ICRA 2020 论文列表

Autonomous navigation in unknown environments with the intent of exploring all traversable areas is a significant challenge for robotic platforms. In this paper, a simple yet reliable method for exploring unknown environments is presented based on bio-inspired reactive control and metric-topological planning. The reactive control algorithm is modeled after the spatial decomposition of wide and small-field patterns of optic flow in the insect visuomotor system. Centering behaviour and small obstacle detection and avoidance are achieved through wide-field integration and Fourier residual analysis of instantaneous measured nearness respectively. A topological graph is estimated using image processing techniques on a continuous occupancy grid. Node paths are rapidly generated to navigate to the nearest unexplored edge in the graph. It is shown through rigorous field-testing that the proposed control and planning method is robust, reliable, and computationally efficient.

We present an approach for estimating the body-frame velocity of a mobile robot. We combine measurements from a millimeter-wave radar-on-a-chip sensor and an inertial measurement unit (IMU) in a batch optimization over a sliding window of recent measurements. The sensor suite employed is lightweight, low-power, and is invariant to ambient lighting conditions. This makes the proposed approach an attractive solution for platforms with limitations around payload and longevity, such as aerial vehicles conducting autonomous exploration in perceptually degraded operating conditions, including subterranean environments. We compare our radar-inertial velocity estimates to those from a visual-inertial (VI) approach. We show the accuracy of our method is comparable to VI in conditions favorable to VI, and far exceeds the accuracy of VI when conditions deteriorate.

It has been demonstrated that the performance of an object detector degrades when it is used outside the domain of the data used to train it. However, obtaining training data for a new domain can be time consuming and expensive. In this work we demonstrate how a radar can be used to generate plentiful (but noisy) training data for image-based vehicle detection. We then show that the performance of a detector trained using the noisy labels can be considerably improved through a combination of noise-aware training techniques and relabelling of the training data using a second viewpoint. In our experiments, using our proposed process improves average precision by more than 17 percentage points when training from scratch and 10 percentage points when fine-tuning a pre-trained model.

With the availability of small system in package realizations, radar systems become more and more attractive for a variety of applications in robotics, in particular also for collaborative robotics. As the simulation of robot systems in realistic scenarios has become an important tool, not only for design and optimization, but also e.g. for machine learning approaches, realistic simulation models are needed. In the case of radar sensor simulations, this means providing more realistic results than simple proximity sensors, e.g. in the presence of multiple objects and/or humans, objects with different relative velocities and differentiation between background and foreground movement. Due to the short wavelength in the millimeter range, we propose to utilize methods known from computer graphics (e.g. z-buffer, Lambertian reflectance model) to quickly acquire depth images and reflection estimates. This information is used to calculate an estimate of the received signal for a Frequency Modulated Continuous Wave (FMCW) radar by superposition of the corresponding signal contributions. Due to the moderate computational complexity, the approach can be used with various simulation environments such as V-Rep or Gazebo. Validity and benefits of the approach are demonstrated by means of a comparison with experimental data obtained with a radar sensor on a UR10 arm in different scenarios.

We present a novel key region extraction method of point cloud, ROI-cloud, for LiDAR odometry and localization with autonomous robots. Traditional methods process massive point cloud data in every region within the field of view. In dense urban environments, however, processing redundant and dynamic regions of point cloud is time-consuming and harmful to the results of matching algorithms. In this paper, a voxelized cube set, ROI-cloud, is proposed to solve this problem by exclusively reserving the regions of interest for better point set registration and pose estimation. 3D space is firstly voxelized into weighted cubes. The key idea is to update their weights continually and extract cubes with high importance as key regions. By extracting geometrical features of a LiDAR scan, the importance of each cube is evaluated as a new measurement. With the help of on-board IMU/odometry data as well as new measurements, the weights of cubes are updated recursively through Bayes filtering. Thus, dynamic and redundant point cloud inside cubes with low importance are discarded by means of Monte Carlo sampling. Our method is validated on various datasets, and results indicate that the ROI-cloud improves the existing method in both accuracy and speed.

