IROS 2018 论文列表

A common solution criterion for partially observable Markov decision processes (POMDPs) is to maximize the expected sum of exponentially discounted rewards, for which a variety of approximate methods have been proposed. Those that plan in the belief space typically provide tighter performance guarantees, but those that plan over the state space (e.g., QMDP and FIB) often require much less memory and computation. This paper presents an encouraging result that shows that reducing the discount factor while planning in the state space can actually improve performance significantly when evaluated on the original problem. This phenomenon is confirmed by both a theoretical analysis as well as a series of empirical studies on benchmark problems. As predicted by the theory and confirmed empirically, the phenomenon is most prominent when the observation model is noisy or rewards are sparse.

Many of the solutions proposed for the object manipulation problem are based on the knowledge of the object features. The approach proposed in this paper intends to provide a simple geometrical approach to securely manipulate an unknown object based only on tactile and kinematic information. The tactile and kinematic data obtained during the manipulation is used to recognize the object shape (at least the local object curvature), allowing to improve the grasping forces when this information is added to the manipulation strategy. The approach has been fully implemented and tested using the Schunk Dexterous Hand (SDH2). Experimental results are shown to illustrate the efficiency of the approach.

This paper presents a novel quadcopter design with an added momentum wheel for enhanced stability. The novel vehicle has improved torque disturbance rejection capabilities compared to a standard quadcopter. An analysis of the vehicle dynamics shows that the effect of torque disturbances decreases monotonically with increasing angular momentum of the momentum wheel. A framework for choosing the mass moment of inertia and speed of the momentum wheel is given based on an upper bound on the allowable energy stored in the wheel. Theoretical results are experimentally validated by comparing responses to torque impulses applied to the vehicle with and without the momentum wheel spinning.

Most state-of-the-art torque control-based legged robots show excellent performance, exceeding that of conventional position control-based robots. Many conventional position control-based legged robots have high gear ratios, but do not have joint torque sensors. In addition, some robots cannot generate current for controlling the motor torque. To apply torque control-based walking algorithms to a position control-based humanoid robot, we proposed current control using a motor thermal model and realized joint torque control by compensating for the joint dynamics and robot dynamics. We conducted experiments to verify the performance of the Hubo2 platform developed in 2008 by applying a full-body dynamics control framework. The results confirmed the possibility of using torque control algorithms with existing position-based robots.

This paper presents a multi-functioning light weight robotic system, the Autonomous Legged Personal Helper Robot with Enhanced Dynamics (ALPHRED), capable of both locomotion and manipulation. In addition, we extended a 2D zero moment point (ZMP) trajectory optimization (TO) algorithm to a 3D implementation. As well as adding the acceleration of the center of mass to the TO cost in order to smooth out the motion of the robot during trajectories with support polygons that do not intersect. By implementing this versatile TO algorithm on a multi-modal robotic platform we showed that many different forms of stable locomotion and manipulation were possible including a dynamic 0.7 m/s trot gait.

Deployment of robotics and remote systems for tasks in unstructured nuclear environment has been impeded by the severe task requirements such as high radiation and dexterous and complex manipulation of heavy materials, which cannot be addressed by the current telerobotics technology. To address such practical challenges, this paper presents an enhanced teleoperator interface incorporating multi-modal augmented reality, and new method of telerobotic operation based on perceptual overlay - `virtual fixtures'. Rather than trying to devise complex robotic systems, innovation is directed to enhancement of teleoperator interface so as to draw more performance and intuition from the human operator. Particular enhancements were made over the current technology basis in 3D sensing and reconstruction, virtual fixture generation, and operator interface. The telerobotic system was developed using ROS (Robot Operating System) to streamline system integration and resource sharing. The presented innovation is expected to allow deployment of simple and rugged robots to perform dexterous manipulation of heavy objects.

This paper presents a 1-link robotic arm that is antagonistically driven by one pair of a high performance super-coiled polymer actuators with an embedded controller. The actuator which is made from Spandex and nylon fibers is low-cost, easy to fabricate and light-weight. Moreover, it can generate large displacement and provide Joule heating capability. The main contribution of the paper is the model-based impedance controller, which enables position control of the antagonistic joint with variable stiffness and damping. The impedance control is a torque-based law, which in turn depends on a proposed backstepping control law to control the force of each actuator. The control system is proved to be stable using dissipativity stability theory and verified through experiments. Experimental results show that our system can track the angular position reference with the worst position error of 0.43deg and root-mean squared error of 0.16deg at steady state for sinusoidal waveform tracking (with the frequency of 0.1Hz), and the worst position error of 0.2deg for set-point regulation.

