IROS 2018 论文列表

In this work, we propose a purely geometrical approach for the robust matching of line segments for challenging stereo streams with severe illumination changes or High Dynamic Range (HDR) environments. To that purpose, we exploit the univocal nature of the matching problem, i.e. every observation must be corresponded with a single feature or not corresponded at all. We state the problem as a sparse, convex, ℓ1-minimization of the matching vector regularized by the geometric constraints. This formulation allows for the robust tracking of line segments along sequences where traditional appearance-based matching techniques tend to fail due to dynamic changes in illumination conditions. Moreover, the proposed matching algorithm also results in a considerable speed-up of previous state of the art techniques making it suitable for real-time applications such as Visual Odometry (VO). This, of course, comes at expense of a slightly lower number of matches in comparison with appearance-based methods, and also limits its application to continuous video sequences, as it is rather constrained to small pose increments between consecutive frames. We validate the claimed advantages by first evaluating the matching performance in challenging video sequences, and then testing the method in a benchmarked point and line based VO algorithm.

This paper deals with the optimization of the geometry of a Cable-Driven Parallel Robot (CDPR) dedicated to large-scale construction applications. Since the maximum cable tension is a critical parameter in the design of the CDPR components, the geometry of the CDPR is optimized by minimizing the lowest maximum cable tension that ensures the validity of wrench-feasibility constraints. The geometric design procedure used in this paper consists of two phases, the CDPR cable connections is selected in the first phase followed by a second phase where the geometric parameters are optimized. The result of this procedure is an original fully-constrained CDPR geometry.

This work presents a methodology for modeling and predicting human behavior in settings with N humans interacting in highly multimodal scenarios (i.e. where there are many possible highly-distinct futures). A motivating example includes robots interacting with humans in crowded environments, such as self-driving cars operating alongside human-driven vehicles or human-robot collaborative bin packing in a warehouse. Our approach to model human behavior in such uncertain environments is to model humans in the scene as nodes in a graphical model, with edges encoding relationships between them. For each human, we learn a multimodal probability distribution over future actions from a dataset of multi-human interactions. Learning such distributions is made possible by recent advances in the theory of conditional variational autoencoders and deep learning approximations of probabilistic graphical models. Specifically, we learn action distributions conditioned on interaction history, neighboring human behavior, and candidate future agent behavior in order to take into account response dynamics. We demonstrate the performance of such a modeling approach in modeling basketball player trajectories, a highly multimodal, multi-human scenario which serves as a proxy for many robotic applications.

Conventional deep learning classifiers are static in the sense that they are trained on a predefined set of classes and learning to classify a novel class typically requires re-training. In this work, we address the problem of Low-Shot network-expansion learning. We introduce a learning framework which enables expanding a pre-trained (base) deep network to classify novel classes when the number of examples for the novel classes is particularly small. We present a simple yet powerful hard distillation method where the base network is augmented with additional weights to classify the novel classes, while keeping the weights of the base network unchanged. We show that since only a small number of weights needs to be trained, the hard distillation excels in low-shot training scenarios. Furthermore, hard distillation avoids detriment to classification performance on the base classes. Finally, we show that low-shot network expansion can be done with a very small memory footprint by using a compact generative model of the base classes training data with only a negligible degradation relative to learning with the full training set.

We propose a scalable, parallelizable policy synthesis framework intended for a robot presented with the decision of exploration or rescue, given some time-varying, stochastic mission conditions, referred to as context. We demonstrate the feasibility of such a solution using physics-based simulations to synthesize a policy in a computationally-efficient manner and exhibit superior performance with regards to the minimization of probability of mission failure when compared to two feasible baseline approaches. Furthermore, we present preliminary results that suggest our approach is robust to errors in the state estimation used to build mission context, which further supports the notion of real-world applicability.

This paper describes a method of generating attending behaviors adaptively to the user state. The method classifies the user state based on user information such as the relative position and the orientation. For each classified state, the method executes the corresponding policy for behavior generation, which has been trained using a deep reinforcement learning, namely DDPG (deep deterministic policy gradient). We use as a state space of DDPG a distance-transformed local map with person information, and define reward functions suitable for respective user states. We conducted attending experiments both in a simulated and a real environment to show the effectiveness of the proposed method.

This work describes the development of a high-resolution tactile-sensing finger for robot grasping. This finger, inspired by previous GelSight sensing techniques (Johnson and Adelson 2009), features an integration that is slimmer, more robust, and with more homogeneous output than previous vision-based tactile sensors. To achieve a compact integration, we redesign the optical path from illumination source to camera by combining light guides and an arrangement of mirror reflections. We parameterize the optical path with geometric design variables and describe the tradeoffs between the finger thickness, camera depth of field, and size of the tactile sensing area. The sensor sustains the wear from continuous use - and abuse - in grasping tasks by combining tougher materials for the compliant gel, a textured fabric skin, a structurally rigid body, and a calibration process that maintains homogeneous illumination and contrast of the tactile images during use. Finally, we evaluate the sensor's durability along four metrics that track the signal quality during more than 3000 grasping experiments.

