ICRA 2020 论文列表

Batch optimization based inertial measurement unit (IMU) and visual sensor fusion enables high rate localization for many robotic tasks. However, it remains a challenge to ensure that the batch optimization is computationally efficient while being consistent for high rate IMU measurements without marginalization. In this paper, we derive inspiration from maximum likelihood estimation with partial-fixed estimates to provide a unified approach for handing both IMU preintegration and time-offset calibration. We present a modularized analytic combined IMU integrator (ACI2) with elegant derivations for IMU integrations, bias Jabcobians and related covariances. To simplify our derivation, we also prove that the right Jacobians for Hamilton quaterions and SO(3) are equivalent. Finally, we present a time offset calibrator that operates by fixing the linearization point for a given time offset. This reduces re-integration of the IMU measurements and thus improve efficiency. The proposed ACI2 and time-offset calibration is verified by intensive Monte-Carlo simulations generated from real world datasets. A proof-of-concept real world experiment is also conducted to verify the proposed ACI2 estimator.

For safe navigation around pedestrians, automated vehicles (AVs) need to plan their motion by accurately predicting pedestrians' trajectories over long time horizons. Current approaches to AV motion planning around crosswalks predict only for short time horizons (1-2 s) and are based on data from pedestrian interactions with human-driven vehicles (HDVs). In this paper, we develop a hybrid systems model that uses pedestrians' gap acceptance behavior and constant velocity dynamics for long-term pedestrian trajectory prediction when interacting with AVs. Results demonstrate the applicability of the model for long-term (> 5 s) pedestrian trajectory prediction at crosswalks. Further, we compared measures of pedestrian crossing behaviors in the immersive virtual environment (when interacting with AVs) to that in the real world (results of published studies of pedestrians interacting with HDVs), and found similarities between the two. These similarities demonstrate the applicability of the hybrid model of AV interactions developed from an immersive virtual environment (IVE) for real-world scenarios for both AVs and HDVs.

In manufacturing, kitting is the process of grouping separate items together to be supplied as one unit to the assembly line. This is a key logistic task, which is usually performed manually by human operators. However, picking objects from the warehouse implies a great repetitiveness in arm motion. Moreover, the weight and position of items may increase the physical strain and induce the development of work-related musculoskeletal disorders. The inclusion of a collaborative robot in the process may help to reduce the operator's effort and increase productivity. This paper introduces an online scheduling algorithm to guide the picking operations of the human and the robot. The proposed approach has been experimentally evaluated and compared with an offline scheduler, as well as with the baseline case of manual kitting.

The field of soft robotics is grounded on the idea that, due to their inherent compliance, soft robots can safely interact with the environment. Thus, the development of effective planning and control pipelines for soft robots should incorporate reliable robot-environment interaction models. This strategy enables soft robots to effectively exploit contacts to autonomously navigate and accomplish tasks in the environment. However, for a class of soft robots, namely vine-inspired, tip-extending or "vine" robots, such interaction models and the resulting planning and control strategies do not exist. In this paper, we analyze the behavior of vine robots interacting with their environment and propose an obstacle-interaction model that characterizes the bending and wrinkling deformation induced by the environment. Starting from this, we devise a novel obstacle-interaction planning method for these robots. We show how obstacle interactions can be effectively leveraged to enlarge the set of reachable workspace for the robot tip, and verify our findings with both simulated and real experiments. Our work improves the capabilities of this new class of soft robot, helping to advance the field of soft robotics.

In this paper, we propose an efficient solution to the calibrated camera motion estimation with a common direction. This case is relevant to smart phones, tablets, and other camera-IMU (Inertial measurement unit) systems, which have accelerometers to measure the gravity direction. We can align one of the axes of the camera with this common direction so that the relative rotation between the views reduces to only 1DOF (degree of freedom). This allows us to use only three point correspondences for relative pose estimation. Unlike previous work, we derive new constraints on the simplified essential matrix using an elimination strategy based on Gröbner basis. In this case, computing the coefficients of these constraints require less computation and we only need to solve a polynomial eigenvalue problem. We show detailed analyses and comparisons against the existing 3-point algorithms, with satisfactory results obtained.

