ICRA 2020 论文列表

We estimate the global pose of a multirotor UAV by visually localizing images captured during a flight with Google Earth images pre-rendered from known poses. We metrically localize real images with georeferenced rendered images using a dense mutual information technique to allow accurate global pose estimation in outdoor GPS-denied environments. We show the ability to consistently localize throughout a sunny summer day despite major lighting changes while demonstrating that a typical feature-based localizer struggles under the same conditions. Successful image registrations are used as measurements in a filtering framework to apply corrections to the pose estimated by a gimballed visual odometry pipeline. We achieve less than 1 m and 1° RMSE on a 303 m flight and less than 3 m and 3° RMSE on six 1132 m flights as low as 36 m above ground level conducted at different times of the day from sunrise to sunset.

We study the problem of learning manipulation skills from human demonstration video by inferring the association relationships between geometric features. Motivation for this work stems from the observation that humans perform eye-hand coordination tasks by using geometric primitives to define a task while a geometric control error drives the task through execution. We propose a graph based kernel regression method to directly infer the underlying association constraints from human demonstration video using Incremental Maximum Entropy Inverse Reinforcement Learning (InMaxEnt IRL). The learned skill inference provides human readable task definition and outputs control errors that can be directly plugged into traditional controllers. Our method removes the need for tedious feature selection and robust feature trackers required in traditional approaches (e.g. feature-based visual ser-voing). Experiments show our method infers correct geometric associations even with only one human demonstration video and can generalize well under variance.

This paper presents a coverage control algorithm for teams of quadcopters with downward facing visual sensors that prevents the appearance of coverage holes in-between the monitored areas while maximizing the coverage quality as much as possible. We derive necessary and sufficient conditions for preventing the appearance of holes in-between the fields of views among trios of robots. Because this condition can be expressed as logically combined constraints, control nonsmooth barrier functions are implemented to enforce it. An algorithm which extends control nonsmooth barrier functions to hybrid systems is implemented to manage the switching among barrier functions caused by the changes of the robots composing trio. The performance and validity of the proposed algorithm are evaluated in simulation as well as on a team of quadcopters.

We address the localization of robots in a multi-MAV system where external infrastructure like GPS or motion capture systems may not be available. Our approach lends itself to implementation on platforms with several constraints on size, weight, and power (SWaP). Particularly, our framework fuses the onboard VIO with the anonymous, visual-based robot-to-robot detection to estimate all robot poses in one common frame, addressing three main challenges: 1) the initial configuration of the robot team is unknown, 2) the data association between each vision-based detection and robot targets is unknown, and 3) the vision-based detection yields false negatives, false positives, inaccurate, and provides noisy bearing, distance measurements of other robots. Our approach extends the Coupled Probabilistic Data Association Filter [1] to cope with nonlinear measurements. We demonstrate the superior performance of our approach over a simple VIO-based method in a simulation with the measurement models statistically modeled using the real experimental data. We also show how onboard sensing, estimation, and control can be used for formation flight.

Legged robots have been highlighted as promising mobile platforms for disaster response and rescue scenarios because of their rough terrain locomotion capability. In cluttered environments, small robots are desirable as they can maneuver through small gaps, narrow paths, or tunnels. However small robots have their own set of difficulties such as limited space for sensors, limited obstacle clearance, and scaled-down walking speed. In this paper, we extensively address these difficulties via effective sensor integration and exploitation of dynamic locomotion and jumping. We integrate two Intel RealSense sensors into the MIT Mini-Cheetah, a 0.3 m tall, 9 kg quadruped robot. Simple and effective filtering and evaluation algorithms are used for foothold adjustment and obstacle avoidance. We showcase the exploration of highly irregular terrain using dynamic trotting and jumping with the small-scale, fully sensorized Mini-Cheetah quadruped robot.