This paper presents a new technique to solve the Indoor Visual Place Recognition problem from the Deep Learning perspective. It consists on an image retrieval approach supported by a novel image similarity metric. Our work uses a 3D laser sensor mounted on a backpack with a calibrated spherical camera i) to generate the data for training the deep neural network and ii) to build a database of geo-referenced images for an environment. The data collection stage is fully automatic and requires no user intervention for labelling. Thanks to the 3D laser measurements and the spherical panoramas, we can efficiently survey large indoor areas in a very short time. The underlying 3D data associated to the map allows us to define the similarity between two training images as the geometric overlap between the observed pixels. We exploit this similarity metric to effectively train a CNN that maps images into compact embeddings. The goal of the training is to ensure that the L2 distance between the embeddings associated to two images is small when they are observing the same place and large when they are observing different places. After the training, similarities between a query image and the geo-referenced images in the database are efficiently retrieved by performing a nearest neighbour search in the embeddings space.

Origami robots are well-suited for jumping maneuvers because of their light weight and ability to incorporate actuation and control strategies directly into the robot body. However, existing origami robots often model fold patterns as rigidly foldable and fail to take advantage of deformation in an origami sheet for potential energy storage. In this paper, we consider a parametric origami tessellation, the Reconfigurable Expanding Bistable Origami (REBO) pattern, which leverages face deformations to act as a nonlinear spring. We present a pseudo-rigid-body model for the REBO for computing its energy stored when compressed to a given displacement and compare that model to experimental measurements taken on a mechanical testing system. This stored potential energy, when released quickly, can cause the pattern to jump. Using our model and experimental data, we design and fabricate a jumping robot, REBOund, that uses the spring-like REBO pattern as its body. Four lightweight servo motors with custom release mechanisms allow for quick compression and release of the origami pattern, allowing the fold pattern to jump over its own height even when carrying 5 times its own weight in electronics and power. We further demonstrate that small geometric changes to the pattern allow us to change the jump height without changing the actuation or control mechanism.

The International Space Station (ISS) serves as a research lab for a wide variety of experiments including some that study the biological effects of microgravity and spaceflight using the Rodent Habitat and Microgravity Science Glovebox (MSG). Astronauts train for onboard dissections of rodents following basic training. An alternative approach for conducting these experiments is teleoperation of a robot located on the ISS from earth by a scientist who is proficient in rodent dissection. This pilot study addresses (1) the effects of extreme time delay on skill degradation during Fundamentals of Laparoscopic Surgery (FLS) tasks and rodent dissections using RAVEN II; (2) derivation and testing of rudimentary interaction force estimation; (3) elicitation of design requirements for an onboard dissection robot, RAVEN-S; and (4) simulation of the RAVEN-S prototype design with dissection data. The results indicate that the tasks' completion times increased by a factor of up to 9 for a 3 s time delay while performing manipulation and cutting tasks (FLS model) and by a factor of up to 3 for a 0.75 s time delay during mouse dissection tasks (animal model). Average robot forces/torques of 14N/0.1Nm (peak 90N/0.75Nm) were measured along with average linear/angular velocities of 0.02m/s/4rad/s (peak 0.1m/s/40rad/s) during dissection. A triangular configuration of three arms with respect to the operation site showed the best configuration given the MSG geometry and the dissection tasks. In conclusion, the results confirm the feasibility of utilizing a surgically-inspired RAVEN-S robot for teleoperated rodent dissection for successful completion of the predefined tasks in the presence of communications time delay between the ISS and ground control.

We introduce R3T, a reachability-based variant of the rapidly-exploring random tree (RRT) algorithm that is suitable for (optimal) kinodynamic planning in nonlinear and hybrid systems. We developed tools to approximate reachable sets using polytopes and perform sampling-based planning with them. This method has a unique advantage in hybrid systems: different dynamic modes in the reachable set can be explicitly represented using multiple polytopes. We prove that under mild assumptions, R3T is probabilistically complete in kinodynamic systems, and asymptotically optimal through rewiring. Moreover, R3T provides a formal verification method for reachability analysis of nonlinear systems. The advantages of R3T are demonstrated with case studies on nonlinear, hybrid, and contact-rich robotic systems.