Online estimation of 3D ground reaction forces (GRFs) is becoming increasingly important for closed-loop control of lower-extremity robotic exoskeletons. Through in-verse dynamics and optimization models, 3D GRFs can be used to estimate net joint torques and approximate muscle forces. Although instrumented footwear to measure vertical GRFs in out-of-the-lab environments is available, accurately measuring shear GRFs with foot-mounted sensors still remains a challenging task. In this paper, a new outsole-embedded optoelectronic sensor configuration that is able to measure biaxial shear GRFs is proposed. Compared with traditional strain-gauge based solutions, optoelectronic sensors allow for a more affordable design. To mitigate the risk of altering the wearer's natural gait, the proposed solution does not involve external modifications to the footwear structure. A preliminary validation of the outsole-embedded sensor was conducted against validated laboratory equipment. The test involved two sessions of treadmill walking at different speeds. Experimental results suggest that the proposed design may be a promising solution for measuring shear GRFs in unconstrained environments.

We present an open-source framework that provides a low barrier to entry for real-time simulation, visualization, and interactive manipulation of user-specifiable soft-bodies, environments, and robots (using a human-readable front-end interface). The simulated soft-bodies can be interacted by a variety of input interface devices including commercially available haptic devices, game controllers, and the Master Tele-Manipulators (MTMs) of the da Vinci Research Kit (dVRK) with real-time haptic feedback. We propose this framework for carrying out multi-user training, user-studies, and improving the control strategies for manipulation problems. In this paper, we present the associated challenges to the development of such a framework and our proposed solutions. We also demonstrate the performance of this framework with examples of soft-body manipulation and interaction with various input devices.

Even though mobile robots have been around for decades, trajectory optimization and continuous time collision avoidance remain subject of active research. Existing methods trade off between path quality, computational complexity, and kinodynamic feasibility. This work approaches the problem using a nonlinear model predictive control (NMPC) framework, that is based on a novel convex inner approximation of the collision avoidance constraint. The proposed Convex Inner ApprOximation (CIAO) method finds kinodynamically feasible and continuous time collision free trajectories, in few iterations, typically one. For a feasible initialization, the approach is guaranteed to find a feasible solution, i.e. it preserves feasibility. Our experimental evaluation shows that CIAO outperforms state of the art baselines in terms of planning efficiency and path quality. Experiments show that it also efficiently scales to high-dimensional systems. Furthermore real-world experiments demonstrate its capability of unifying trajectory optimization and tracking for safe motion planning in dynamic environments.

Iterative linear-quadratic (ILQ) methods are widely used in the nonlinear optimal control community. Recent work has applied similar methodology in the setting of multi-player general-sum differential games. Here, ILQ methods are capable of finding local equilibria in interactive motion planning problems in real-time. As in most iterative procedures, however, this approach can be sensitive to initial conditions and hyperparameter choices, which can result in poor computational performance or even unsafe trajectories. In this paper, we focus our attention on a broad class of dynamical systems which are feedback linearizable, and exploit this structure to improve both algorithmic reliability and runtime. We showcase our new algorithm in three distinct traffic scenarios, and observe that in practice our method converges significantly more often and more quickly than was possible without exploiting the feedback linearizable structure.

This study proposes an intelligent spraying system with semantic segmentation of fruit trees in a pear orchard. A fruit tree detection system was developed using the SegNet model, a semantic segmentation structure. The system is trained with images categorized into five distinct classes. The learned deep learning model performed with an accuracy of 83.79%. Further, we fusion depth data from an RGB-D camera to prevent the tree in the background from being detected. To operate the nozzles, each image captured from the camera is separated lengthwise into quarters and mapped to the nozzles. Then, the nozzle was opened when the area of fruit trees in each zone exceeded 20%. Two types of field experiments were performed in a pear orchard to verify the effectiveness of our system. From the results obtained, we can confirm the satisfactory performance of our deep learning-based intelligent spraying system. It is expected that the introduction of this system to actual farms will signicantly reduce the amount of pesticide used and will make the work environment safer for farmers.