This paper demonstrates the experimental validation of canonical peg-in-hole manipulation task using an aerial robot. The robot consists of a multirotor platform equipped with a dual arm multi degree of freedom manipulator. The paper discusses the introduced kinematic constraints which make sure the robot holds a bolt with both arms. We build our peg-in-hole approach using impedance control which is the foundation of compliant interaction with the environment. We utilize a finite state automaton to plan a multi stage strategy which relies on tactile perception in order to pin point the target. Finally, the whole body locomotion is considered, meaning both the degrees of freedom of multirotor base and the dual arm manipulator are considered.

This study proposes an imitation learning method based on force and position information. Force information is required for precise object manipulation but is difficult to obtain because the acting and reaction forces cannot be separated. To separate the forces, we proposed to introduce bilateral control, in which the acting and reaction forces are divided using two robots. In the proposed method, two models of neural networks learn a task; to draw a line along a ruler. We verify the possibility that force information is essential to imitate the human skill of object manipulation.

In this paper, we propose a novel Image-Based Visual Servoing (IBVS) controller for multirotor aerial robots based on a recent deep reinforcement learning algorithm named Deep Deterministic Policy Gradients (DDPG). The proposed RL-IBVS controller is successfully trained in a Gazebo-based simulation scenario in order to learn the appropriate IBVS policy for directly mapping a state, based on errors in the image, to the linear velocity commands of the aerial robot. A thorough validation of the proposed controller has been conducted in simulated and real flight scenarios, demonstrating outstanding capabilities in object following applications. Moreover, we conduct a detailed comparison of the RL-IBVS controller with respect to classic and partitioned IBVS approaches.

Objects in space often exhibit a tumbling motion around the major inertial axis. In this paper, we address the image based visual servoing of a robotic system towards an uncooperative tumbling object. In contrast to previous approaches that require explicit reconstruction of the object and an estimation of its velocity, we propose a novel controller that is able to minimize the feature error directly in image space. This is achieved by observing that the feature points on the tumbling object follow a circular path around the axis of rotation and their projection creates an elliptical track in the image plane. Our controller minimizes the error between this elliptical track and the desired features, such that at the desired pose the features lie on the circumference of the ellipse. The effectiveness of our framework is exhibited by implementing the algorithm in simulation as well on a mobile robot.

We present a novel approach to automatically identify driver behaviors from vehicle trajectories and use them for safe navigation of autonomous vehicles. We propose a novel set of features that can be easily extracted from car trajectories. We derive a data-driven mapping between these features and six driver behaviors using an elaborate web-based user study. We also compute a summarized score indicating a level of awareness that is needed while driving next to other vehicles. We also incorporate our algorithm into a vehicle navigation simulation system and demonstrate its benefits in terms of safer realtime navigation, while driving next to aggressive or dangerous drivers.

This work proposes a system able to recognize the user habits based on his daily activities in a smart house. The habit estimation system uses the information provided by an activity recognition module, which provides the sequence and the duration of the activities performed by the user. Based on those parameters, the activities are represented as a signal by using Fourier series representation. Several output signals from different users are clustered into groups using the k-means method, where each cluster corresponds to a specific habit from a group of people. The proposed system was tested with dataset from the experiment that took place in an environment similar to a smart house. The users were asked to perform a set of 6 activities in any desired orders. In total, twenty-four subjects took part in the experiments. All activities were successfully recognized by the system and three different habits were found. The proposed system along with its habit representation can be potentially used to trace the relationships between the habits observed and some aspects of the user personality traits.

It's a hot summer in 2021 and the blistering sun is shining upon Madrid. You are enjoying some tinto de verano on your terraza. Sizzling in the scorching heat, you are trying to relax. With a simple gesture you call your robotic assistant to help you cool down a little bit. Without further ado, your robot provides some relaxing shade holding a parasol for you, picks up a fan autonomously and starts waving it and the gentle breeze brings you some light relief.

