IROS 2016 Papers

In this paper a robust PID controller for quadrotor unmanned aerial vehicles is proposed that uses the parameter space approach. Stability and robustness analyses are carried out in the controller parameter space to determine a set of stable controller gains that guarantee also robustness against system parameter uncertainties. Additionally, the trade-off between robustness and performance is included in the control gain choice. Experimental results validate the proposed approach, where the robust behavior of a quadrotor is shown for step response and path following.

Despite the availability in the literature of several constraint-based motion generation algorithms, modest attention has been paid to their robustness with respect to noise, and more in general, to unstructured uncertainties. Especially in the case of sensor-related constraints, the envisaged robustness properties are clearly crucial to enforce the correct and expected behaviour of these algorithms. This paper contributes with a method to explicitly account for different sources of uncertainty. We also suggest a computational efficient way to consistently modify the constraint specification in order to obtain such robustness. An experimental verification on a visual aided grasping task, where visibility of the object is to be maintained, enlightens the benefits of the proposed approach in terms of achieving the desired robustness.

Industrial robots are often used for the simulation of satellites during on orbit servicing. In order to cover also the docking phase, both robots are equipped with force-torque sensors, and the measured forces and torques are taken to compute the desired motion of the position controlled robots. Since the system dynamics of robots and of free floating bodies obviously differ, for each robot we distinguish between the really executed and the assumed satellite motion. The difference between the two motions is used to adapt the measured forces in such a way that they correspond to the satellite's trajectory. In this way the docking procedure can be visualized by two robots which closely follow the satellites' trajectories. Stability of the robot control is not compromised even if the dynamics of the satellites and the robots are totally different. Simulation results verify the approach.

In complex manipulation scenarios (e.g. tasks requiring complex interaction of two hands or in-hand manipulation), generalization is a hard problem. Current methods still either require a substantial amount of (supervised) training data and / or strong assumptions on both the environment and the task. In this paradigm, controllers solving these tasks tend to be complex. We propose a paradigm of maintaining simpler controllers solving the task in a small number of specific situations. In order to generalize to novel situations, the robot transforms the environment from novel situations into a situation where the solution of the task is already known. Our solution to this problem is to play with objects and use previously trained skills (basis skills). These skills can either be used for estimating or for changing the current state of the environment and are organized in skill hierarchies. The approach is evaluated in complex pick-and-place scenarios that involve complex manipulation. We further show that these skills can be learned by autonomous playing.

Recent works of non-rigid registration have shown promising applications on tasks of deformable manipulation. Those approaches use thin plate spline-robust point matching (TPS-RPM) algorithm to regress a transformation function, which could generate a corresponding manipulation trajectory given a new pose/shape of the object. However, this method regards the object as a bunch of discrete and independent points. Structural information, such as shape and length, is lost during the transformation. This limitation makes the object's final shape to differ from training to test, and can sometimes cause damage to the object because of excessive stretching. To deal with these problems, this paper introduces a tangent space mapping (TSM) algorithm, which maps the deformable object in the tangent space instead of the Cartesian space to maintain structural information. The new algorithm is shown to be robust to the changes in the object's pose/shape, and the object's final shape is similar to that of training. It is also guaranteed not to overstretch the object during manipulation. A series of rope manipulation tests are performed to validate the effectiveness of the proposed algorithm.

Due to the faster-is-slower phenomenon in emergency escape, it is desirable to regulate pedestrian flow at the exit or a bottleneck. Modification of pedestrian facilities was previously studied to increase the efficiency and safety by the transportation community. We propose a robot-assisted pedestrian regulation scheme and study passive human-robot interaction (HRI), where the robot acts as a dynamic obstacle that interacts with pedestrians. Such a robot-assisted solution replaces expensive infrastructure modification with real-time reconfigurability. In the paper, we first formulate a robot-assisted flow optimization problem based on the social force models of pedestrian dynamics with embedded HRI forces. We then present an online learning algorithm based on adaptive dynamic programming (ADP) to generate motion control so that the robot can replan and adapt its motion to real-time pedestrian flows. The ADP control process uses observed flow information only but not the models of pedestrians, and provides feedback control with online learning and control capability. Simulation results demonstrate efficiency of the proposed method.