Non-linear dynamical systems represent a compact, flexible, and robust tool for reactive motion generation. The effectiveness of dynamical systems relies on their ability to accurately represent stable motions. Several approaches have been proposed to learn stable and accurate motions from demonstration. Some approaches work by separating accuracy and stability into two learning problems, which increases the number of open parameters and the overall training time. Alternative solutions exploit single-step learning but restrict the applicability to one regression technique. This paper presents a single-step approach to learn stable and accurate motions that work with any regression technique. The approach makes energy considerations on the learned dynamics to stabilize the system at run-time while introducing small deviations from the demonstrated motion. Since the initial value of the energy injected into the system affects the reproduction accuracy, it is estimated from training data using an efficient procedure. Experiments on a real robot and a comparison on a public benchmark shows the effectiveness of the proposed approach.

Industrial arm-type robots have multiple degrees-of-freedom (DoFs) and high dexterity but the use of the roll-pitch-roll wrist configuration yields singularities inside the reachable workspace. Excessive joint velocities will occur when encountering these singularities. Arm-type robots currently don't have enough dexterity to move the end-effector path away from the wrist singularities. Robots with redundant DoFs can be used to provide additional dexterity to avoid the singularities and reduce the excessive joint velocity. An end-effector wrist module is proposed to provide two redundant DoFs when interfaced with an existing 6-DoF robot. The new 8-DoF robot has a compact roll-pitch-yaw wrist that has no singularities inside the reachable workspace. The highly redundant robot can also be used to avoid collisions in various directions. Path tracking simulation examples are provided to show the advantages of the proposed design when compared with existing redundant or nonredundant robots. We expect that this module can serve as a cost-effective solution in applications where singularity-free motion or collision-free motion is required.

Exploration of unknown environments is embedded and essential in many robotics applications. Traditional algorithms, that decide where to explore by computing the expected information gain of an incomplete map from future sensor measurements, are limited to very powerful computational platforms. In this paper, we describe a novel approach for computing this expected information gain efficiently, as principally derived via mutual information. The key idea behind the proposed approach is a continuous occupancy map framework and the recursive structure it reveals. This structure makes it possible to compute the expected information gain of sensor measurements across an entire map much faster than computing each measurements’ expected gain independently. Specifically, for an occupancy map composed of |M| cells and a range sensor that emits |Θ| measurement beams, the algorithm (titled FCMI) computes the information gain corresponding to measurements made at each cell in O(|Θ||M|) steps. To the best of our knowledge, this complexity bound is better than all existing methods for computing information gain. In our experiments, we observe that this novel, continuous approach is two orders of magnitude faster than the state-of-the-art FSMI algorithm.

Dormant pruning is a major cost component of fresh market tree fruit production, nearly equal in scale to harvesting the fruit. However, relatively little focus has been given to the problem of pruning trees autonomously. In this paper, we introduce a robotic system consisting of an industrial manipulator, an eye-in-hand RGB-D camera configuration, and a custom pneumatic cutter. The system is capable of planning and executing a sequence of cuts while making minimal assumptions about the environment. We leverage a novel planning framework designed for high-throughput operation which builds upon previous work to reduce motion planning time and sequence cut points intelligently. In end-to-end experiments with a set of ten different branch configurations, the system achieved a high success rate in plan execution and a 1.5x speedup in throughput versus a baseline planner, representing a significant step towards the goal of practical implementation of robotic pruning.

