ICRA 2017 Papers

Visual scene decomposition into semantic entities is one of the major challenges when creating a reliable object grasping system. Recently, we introduced a bottom-up hierarchical clustering approach [1] which is able to segment objects and parts in a scene. In this paper, we introduce a transform from such a segmentation into a corresponding, hierarchical saliency function. In comprehensive experiments we demonstrate its ability to detect salient objects in a scene. Furthermore, this hierarchical saliency defines a most salient corresponding region (scale) for every point in an image. Based on this, an easy-to-use pick and place manipulation system was developed and tested exemplarily.

Assistive passive robotic walkers are naturally modelled as rear-driven bicycle with control on the front steering wheels. Standard path following algorithms used for unicycle-like robots can then be readily available, e.g. using backstepping techniques, to control the walker on desired paths. However, such an approach is usually singular in the very common situation of zero velocity, e.g. whenever the vehicle starts its motion or the user stops for any reason. The paper proposes a non-singular passive path following algorithm for an assistive robotic walker equipped with front steering wheels. The control law avoids the singularities, since it is velocity-independent, and allows the designer to specify saturation limits on the steering angles. The converging properties of the non-singular velocity-independent controller in the presence of saturation constraints are firstly formally proved and tested in simulations. Then extensive experiments performed on 14 testers are presented. These tests underline the promising performance of the proposed controller and the importance of singularities-avoidance in real-world scenarios to increase human comfort along the planned trajectory.

In this paper, we introduce a communication-aware algorithm to navigate mobile agents with non-trivial dynamics in cluttered environments. The navigation technique is based on the Harmonic Potential Field (HPF) approach to motion-planning. The proposed approach employs beamforming at the Base-Station (BS) simultaneously with the robot's motion to increase the Channel Spectral Efficiency (CSE). The approach is developed and basic proofs of performance are provided. Realistic simulations using the WINNER-II wireless channel model are used to demonstrate the navigation algorithm.

There is a considerable need for assistive dressing among people with disabilities, and robots have the potential to fulfill this need. However, training such a robot would require extensive trials in order to learn the skills of assistive dressing. Such training would be time-consuming and require considerable effort to recruit participants and conduct trials. In addition, for some cases that might cause injury to the person being dressed, it is impractical and unethical to perform such trials. In this work, we focus on a representative dressing task of pulling the sleeve of a hospital gown onto a person's arm. We present a system that learns a haptic classifier for the outcome of the task given few (2-3) real-world trials with one person. Our system first optimizes the parameters of a physics simulator using real-world data. Using the optimized simulator, the system then simulates more haptic sensory data with noise models that account for randomness in the experiment. We then train hidden Markov Models (HMMs) on the simulated haptic data. The trained HMMs can then be used to classify and predict the outcome of the assistive dressing task based on haptic signals measured by a real robot's end effector. This system achieves 92.83% accuracy in classifying the outcome of the robot-assisted dressing task with people not included in simulation optimization. We compare our classifiers to those trained on real-world data. We show that the classifiers from our system can categorize the dressing task outcomes more accurately than classifiers trained on ten times more real data.

Multiple robotic agents, autonomous or teleoperated, can be employed to synergise and cooperate to achieve a common objective more effectively. Tasks using robotic manipulators can be eased and improved in terms of reliability, adaptability and ergonomics via robot cooperation. In spite of visual and haptic feedback, cooperative telemanipulation of multiple robots by distant operators can still be challenging due to practical limitations in synchronisation and supervision. This paper presents a new control approach for haptic intention augmentation between two human operators handling objects via teleoperation in a cooperative manner. The force feedback to each operator is enhanced by information on the motion intention of the other operator observed by a force sensor at the input devices. Besides on-ground experiments, an experiment is presented that involves the cooperative teleoperation of an on-ground robot by a cosmonaut on the International Space Station and another distant operator on ground.

We study the dynamics of hovering mechanisms inspired by hummingbirds and bats to extract practical guidelines for the design and control of aerial robots. Using a dynamical model that incorporates wing flapping, pronation and folding, we determine the parameters for which hovering is possible. We show that the pronation angle (rotational degree of freedom for the wing about its span) significantly effects energy consumption of the system and demonstrate an optimal wing-beat frequency that minimizes the average power consumption. Furthermore, we show that increased body rotational inertia can significantly increase the energy consumption of hovering and can lead to implausible hovering trajectories. Last, we show that a bat-inspired hovering mechanism that allows for wing folding can generate plausible hovering behavior for larger aerial systems.