This paper explores the estimation of user attention in the setting of a cooperative handheld robot - a robot designed to behave as a handheld tool but that has levels of task knowledge. We use a tool-mounted gaze tracking system, which, after modelling via a pilot study, we use as a proxy for estimating the attention of the user. This information is then used for cooperation with users in a task of selecting and engaging with objects on a dynamic screen. Via a video game setup, we test various degrees of robot autonomy from fully autonomous, where the robot knows what it has to do and acts, to no autonomy where the user is in full control of the task. Our results measure performance and subjective metrics and show how the attention model benefits the interaction and preference of users.

This paper focuses on numerical methods that can be used to identify hydrodynamic parameters of submerged bodies, namely added mass, linear and quadratic friction coefficients. The mechanical setup is a free decay pendulum that is equipped with an encoder. The first contribution of this paper deals with the comparison of two estimation methods: one method that fits the acceleration of the dynamical model with the acceleration obtained from derivatives of the measured angular position, and another method that fits this position with the angle obtained by numerical integration. The second contribution consists of investigating to what extent estimated added mass and friction coefficient parameters of the dynamical model match the empirical or theoretical values in the case of a spherical object. The results obtained show that the numerical integration method allows to determine the added mass with a good accuracy and a single friction coefficient could be used for the dynamic model without loosing validity.

Visual servoing is a well-known task in robotics. However, there are still challenges when multiple visual sources are combined to accurately guide the robot or occlusions appear. In this paper we present a novel visual servoing approach using hybrid multi-camera input data to lead a robot arm accurately to dynamically moving target points in the presence of partial occlusions. The approach uses four RGBD sensors as Eye-to-Hand (EtoH) visual input, and an arm-mounted stereo camera as Eye-in-Hand (EinH). A Master supervisor task selects between using the EtoH or the EinH, depending on the distance between the robot and target. The Master also selects the subset of EtoH cameras that best perceive the target. When the EinH sensor is used, if the target becomes occluded or goes out of the sensor's view-frustum, the Master switches back to the EtoH sensors to re-track the object. Using this adaptive visual input data, the robot is then controlled using an iterative planner that uses position, orientation and joint configuration to estimate the trajectory. Since the target is dynamic, this trajectory is updated every time-step. Experiments show good performance in four different situations: tracking a ball, targeting a bulls-eye, guiding a straw to a mouth and delivering an item to a moving hand. The experiments cover both simple situations such as a ball that is mostly visible from all cameras, and more complex situations such as the mouth which is partially occluded from some of the sensors.

The factor graph framework is a convenient modeling technique for robotic state estimation where states are represented as nodes, and measurements are modeled as factors. When designing a sensor fusion framework for legged robots, one often has access to visual, inertial, joint encoder, and contact sensors. While visual-inertial odometry has been studied extensively in this framework, the addition of a preintegrated contact factor for legged robots has been only recently proposed. This allowed for integration of encoder and contact measurements into existing factor graphs, however, new nodes had to be added to the graph every time contact was made or broken. In this work, to cope with the problem of switching contact frames, we propose a hybrid contact preintegration theory that allows contact information to be integrated through an arbitrary number of contact switches. The proposed hybrid modeling approach reduces the number of required variables in the nonlinear optimization problem by only requiring new states to be added alongside camera or selected keyframes. This method is evaluated using real experimental data collected from a Cassie-series robot where the trajectory of the robot produced by a motion capture system is used as a proxy for ground truth. The evaluation shows that inclusion of the proposed preintegrated hybrid contact factor alongside visual-inertial navigation systems improves estimation accuracy as well as robustness to vision failure, while its generalization makes it more accessible for legged platforms.

In this paper we present a new robotic control approach inspired in the animal nervous system control. The system implements the binomial Brain-Peripheral Nervous System (CNS-PNS) combining the use of microprocessors as the high level control, or brain, and FPGAs as the low level control, or nervous system. Thanks to the new open source tools for FPGAs, we are able to apply them in the field of robotics in new ways that were impossible before. In this paper, we will demonstrate that our approach is not only able to control the movements of robots using digital circuits built inside an FPGA, but is also capable of generating, synthesizing and uploading them inside the FPGA in real time and on demand.