Electrical resistance tomography (ERT) has previously been utilized to develop a large-scale tactile sensor because this approach enables the estimation of the conductivity distribution among the electrodes based on a known physical model. Such a sensor made with a stretchable material can conform to a curved surface. However, this sensor cannot fully cover a cylindrical surface because in such a configuration, the edges of the sensor must meet each other. The electrode configuration becomes irregular in this edge region, which may degrade the sensor performance. In this paper, we introduce an ERT-based robotic skin with evenly and sparsely distributed electrodes. For implementation, we sprayed a carbon nanotube (CNT)-dispersed solution to form a conductive sensing domain on a cylindrical surface. The electrodes were firmly embedded in the surface so that the wires were not exposed to the outside. The sensor output images were estimated using a deep neural network (DNN), which was trained with noisy simulation data. An indentation experiment revealed that the localization error of the sensor was 5.2 ± 3.3 mm, which is remarkable performance with only 30 electrodes. A frame rate of up to 120 Hz could be achieved with a sensing domain area of 90 cm2. The proposed approach simplifies the fabrication of 3D-shaped sensors, allowing them to be easily applied to existing robot arms in a seamless and robust manner.

For most people on the ground, facing an unwanted drone buzzing around overhead, there is not a lot that we can do, especially if it is out of gun (radio wave gun or shotgun) range. A solution to this is to use intercept drones that seek out and bring down other drones. In order to make the interception autonomous, an image-based visual servo algorithm is designed with a forward-looking monocular camera. The control command, namely the angular velocity and thrust, is generated for intercept drones to implement accurate and fast interception. The proposed method is demonstrated in both hardware-in-the-loop simulation and demonstrative flight experiments.

This paper presents an adaptive control scheme for robotic systems that operate in the face of-potentially large-structured uncertainty. The proposed adaptive controller employs an on-line supervisor that utilizes logic-based switching among a finite set of controllers to identify uncertain parameters, and adapt the behavior of the system based on a current estimate of their value. To achieve this, the adaptive control approach in this paper combines on-line parameter estimation and feedback control while avoiding some of the inherent difficulties of classical adaptive control strategies. Furthermore, the proposed supervisory control architecture is modular as it relies on established "off-the-shelf" feedback control law and estimator design approaches, instead of cus-tomizing the overall design to the specific requirements of an adaptive control algorithm. We demonstrate the efficacy of the method on the problem of a dynamically-walking bipedal robot delivering a payload of unknown mass, and show that, by switching to the controller that is the "best" according to a current estimate of the uncertainty, the system maintains a low energy cost during its operation.

This paper presents a novel approach to fault tolerant reconfiguration of modular self-folding robots. Among various types of faults that probably occur in the modular system, we focus on the tolerance of complete actuation failure of active modules that might cause imprecise robotic motion and even reconfiguration failure. Our approach is to utilize the reconfigurability of modular self-folding robots and investigate intra-module connection to determine initial patterns that are inherently fault tolerant. We exploit the redundancy of actuation and distribute active modules in both layout-based and target-based scenarios, such that reconfiguration schemes with user-specified fault tolerant capability can be generated for an arbitrary input initial pattern or 3D configuration. Our methods are demonstrated in computer-aided simulation on the robotic platform of Mori, a modular origami robot. The simulation results validate that the proposed algorithms yield fault tolerant initial patterns and distribution schemes of active modules for several 2D and 3D configurations with Mori, while retaining generalizability for a large number of modular self-folding robots.

In human-robot collaboration, it is crucial for the robot to make its intentions clear and predictable to the human partners. Inspired by the mutual learning and adaptation of human partners, we suggest an actor-critic approach for a legible robot motion planner. This approach includes two neural networks and a legibility evaluator: 1) A policy network based on deep reinforcement learning (DRL); 2) A Recurrent Neural Networks (RNNs) based sequence to sequence (Seq2Seq) model as a motion predictor; 3) A legibility evaluator that maps motion to legible reward. Through a series of human-subject experiments, we demonstrate that with a simple handicraft function and no real-human data, our method lead to improved collaborative performance against a baseline method and a non-prediction method.