This paper presents an algorithm for solving the problem of tracking smooth curves by a fixed wing unmanned aerial vehicle travelling with a constant airspeed and under a constant wind disturbance. The algorithm is based on the idea of following a guiding vector field which is constructed from the implicit function that describes the desired (possibly time-varying) trajectory. The output of the algorithm can be directly expressed in terms of the bank angle of the UAV in order to achieve coordinated turns. Furthermore, the algorithm can be tuned offline such that physical constraints of the UAV, e.g. the maximum bank angle, will not be violated in a neighborhood of the desired trajectory. We provide the corresponding theoretical convergence analysis and performance results from actual flights.

Crop grow measurement technologies are important to increase the farm productivity. Detection and measurement of fruit volume are useful for forecasting and harvesting applications. Some environmental challenges such as lighting conditions or occlusions make the fruit detection difficult. Our approach is based on features extraction from images through a sub-image clustering technique. Then images being described as a number of pixel in various labels are used in a regression model to estimate the fruit volume. The validity of the proposed method in experimental condition is successfully verified. The method is evaluated also in a field condition but results were inferior to the expectation. This paper tries to elucidate the reasons of the insufficient performance and tries to improve the proposed method in terms of illumination condition, precision and calculation time.

In order to realize adaptive quadruped locomotion on terrains with different properties (such as surface friction or elasticity modulus), we plan to collect the foot-ground contact force and gyroscope information during locomotion on different ground substrates, then classify the ground substrates with the feature vector extracted from the collected data using Support Vector Machine (SVM) algorithm. However, the quadruped walk gait generated by Central Pattern Generators (CPGs) does not perform well on certain ground substrates, e.g., robots may be stuck in the soft ground substrates with small elasticity modulus. Therefore, for one thing, we present a Center of Gravity (COG) adjustment method to eliminate the offset between the control signal generated by CPGs and the actual phase of the quadruped limb, so the limb in theoretical swing phase is able to lift off the ground. For another, we combine CPGs with a foot path planning method to make the lift height controllable. Using these methods, the quadruped robot Biodog realizes the sensor data collection on five different ground substrates. Then we train and classify the sensor data with the SVM. About 99.33% of the five ground substrates can be classified correctly.

We present a friction model for the curved contact area between a deformable object and soft parallel gripper jaws for grasping posture estimation. We show that the assumption of a planar contact area leads to an overestimation of the frictional force and torque, which might cause the object to slip. We simulate the contact with the Finite Element Method, then compute the friction wrenches, which are fitted with two limit surface models: an ellipsoid and a convex 4th-order polynomial. Despite a slightly higher fitting error, the ellipsoid limit surface is chosen to compute the grasp quality because of its simplicity. We compare the limit surfaces of our friction model with the planar contact model and show the improved accuracy obtainable with our model. We then apply the presented model for grasping posture estimation by simulating the contact for all grasp candidates. We show a grasp quality map (quality of all grasp candidates) and the best possible grasp location for several deformable objects.

Grasping tasks have always been challenging for robots, despite recent innovations in vision-based algorithms and object-specific training. If robots are to match human abilities and learn to pick up never-before-seen objects, they must combine vision with tactile sensing. This paper present a novel way to improve robotic grasping: by using tactile sensors and an unsupervised feature-learning approach, a robot can find the common denominators behind successful and failed grasps, and use this knowledge to predict whether a grasp attempt will succeed or fail. This method is promising as it uses only high-level features from two tactile sensors to evaluate grasp quality, and works for the training set as well as for new objects. In total, using a total of 54 different objects, our system recognized grasp failure 83.70% of time.

Wrench-based quality metrics play an important role in many applications such as grasp planning or grasp success prediction. In this work, we study the following discrepancy which is frequently overlooked in practice: the quality metrics are commonly computed under the assumption of sum-magnitude bounded contact forces, but the corresponding grasps are executed by a fully actuated device where the contact forces are limited independently. By means of experiments carried out in simulation and on real hardware, we show that in this setting the values of these metrics are severely underestimated. This can lead to erroneous conclusions regarding the actual capabilities of the grasps under consideration. Our findings highlight the importance of matching the physical properties of the task and the grasping device with the chosen quality metrics.