This paper introduces an intrinsically safe parallel manipulator dedicated to fast pick-and-place operations, called R-Min. It has been designed to reduce the risk of injury during a collision with a human operator, while maintaining high speed and acceleration. The proposed architecture is based on a modification of the well-known planar five-bar mechanism, where additional passive joints are introduced to the distal links in order to create a planar seven-bar mechanism with two degrees of underactuation, so that it can passively reconfigure in case of collision. A supplementary passive leg, in which a tension spring is mounted, is added between the base and the end-effector in order to constrain the additional degrees of freedom. A prototype of this new collaborative parallel robot is designed and its equilibrium configurations under several types of loadings are analyzed. Its dynamics is also studied. We analyze the impact force occurring during a collision between our prototype and the head of an operator and compare these results with those that would have been obtained with a rigid five-bar mechanism. Simulation results of impact during a standard pick-and-place trajectory of duration 0.3 s show that a regular five-bar mechanism would injure a human, while our robot would avoid the trauma.

The current state-of-the-art for testing and evaluation of autonomous surface vehicle (ASV) decision-making is currently limited to one-versus-one vessel interactions by determining compliance with the International Regulations for Prevention of Collisions at Sea, referred to as COLREGS. Strict measurement of COLREGS compliance, however, loses value in multi-vessel encounters, as there can be conflicting rules which make determining compliance extremely subjective. This work proposes several performance metrics to evaluate ASV decision-making based on the concept of "good seamanship," a practice which generalizes to multi-vessel encounters. Methodology for quantifying good seamanship is presented based on the criteria of reducing the overall collision risk of the situation and taking early, appropriate actions. Case study simulation results are presented to showcase the seamanship performance criteria against different ASV planning strategies.

This paper presents a method of computing free motions of a planar assembly of rigid bodies connected by loose joints. Joints are modeled using local distance constraints, which are then linearized with respect to configuration space velocities, yielding a linear programming formulation that allows analysis of systems with thousands of rigid bodies. Potential applications include analysis of collections of modular robots, structural stability perturbation analysis, tolerance analysis for mechanical systems, and formation control of mobile robots.

Selectively picking a target fruit surrounded by obstacles is one of the major challenges for fruit harvesting robots. Different from traditional obstacle avoidance methods, this paper presents an active obstacle separation strategy that combines push and drag motions. The separation motion and trajectory are generated based on the 3D visual perception of the obstacle information around the target. A linear push is used to clear the obstacles from the area below the target, while a zig-zag push that contains several linear motions is proposed to push aside more dense obstacles. The zig-zag push can generate multi-directional pushes and the side-to-side motion can break the static contact force between the target and obstacles, thus helping the gripper to receive a target in more complex situations. Moreover, we propose a novel drag operation to address the issue of mis-capturing obstacles located above the target, in which the gripper drags the target to a place with fewer obstacles and then pushes back to move the obstacles aside for further detachment. Furthermore, an image processing pipeline consisting of color thresholding, object detection using deep learning and point cloud operation, is developed to implement the proposed method on a harvesting robot. Field tests show that the proposed method can improve the picking performance substantially. This method helps to enable complex clusters of fruits to be harvested with a higher success rate than conventional methods.

This paper presents a monocular vision based proximity estimation system using abstract features, such as corner points, blobs and edges, as inputs to a neural network. An experimental vehicle was built using a vision system integrating the SCAMP-5 vision chip, a micro-controller, and an RC model car. The vision chip includes image sensor with embedded 256×256 processor SIMD array. The pixel processor array chip was programmed to capture images and run the feature algorithms directly on the focal plane, and then digest them so that only sparse feature description data were read-out in the form of 40 values. By logging the vision output and the output from three infrared proximity sensors, training data were obtained to train three fully connected layer-recurrent neural networks with fewer than 700 parameters each. The trained neural network was able to estimate the proximity to the level of accuracy sufficient for a reactive collision avoidance behaviour to be achieved. The latency of the control system, from image capture to neural network output, was under 4ms, enabling the vehicles to avoid obstacles while moving at 0.64m/s to 1.8m/s in the experiment.