Recent work on semantic simultaneous localization and mapping (SLAM) have shown the utility of natural objects as landmarks for improving localization accuracy and robustness. In this paper we present a monocular semantic SLAM system that uses object identity and inter-object geometry for view-invariant loop detection and drift correction. Our system's ability to recognize an area of the scene even under large changes in viewing direction allows it to surpass the mapping accuracy of ORB-SLAM, which uses only local appearance-based features that are not robust to large viewpoint changes. Experiments on real indoor scenes show that our method achieves mean drift reduction of 70% when compared directly to ORB-SLAM. Additionally, we propose a method for object orientation estimation, where we leverage the tracked pose of a moving camera under the SLAM setting to overcome ambiguities caused by object symmetry. This allows our SLAM system to produce geometrically detailed semantic maps with object orientation, translation, and scale.

Car models have been extensively studied at varying levels of abstraction, and planning and executing motions under ideal conditions is well researched and understood. For more aggressive maneuvers, for example when drifting or skidding, empirical and/or discontinuous friction models have been used to explain and approximate real world contact behavior. Separately, contact dynamics have been extensively studied by the robotics community, often times formulated as a linear complementarity problem (LCP) for dynamic multi-rigid-body contact problems with Coulomb friction cone approximations. In this work, we explore the validity of using such an anisotropic Coulomb friction cone to model tire dynamics to plan for vehicle motion, and present a versatile trajectory optimization framework using this model that can both avoid and/or exploit wheel skidding, depending on the cost function and planning horizon. Experimental evidence of planning and executing dynamic drift parking is shown on a 1/16 scale model car.

There are limitations of conventional active-assistive control for upper limb rehabilitation exoskeleton robot, such as 1). prior time-dependent trajectories are generally required, 2). task-based rehabilitation exercise involving multi-joint motion is hard to implement, and 3). assistive mechanism normally is so inflexible that the resulting exercise performed by the subjects becomes inefficient. In this paper, we propose a novel velocity field based active-assistive control system to address these issues. First, we design a Kalman filter based interactive torque observer to obtain subjects' active intention of motion. Next, a joint-position-dependent velocity field which can be automatically generated via the task motion pattern is proposed to provide the time-independent assistance to the subjects. We further propose a novel integration method that combines the active and assistive motions based on the performance and the involvement of subjects to guide them to perform the task more voluntarily and precisely. The experiment results show that both the execution time and the subjects' torque exertion are reduced while performing both given single joint tasks and task-oriented multi-joint tasks as compared with the related work in the literature. To sum up, the proposed system not only can efficiently retain subjects' active intention but also can assist them to accomplish the rehabilitation task more precisely.

Accurate and reliable localization is crucial to autonomous vehicle navigation and driver assistance systems. This paper presents a novel approach for online vehicle localization in a digital map. Two distinct map matching algorithms are proposed: i) Iterative Closest Point (ICP) based lane level map matching is performed with visual lane tracker and grid map ii) decision-rule based approach is used to perform topological map matching. Results of both the map matching algorithms are fused together with GPS and dead reckoning using Extended Kalman Filter to estimate vehicle's pose relative to the map. The proposed approach has been validated on real life conditions on an equipped vehicle. Detailed analysis of the experimental results show improved localization using the two aforementioned map matching algorithms.

We propose to formulate the problem of representing a distribution of robot configurations (e.g. joint angles) as that of approximating a product of experts. Our approach uses variational inference, a popular method in Bayesian computation, which has several practical advantages over sampling-based techniques. To be able to represent complex and multimodal distributions of configurations, mixture models are used as approximate distribution. We show that the problem of approximating a distribution of robot configurations while satisfying multiple objectives arises in a wide range of problems in robotics, for which the properties of the proposed approach have relevant consequences. Several applications are discussed, including learning objectives from demonstration, planning, and warm-starting inverse kinematics problems. Simulated experiments are presented with a 7-DoF Panda arm and a 28-DoF Talos humanoid.