Autonomous self-assembly allows to create structures and scaffolds on demand and automatically. The desired structure may be predetermined or alternatively it is the result of an artificial growth process that adapts to environmental features and to the intermediate structure itself. In a self-organizing and decentralized control approach the robots interact only locally and form the structure collectively. Designing a complete approach that allows the robot group to collectively decide on where to start the self-assembly, that adapts at runtime to environmental conditions, and that guarantees the structural stability, is challenging and does not yet exist. We present an approach to self-assembly inspired by diffusion-limited aggregation that generates an adaptive structure reacting to environmental conditions in an artificial growth process. During a preparatory stage the robots collectively decide where to start the self-assembly also depending on environmental conditions. In the actual self-assembly stage, the robots create tree-like structures that grow towards light. We report the results of robot self-assembly experiments with 50 Kilobots. Our results demonstrate how an adaptive growth process can be implemented in robots. We briefly describe our future work of how to extend the approach to a 3-d growth process and how robot self-assembly as an open-ended adaptive growth process opens up a multiplicity of future opportunities.

Robots are at the position to become our everyday companions in the near future. Still, many hurdles need to be cleared to achieve this goal. One of them is the fact that robots are still not able to perceive some important communication cues naturally used by humans, e.g. gaze. In the recent past, eye gaze in robot perception was substituted by its proxy, head orientation. Such an approach is still adopted in many applications today. In this paper we introduce performance improvements to an eye tracking system we previously developed and use it to explore if this approximation is appropriate. More precisely, we compare the impact of the use of eye- or head-based gaze estimation in a human robot interaction experiment with the iCub robot and naïve subjects. We find that the possibility to exploit the richer information carried by eye gaze has a significant impact on the interaction. As a result, our eye tracking system allows for a more efficient human-robot collaboration than a comparable head tracking approach, according to both quantitative measures and subjective evaluation by the human participants.

This paper proposes a particular case of a robot body design method which determines a degrees of freedom (DOFs) number and link parameters to maximize a target task performance. The DOFs number is an essential point to be considered in the robot body design problem. In this paper, the target task is to make a long throw, and multi DOFs ball throwing robot is designed. Design parameters are the robot body parameters and its motion pattern, and they are designed to maximize ball flying distance under long throw task conditions. To define the link lengths and the robot DOFs number as design parameters, it is assumed that intermediate links of the robot have identical actuators, and these link parameters are defined as functions of link lengths. These links are chained to construct the whole link system. Because of this assumption, the motion equation, which is utilized in the task conditions, is determined by the given robot DOFs number and link lengths. The proposed method was applied to the ball throwing robot model, and its body parameters and motion pattern were designed in the proposed calculation algorithm. As a result, 5 DOFs robot and its throwing motion were obtained, and the ball flying distance was maximized. The ball flying distance was changed along with the DOFs number, and the effectiveness of the proposed design method was demonstrated.

This paper presents a robot capable of painting colorful pictures with a visual control system like human artists. It can use only five basic colors (cyan, magenta, yellow, white and black) to mix a variety of colors. After receiving a picture, the Robot Artist creates the artwork in two stages—the underpainting and the refinement. In the underpainting stage, it covers the canvas with a thin layer of acrylic paint, which can effectively set the basic tone of the painting. In the second stage, the robot uses the camera attached to its wrist to obtain the current artwork and refine it repeatedly. During the refinement, the methodology of non-photorealistic rendering (NPR) is also introduced to generate hand-painted strokes, which will be painted on the canvas to refine the artworks.

Traversability analysis is an important aspect of autonomous navigation in robotics. In this paper, we relate the idea of traversability to safety and ease of road usage by defining a novel sensor-data driven metric called Road Traversability Index (RTI). The RTI translate the geometric interaction of vehicle with road into a distance modulated index that can be used as advice for a human driver or an autonomous agent intending to traverse a particular road segment using a specific vehicle. We present a framework in which 3D sensor data is converted into a road model, which in turn is converted into a roadmap based motion planning graph to represent the underlying configuration space. The RTI is defined as a function of the roadmap by axiomatically satisfying all required properties of road traversability. We have tested our algorithmic framework on simulated scenarios to explore safety; and real-world data sets to discover aspects of traversability for vehicles of various types. Experimental results show that RTI is a practical tool that reveals information that may be hidden to human inspection or other methods of assessment that do not explicitly model a vehicle.