Pneumatic transmission has several advantages in developing powered ankle foot orthosis (AFO) systems, such as the flexibility in placing pneumatic components for mass distribution and providing high back-drivability via simple valve control. However, pneumatic systems are generally tethered to large stationary air compressors that restrict them for being used as daily assistive devices. In this study, we improved a previously developed wearable (untethered) custom compressor that can be worn (1.5 kg) at the waist of the body and can generate adequate amount of pressurized air (maximum pressure of 1050 kPa and a flow rate of 15.1 mL/sec at 550 kPa) to power a unilateral active AFO used to assist the dorsiflexion (DF) motion of drop-foot patients. The finalized system can provide a maximum assistive torque of 10 Nm and induces an average 0.03±0.06 Nm resistive torque when free movement is provided. The system was tested for two unilateral drop-foot patients. The proposed system showed an average improvement of 13.6° of peak dorsiflexion angle during the swing phase of the gait cycle.

Multi-legged robots are effective at traversing rough terrain. However, terrains that include collapsible footholds (i.e. regions that can collapse when stepped on) remain a significant challenge, especially since such situations can be extremely difficult to anticipate using only exteroceptive sensing. State-of-the-art methods typically use various stabilisation techniques to regain balance and counter changing footholds. However, these methods are likely to fail if safe footholds are sparse and spread out or if the robot does not respond quickly enough after a foothold collapse. This paper presents a novel method for multi-legged robots to probe and test the terrain for collapses using its legs while walking. The proposed method improves on existing terrain probing approaches, and integrates the probing action into a walking cycle. A follow-the-leader strategy with a suitable gait and stance is presented and implemented on a hexapod robot. The proposed method is experimentally validated, demonstrating the robot can safely traverse terrain containing collapsible footholds.

In this paper, we address a 3D reconstruction problem using depth prediction from a single RGB image. With the recent advances in deep learning, depth prediction shows high performance. However, due to the discrepancy between training environment and test environment, 3D reconstruction can be vulnerable to the uncertainty of depth prediction. To consider the uncertainty of depth prediction for robust 3D reconstruction, we adopt Bayesian deep learning framework. Conventional Bayesian deep learning requires a large amount of time and GPU memory to perform Monte Carlo sampling. To address this problem, we propose a lightweight Bayesian neural network consisting of U-net structure and summation-based skip connections, which is performed in real-time. Estimated uncertainty is utilized in probabilistic TSDF fusion for dense 3D reconstruction by maximizing the posterior of TSDF value per voxel. As a result, global TSDF robust to erroneous depth values can be obtained and then dense 3D reconstruction from the global TSDF is achievable more accurately. To evaluate the performance of depth prediction and 3D reconstruction using our method, we utilized two official datasets and demonstrated the outperformance of the proposed method over other conventional methods.

Nowadays, robots are mechanically able to perform highly demanding tasks, where AI-based planning methods are used to schedule a sequence of actions that result in the desired effect. However, it is not always possible to know the exact outcome of an action in advance, as failure situations may occur at any time. To enhance failure tolerance, we propose to predict the effects of robot actions by augmenting collected experience with semantic knowledge and leveraging realistic physics simulations. That is, we consider semantic similarity of actions in order to predict outcome probabilities for previously unknown tasks. Furthermore, physical simulation is used to gather simulated experience that makes the approach robust even in extreme cases. We show how this concept is used to predict action success probabilities and how this information can be exploited throughout future planning trials. The concept is evaluated in a series of real world experiments conducted with the humanoid robot Rollin' Justin.

Modern robotic systems sense the environment geometrically, through sensors like cameras, lidar, and sonar, as well as semantically, often through visual models learned from data, such as object detectors. We aim to develop robots that can use all of these sources of information for reliable navigation, but each is corrupted by noise. Rather than assume that object detection will eventually achieve near perfect performance across the lifetime of a robot, in this work we represent and cope with the semantic and geometric uncertainty inherent in object detection methods. Specifically, we model data association ambiguity, which is typically non-Gaussian, in a way that is amenable to solution within the common nonlinear Gaussian formulation of simultaneous localization and mapping (SLAM). We do so by eliminating data association variables from the inference process through max-marginalization, preserving standard Gaussian posterior assumptions. The result is a max-mixture-type model that accounts for multiple data association hypotheses. We provide experimental results on indoor and outdoor semantic navigation tasks with noisy odometry and object detection and find that the ability of the proposed approach to represent multiple hypotheses, including the "null" hypothesis, gives substantial robustness advantages in comparison to alternative semantic SLAM approaches.