General-purpose simulators can be a valuable data source for flexible learning and control approaches. However, training models or control policies in simulation and then directly applying to hardware can yield brittle control. Instead, we propose a novel way to use simulators as regularizers. Our approach regularizes a decoder of a variational autoencoder to a black-box simulation, with the latent space bound to a subset of simulator parameters. This enables successful encoder training from a small number of real-world trajectories (10 in our experiments), yielding a latent space with simulation parameter distribution that matches the real-world setting. We use a learnable mixture for the latent prior/posterior, which implies a highly flexible class of densities for the posterior fit. Our approach is scalable and does not require restrictive distributional assumptions. We demonstrate ability to recover matching parameter distributions on a range of benchmarks, challenging custom simulation environments and several real-world scenarios. Our experiments using ABB YuMi robot hardware show ability to help reinforcement learning approaches overcome cases of severe sim-to-real mismatch.

Theory suggests an inverse relation between the stiffness and the energy storage capacity for linear helical springs: reducing the active length of the spring by 50% increases its stiffness by 100%, but reduces its energy storage capacity by 50%. State-of-the-art variable stiffness actuators used to drive robots are characterized by a similar inverse relation, implying reduced energy storage capacity for increased spring stiffness. This relation limits the potential of the variable stiffness actuation technology when it comes to human performance augmentation in natural tasks, e.g., jumping, weight-bearing and running, which may necessitate a spring exoskeleton with large stiffness range and high energy storage capacity. In this paper, we theoretically show that the trade-off between stiffness range and energy storage capacity is not fundamental; it is possible to develop variable stiffness springs with simultaneously increasing stiffness and energy storage capacity. Consistent with the theory, we experimentally show that a controllable volume air spring, has a direct relation between its stiffness range and energy storage capacity. The mathematical conditions presented in this paper may be used to develop actuators that could bypass the limited energy storage capacity of current variable stiffness spring technology.

Human-robot interaction is one of the keys of assistive robots. Robots are expected to be compliant with people but at the same time correctly perform the tasks. In such applications, Cartesian impedance control is preferred over joint control, as the desired interaction and environmental feedback can be described more naturally, and the force to be exerted by the robot can be readily adjusted.This paper addresses the problem of controlling a robot arm in the operational space with variable stiffness so as to continuously adapt the force exerted in each phase of motion according to the precision requirements. Moreover, performing dimensionality reduction we can separate the degrees of freedom (DoF) relevant for the task from the redundant ones. The stiffness of the former can be adjusted constantly to achieve the required accuracy, while task-redundant DoF can be used to achieve other goals such as avoiding obstacles by moving in the directions where accuracy is not critical. The designed method is tested teaching the robot to give water to drink to a model of human head. Our empirical results demonstrate that the robot can learn precision requirements from demonstration. Furthermore, dimensionality reduction is proved to be useful to avoid obstacles.

This paper presents a variable damping controller to improve the trade-off between agility and stability in physical human-robot interaction (pHRI), while reducing user effort. Variable robotic damping, defined as a dual-sided logistic function, was determined in real time throughout a range of negative to positive values based on the user's intent of movement. To evaluate the effectiveness of the proposed controller, we performed a set of human experiments with subjects interacting with the end-effector of a 7 degree-of-freedom robot. Twelve subjects completed target reaching tasks under three robotic damping conditions: fixed positive, fixed negative, and variable damping. On average, the variable damping controller significantly shortened the rise time by 22.4% compared to the fixed positive damping. It is also important to note that the rise time in the variable damping condition was as fast as that in the fixed negative damping condition and there was no statistical difference between the two conditions. The variable damping controller significantly decreased the percentage overshoot by 49.6% and shortened the settling time by 29.0% compared to the fixed negative damping. Both the maximum and mean root-mean-squared (RMS) interaction forces were significantly lower in the variable damping condition than the other two fixed damping conditions, i.e., the variable damping controller reduced user effort. The maximum and mean RMS interaction forces were at least 17.3% and 20.3% lower than any of the fixed damping conditions, respectively. The results of this study demonstrate that humans can extract the benefits of the variable damping controller in the context of pHRI, as it significantly improves the trade-off between agility and stability and reduces user effort in comparison to fixed damping controllers.