Adapting robot actions to human motions can make human-robot interactions (HRI) more effective. Here, we aim to optimize guidance from haptic devices based on a user's response to produce better task performance. We used Gaussian processes to model the motions a human user made in response to applied torques from an ungrounded control moment gyroscope haptic device. We then used Gaussian process dynamic programming to generate optimized haptic cues to guide the user to rotate the device toward 3D targets. We compared the performance of naive and optimized policies in simulations and with a human user, and found that dynamic programming can significantly improve haptic guidance in cases where human responses are highly variable or inconsistent with the cued haptic direction.

Self-driving vehicles are being increasingly deployed in the wild. One of the most important next hurdles for autonomous driving is how such vehicles will optimally interact with one another and with their surroundings. In this paper, we consider the lane changing problem that is fundamental to road-bound multi-vehicle systems, and approach it through a combination of deep reinforcement learning (DRL) and game theory. We introduce a proactive-passive lane changing framework and formulate the lane changing problem as a Markov game between the proactive and passive vehicles. Based on different approaches to carry out DRL to solve the Markov game, we propose an asynchronous lane changing scheme as in a single-agent RL setting and a synchronous cooperative lane changing scheme that takes into consideration the adaptive behavior of the other vehicle in a vehicle's decision. Experimental results show that the synchronous scheme can effectively create and find proper merging moment after sufficient training. The framework and solution developed here demonstrate the potential of using reinforcement learning to solve multi-agent autonomous vehicle tasks such as the lane changing as they are formulated as Markov games.

Soft microrobots based on photoresponsive materials and controlled by light fields can generate a variety of different gaits. This inherent flexibility can be exploited to maximize their locomotion performance in a given environment and used to adapt them to changing conditions. Albeit, because of the lack of accurate locomotion models, and given the intrinsic variability among microrobots, analytical control design is not possible. Common data-driven approaches, on the other hand, require running prohibitive numbers of experiments and lead to very sample-specific results. Here we propose a probabilistic learning approach for light-controlled soft microrobots based on Bayesian Optimization (BO) and Gaussian Processes (GPs). The proposed approach results in a learning scheme that is data-efficient, enabling gait optimization with a limited experimental budget, and robust against differences among microrobot samples. These features are obtained by designing the learning scheme through the comparison of different GP priors and BO settings on a semi-synthetic data set. The developed learning scheme is validated in microrobot experiments, resulting in a 115% improvement in a microrobot's locomotion performance with an experimental budget of only 20 tests. These encouraging results lead the way toward self-adaptive microrobotic systems based on light-controlled soft microrobots and probabilistic learning control.

Next-best-view algorithms are commonly used for covering known scenes, for example in search, maintenance, and mapping tasks. In this paper, we consider the problem of planning a strategy for covering articulated environments where the robot also has to manipulate objects to inspect obstructed areas. This problem is particularly challenging due to the many degrees of freedom resulting from the articulation. We propose to exploit graphics processing units present in many embedded devices to parallelize the computations of a greedy next-best-view approach. We implemented algorithms for costmap computation, path planning, as well as simulation and evaluation of viewpoint candidates in OpenGL for Embedded Systems and benchmarked the implementations on multiple device classes ranging from smartphones to multi-GPU servers. We introduce a heuristic for estimating a utility map from images rendered with strategically placed spherical cameras and show in simulation experiments that robots can successfully explore complex articulated scenes with our system.

We present semi-supervised deep learning approaches for traversability estimation from fisheye images. Our method, GONet, and the proposed extensions leverage Generative Adversarial Networks (GANs) to effectively predict whether the area seen in the input image(s) is safe for a robot to traverse. These methods are trained with many positive images of traversable places, but just a small set of negative images depicting blocked and unsafe areas. This makes the proposed methods practical. Positive examples can be collected easily by simply operating a robot through traversable spaces, while obtaining negative examples is time consuming, costly, and potentially dangerous. Through extensive experiments and several demonstrations, we show that the proposed traversability estimation approaches are robust and can generalize to unseen scenarios. Further, we demonstrate that our methods are memory efficient and fast, allowing for real-time operation on a mobile robot with single or stereo fisheye cameras. As part of our contributions, we open-source two new datasets for traversability estimation. These datasets are composed of approximately 24h of videos from more than 25 indoor environments. Our methods outperform baseline approaches for traversability estimation on these new datasets.