We consider the problem of grasping concave objects, i.e., objects whose surface includes regions with negative curvature. When a multifingered hand is used to restrain these objects, these areas can be advantageously used to determine grasps capable of more robustly resisting to external disturbance wrenches. We propose a new grasp quality metric specifically suited for this case, and we use it to inform a grasp planner searching the space of possible grasps. Our findings are validated both in simulation and on a real robot system executing a bin picking task. Experimental validation shows that our method is more effective than those not explicitly considering negative curvature.

Robotic teleoperation from a human operator's pose demonstrations provides an intuitive and effective means of control that has been made feasible by improvements in sensor technologies in recent years. However, the imprecision of low-cost depth cameras and the difficulty of calibrating a frame of reference for the operator introduce inefficiencies in this process when performing tasks that require interactions with objects in the robot's workspace. We develop a goal-predictive teleoperation system that aids in “de-noising” the controls of the operator to be more goal-directed. Our approach uses inverse optimal control to predict the intended object of interaction from the current motion trajectory in real time and then adapts the degree of autonomy between the operator's demonstrations and autonomous completion of the predicted task. We evaluate our approach using the Microsoft Kinect depth camera as our input sensor to control a Rethink Robotics Baxter robot.

Publicly available map services are widely used by humans for navigation and nowadays provide almost complete road network data. When utilizing such maps for autonomous navigation with mobile robots one is faced with the problem of inaccuracies of the map and the uncertainty about the position of the robot relative to the map. In this paper, we present a probabilistic approach to autonomous robot navigation using data from OpenStreetMap that associates tracks from Open-StreeetMap with the trails detected by the robot based on its 3D-LiDAR data. It combines semantic terrain information, derived from the 3D-LiDAR data, with a Markov-Chain Monte-Carlo technique to match the tracks from OpenStreetMap with the sensor data. This enables our robot to utilize OpenStreetMap for navigation planning and to still stay on the trails during the execution of these plans. We present the results of extensive experiments carried out in real world settings that demonstrate the robustness of our system regarding the alignment of the vehicle pose relative to the OpenStreetMap data.

In this paper we introduce a gesture-based robot control framework, we discuss the adopted design principles and we report results about its evaluation with humans. Gesture-based control using wearable devices may constitute a novel form of human-robot interaction, but its implications have not been discussed in the literature. We discuss the main challenging issues, possible design guidelines and an open source, freely available implementation using commercially available devices and robots. The overall performance of the architecture, as well as its validation with 27 untrained volunteers, is reported.

Aerial robots are becoming popular among general public, and with the development of artificial intelligence (AI), there is a trend to equip aerial robots with a natural user interface (NUI). Hand/arm gestures are an intuitive way to communicate for humans, and various research works have focused on controlling an aerial robot with natural gestures. However, the techniques in this area are still far from mature. Many issues in this area have been poorly addressed, such as the principles of choosing gestures from the design point of view, hardware requirements from an economic point of view, considerations of data availability, and algorithm complexity from a practical perspective. Our work focuses on building an economical monocular system particularly designed for gesture-based piloting of an aerial robot. Natural arm gestures are mapped to rich target directions and convenient fine adjustment is achieved. Practical piloting scenarios, hardware cost and algorithm applicability are jointly considered in our system design. The entire system is successfully implemented in an aerial robot and various properties of the system are tested.

Controlled hovering of motor driven flapping wing micro aerial vehicles (FWMAVs) is challenging due to its limited control authority, large inertia, vibration produced by wing strokes, and limited components accuracy due to fabrication methods. In this work, we present a hummingbird inspired FWMAV with 12 grams of weight and 20 grams of maximum lift. We present its full non-linear dynamic model including the full inertia tensor, non-linear input mapping, and damping effect from flapping counter torques (FCTs) and flapping counter forces (FCFs). We also present a geometric flight controller to ensure exponentially stable and globally exponential attractive properties. We experimentally demonstrated the vehicle lifting off and hover with attitude stabilization.

Structural Analysis is a lucrative option for Fault Detection and Identification in Unmanned Aerial Vehicles (UAVs), because it handles detailed, large-scale mathematical models. It can be employed by an on-board flight computer to generate residual generators and implement automatic fault-detection. Contemporary algorithms applied on dynamic systems may yield residual generators which require the real-time solution of Differential-Algebraic Equation (DAE) systems. Depending on the form and differential index of each DAE system, its solution may not be possible exclusively by computational means. In this paper we explore the relation between Structural Analysis algorithms and the forms of DAE systems they produce, propose conditions under which all generated DAEs are Structural Index-1 and semi-explicit and provide a large-scale fixed-wing UAV model with that property.