a novel design and experimental investigation for a self-propelled capsule robot that can be used for painless colonoscopy during a retrograde progression from the patient's rectum. The steerable robot is driven forward and backward via its internal vibration and impact with orientation control by using an electromagnetic actuator. The actuator contains four sets of coils and a shaft made by permanent magnet. The shaft can be excited linearly in a controllable and tilted angle, so guide the progression orientation of the robot. Two control strategies are studied in this work and compared via simulation and experiment. Extensive results are presented to demonstrate the progression efficiency of the robot and its potential for robotic colonoscopy.

In this paper, a novel adjustable-stiffness force sensor is developed for multitask measurements requiring different force resolutions and ranges. The applied force of the force sensor is indirectly measured through the linear deformation instead of the structure strain through an optical linear encoder. The main structure of the force sensor is actually a linear variable stiffness mechanism with a compact size and a large stiffness change. Its stiffness can be continuously adjusted by changing the effective second moment of area of the structure. Thus, the force sensor has an adjustable range and resolution since the displacement resolution of the optical linear encoder is constant. The stiffness modeling of the sensor is performed based on the matrix method, which is then evaluated by the finite element analysis. A principle prototype is finally fabricated for the adjustable-stiffness test and a concrete application example. The testing results show that the stiffness and resolution of the force sensor can be changed by the proposed stiffness adjustment. Moreover, it is effective to measure different-resolution forces. This adjustable-stiffness approach can be also extended to the design of a torque sensor or a force/torque sensor.

Continuum surgical robots can navigate anatomical pathways to reach pathological locations deep inside the human body. Their flexibility, however, generally comes with reduced dexterity at their tip and limited workspace. Building on recent work on eccentric tube robots, this paper proposes a new continuum robot architecture and theoretical framework that combines the flexibility of push/pull actuated snake robots and the dexterity offered by concentric tube robotic end-effectors. We designed and present a prototype system as a proof-of-concept, and developed a tailored quasistatic mechanics-based model that describes the shape and end-effector's pose for this new type robotic architecture. The model can accommodate an arbitrary number of arms placed eccentrically with respect to the backbone's neutral axis. Our experiments show that the error between model and experiment is on average 3.56% of the manipulator's overall length. This is in agreement with state of the art models of single type continuum architecture.

In this paper, a new capture system for UAV precision landing in a disturbed environment is proposed. Compared with the traditional visual guided landing methods, perching mechanism based methods, and tethered landing methods, the proposed system takes into account the stability during landing process and retains the high accessibility of the UAV. The proposed system consists of a winch subsystem and a magnetic catcher device. They establish an automatic tethered-UAV system for landing before the UAV touchdown. We analyzed the design principle as well as the feasibility of the magnetic catcher. An optimization problem is formulated to obtain a better layout of magnets on the catcher. The problem is relaxed based on interpolation simulation of attraction force. Experiments are conducted both in indoor and outdoor environments based on different UAV platforms respectively. The results validate that the catcher design and the capture system can achieve a successful landing in both cases.

Rolling spherical robots have been studied in the past few years as an alternative to legged and wheeled robots in unstructured environments. These systems are of uttermost interest for space exploration: fast, robust to collision and able to handle various terrain topologies. This paper introduces a novel barycentric spherical robot, dubbed the Autonomous Robotic Intelligent Explorer Sphere (ARIES). Equipped with an actuated cylindrical joint acting as a pendulum with two degrees-of-freedom (DoF), the ARIES has a continuous differential transmission to allow simultaneous rolling and steering. This mechanism allows an unprecedented mass allocation optimization, notably to provide a low center of mass. Kinematics and dynamics of this novel system are detailed. An analysis of the steering mechanism proves that it is more efficient than a more conventional 2-DoF tilting mechanism, while also retaining more space for a payload, for instance to host sensors for simultaneous localization and mapping, in the upper part of the sphere. Moreover, the kinematic input/output equations obtained significantly simplify the device's control. Finally, we present a first complete prototype with preliminary experimental tests.