Nascent endoscopic therapeutic procedures, such as endoscopic submucosal dissection, enable unparalleled access to and removal of mid-size cancerous neoplasia from within the gastrointestinal tract. However, the remote locations of these lesions often incur substantial distal dexterity which imparts appreciable cognitive loading on the clinician and opens up the possibility of adverse events such as intestinal perforation due to limited dexterity and a lack of sensory feedback. In this work, we introduce a mm-scale, tip-mounted robotic system, EndoMODRA (Endoscopic Module for On-Demand Robotic Assistance), which interfaces with commercially-available endoscopic tools and provides additional dexterity and feedback sensing using on-board actuators and sensors, decoupling tool motion from endoscope motion. Leveraging alternative high-energy-density actuation strategies and monolithic, printed-circuit-inspired manufacturing processes, all actuation and sensing is fully contained within the distally-mounted module, obviating the need for a continuous mechanical transmission to a proximal motor package. We develop a fuzzy-tuned PID/PWM controller for closing the loop distally to enable closed-loop position-controlled trajectory execution using onboard actuation and sensing, realizing fully -distal loop closure in an endoscope-mounted robotic module with no proximal actuation or sensing component. Controller performance is validated on a fully-integrated module with on-board sensing, demonstrating the ability to execute pre-determined trajectories as well as real-time rate-based teleoperation.

Just like the well-established Euler angles representation, fused angles are a convenient parameterisation for rotations in three-dimensional Euclidean space. They were developed in the context of balancing bodies, most specifically walking bipedal robots, but have since found wider application due to their useful properties. A comparative analysis between fused angles and Euler angles is presented in this paper, delineating the specific differences between the two representations that make fused angles more suitable for representing orientations in balance-related scenarios. Aspects of comparison include the locations of the singularities, the associated parameter sensitivities, the level of mutual independence of the parameters, and the axisymmetry of the parameters.

In this paper, we present a real-time approach to predict multiple grasping poses for a parallel-plate robotic gripper using RGB images. A model with oriented anchor box mechanism is proposed and a new matching strategy is used during the training process. An end-to-end fully convolutional neural network is employed in our work. The network consists of two parts: the feature extractor and multi-grasp predictor. The feature extractor is a deep convolutional neural network. The multi-grasp predictor regresses grasp rectangles from predefined oriented rectangles, called oriented anchor boxes, and classifies the rectangles into graspable and ungraspable. On the standard Cornell Grasp Dataset, our model achieves an accuracy of 97.74% and 96.61% on image-wise split and object-wise split respectively, and outperforms the latest state-of-the-art approach by 1.74% on image-wise split and 0.51% on object-wise split.

In this work, we present an adaptive motion planning approach for impedance-controlled robots to modify their tasks based on human physical interactions. We use a class of parameterized time-independent dynamical systems for motion generation where the modulation of such parameters allows for motion flexibility. To adapt to human interactions, we update the parameters of our dynamical system in order to reduce the tracking error (i.e., between the desired trajectory generated by the dynamical system and the real trajectory influenced by the human interaction). We provide analytical analysis and several simulations of our method. Finally, we investigate our approach through real world experiments with a 7-DOF KUKA LWR 4+ robot performing tasks such as polishing and pick-and-place.

For generating interactive behavior of robot to build a long-term relationship between humans and robots, we focus on the difference in familiarity of the human behaviors during conversation. It is difficult to extract interaction motion features correlated to such familiarity as a model in manual. Therefore, we use a machine learning technique: convolution neural network to learn and generate interaction behavior with different familiarity. In the evaluation experiment, we generated interaction behavior using a convolution neural network, which learned from the behaviors of friendship and unknown relationship, who have high and low familiarity respectively. We evaluated how much such interaction behavior affect the human impression by questionnaire survey.

Patients suffering from quadriplegia have limited body motion which prevents them from performing daily activities. We have developed an assistive robotic system with an intuitive free-view gaze interface. The user's point of regard is estimated in 3D space while allowing free head movement and is combined with object recognition and trajectory planning. This framework allows the user to interact with objects using fixations. Two operational modes have been implemented to cater for different eventualities. The automatic mode performs a pre-defined task associated with a gaze-selected object, while the manual mode allows gaze control of the robot's end-effector position on the user's frame of reference. User studies reported effortless operation in automatic mode. A manual pick and place task achieved a success rate of 100% on the users' first attempt.

This research seeks greater efficiency for walking robots. Efficiency can be improved in two ways: better performance (i.e., less wasted motion) and reduced energy consumption. Fractional-order control is a pathway to both of these improvements because of the flexibility it offers in designing a control strategy. Compared to the existing proportional-derivative architecture, changing the order of the derivative - the number of derivatives taken - to real numbers other than 1 has yielded both types of improvement for a simulated walker. The evidence of better performance is the leg angles' improvement in maintaining a desired relationship with respect to one another. Depending on the controller chosen, the walker can also be made to achieve the original level of performance with reduced control signals and less torque delivered to the hip joint, implying greater energy efficiency.