Unlike robots, humans do not have constant movements. Their pathways are individually changeable and influenced by circumstances. This paper presents a method to investigate human pathway variations in a real study. In systematically selected tasks, human pathways are examined for 100 participants in manual and human-robot collaboration (HRC) scenarios. As a result, the variations of pathways are presented depending on various features: e.g. in nearly all cases the variance of women's walking pathways is smaller than that of men. VariPath database can be used in any planning process of manual or HRC scenarios to ensure safety and efficiency.

For those with irregular gait, re-calibration of motor control strategies and retraining of coordination are key goals. Thoughtful external forces or resistances during repetitive tasks can reprogram motor control patterns and strategies. Prior work in our lab has utilized this theory to improve gait in various patient groups using the Tethered Pelvic Assist Device (TPAD), a treadmill-based robotic trainer. In this paper, we propose a new, portable extension of the TPAD, which relies on an open-loop, forward kinematics based controller to remove the restriction of walking in the laboratory on a treadmill, and therefore accommodates overground ambulation. To evaluate the effects of this new control scheme and the effects of the users holding the mobile TPAD frame, a dataset of walking in four conditions was collected from eight healthy individuals. When applying a constant pelvic loading force of 10% body weight, the mean ground reaction force increased by 8.2±7.7% when the individual holds the walker frame and 11.1±7.8% when no hand contact is made. The mobile TPAD was shown to still induce a targeted loading on individuals during treadmill walking. The validation of this mobile device's controller and characterization of holding the frame allow overground studies to be conducted, and now opens the door to new training paradigms for overground gait training.

Aerial imagery plays an important role in land-use planning, population analysis, precision agriculture, and unmanned aerial vehicle tasks. However, existing aerial image datasets generally suffer from the problem of inaccurate labeling, single ground truth type, and few category numbers. In this work, we implement a simulator that can simultaneously acquire diverse visual ground truth data in the virtual environment. Based on that, we collect a comprehensive Virtual AeriaL Image Dataset named VALID, consisting of 6690 high-resolution images, all annotated with panoptic segmentation on 30 categories, object detection with oriented bounding box, and binocular depth maps, collected in 6 different virtual scenes and 5 various ambient conditions (sunny, dusk, night, snow and fog). To our knowledge, VALID is the first aerial image dataset that can provide panoptic level segmentation and complete dense depth maps. We analyze the characteristics of VALID and evaluate state-of-the-art methods for multiple tasks to provide reference baselines. The experiment results demonstrate that VALID is well presented and challenging. The dataset is available at https://sites.google.com/view/valid-dataset/.

In certain challenging environments, such as inside buildings on fire, the main sensors (e.g. cameras, LiDARs and GPS systems) used for multicopter localization can become unavailable. Direct integration of the inertial navigation sensors (the accelerometer and rate gyroscope), is however unaffected by external disturbances, but the rapid error accumulation quickly makes a naive application of such a strategy feasible only for very short durations. In this work we propose a motion strategy for reducing the inertial navigation state estimation error of multicopters. The proposed strategy breaks a long duration flight into multiple short duration hops between which the vehicle remains stationary on the ground. When the vehicle is stationary, zero-velocity pseudo-measurements are introduced to an extended Kalman Filter to reduce the state estimation error. We perform experiments for closed-loop control of a multicopter for evaluation. The mean absolute position estimation error was 3.4% over a total flight distance of 5m in the experiments. The results showed a 80% reduction compared to the standard inertial navigation method without using this strategy. In addition, an additional experiment with total flight distance of 10m is conducted to demonstrate the ability of this method to navigate a multicopter in real-world environment. The final trajectory tracking error was 3% of the total flight distance.

A segmentation-based architecture is proposed to decompose objects into multiple primitive shapes from monocular depth input for robotic manipulation. The backbone deep network is trained on synthetic data with 6 classes of primitive shapes generated by a simulation engine. Each primitive shape is designed with parametrized grasp families, permitting the pipeline to identify multiple grasp candidates per shape primitive region. The grasps are priority ordered via proposed ranking algorithm, with the first feasible one chosen for execution. On task-free grasping of individual objects, the method achieves a 94% success rate. On task-oriented grasping, it achieves a 76% success rate. Overall, the method supports the hypothesis that shape primitives can support task-free and task-relevant grasp prediction.