The perception of the dynamic aspects of the environment is a highly relevant precondition for the realization of autonomous robot system acting in the real world. In this paper, we propose a novel method for estimating dense rigid scene flow in 3D LiDAR scans. We formulate the problem as an energy minimization problem, where we assume local geometric constancy and incorporate regularization for smooth motion fields. Analyzing the dynamics at point level helps in inferring the fine-grained details of motion. We show results on multiple sequences of the KITTI odometry dataset, where we seamlessly estimate multiple motions pertaining to different dynamic objects. Furthermore, we test our approach on a dataset with pedestrians to show how our method adapts to a case with non-rigid motion. For comparison we use the ground truth from KITTI and show how our method outperforms different ICP-based methods.

This paper addresses the problem of sound-source localization (SSL) with a robot head, which remains a challenge in real-world environments. In particular we are interested in locating speech sources, as they are of high interest for human-robot interaction. The microphone-pair response corresponding to the direct-path sound propagation is a function of the source direction. In practice, this response is contaminated by noise and reverberations. The direct-path relative transfer function (DP-RTF) is defined as the ratio between the direct-path acoustic transfer function (ATF) of the two microphones, and it is an important feature for SSL. We propose a method to estimate the DP-RTF from noisy and reverberant signals in the short-time Fourier transform (STFT) domain. First, the convolutive transfer function (CTF) approximation is adopted to accurately represent the impulse response of the microphone array, and the first coefficient of the CTF is mainly composed of the direct-path ATF. At each frequency, the frame-wise speech auto- and cross-power spectral density (PSD) are obtained by spectral subtraction. Then a set of linear equations is constructed by the speech auto- and cross-PSD of multiple frames, in which the DP-RTF is an unknown variable, and is estimated by solving the equations. Finally, the estimated DP-RTFs are concatenated across frequencies and used as a feature vector for SSL. Experiments with a robot, placed in various reverberant environments, show that the proposed method outperforms two state-of-the-art methods.

Achieving resonance in flapping wings has been recognized as one of the most important principles to enhance power efficiency, lift generation, and flight control performance of high-frequency flapping wing micro air vehicles (MAVs). Most of work on the development of such vehicles have attempted to achieve wing flapping resonance. However, the theoretical understanding of its effects on the response and energetics of flapping motion has lagged behind, leading to sub-optimal design decisions and misinterpretations of experimental results. In this work, we systematically model the dynamics of flapping wing as a forced nonlinear resonant system. Using linear approximation approach, we derived analytic solution for steady-state flapping amplitude, energetics, and characteristic frequencies including natural frequency, damped natural frequency, and peak frequency. Our results showed that both aerodynamic lift and power efficiency are maximized by driving the wing at natural frequency, instead of other frequencies. Interestingly, the flapping velocity is maximized at natural frequency as well, which can lead to an easy experimental approach to identify natural frequency and validate the resonance design. Our models and analysis were validated with both simulations and experiments on ten different wings mounted a direct-motor-drive flapping wing MAV. The result can serve as a systematic design principle and guidance in the interpretations of empirical results.

This paper proposes the concept of the center of gravity (COG) viscoelasticity to associate joint viscoelasticity with the COG-zero moment point (ZMP) model of humanoid dynamics. Although COG viscoelasticity is based on the well-known kinematic relationship between joint stiffness and end-effector stiffness, it provides practical advantages for humanoid motion control. Once the feedback gain in the COG-ZMP model is designed using the control theory, COG viscoelasticity can be applied to directly transform it to joint viscoelasticity. The author names this method as resolved COG viscoelasticity control (RCVC). In particular, this paper proposes RCVC in which the null-space of the COG Jacobian is employed. The validity of the RCVC is verified by simulating whole-body dynamics.

Geometric distortion associated with underwater flat port housings considerably affects the accuracy of scene depth and visual pose estimation. To compensate for this, a new solution for the forward projection is proposed based on the refractive camera model and Taylor expansion. This allows for an easy integration of the flat refractive camera model into non-linear optimization problems like bundle adjustment.

For generating motions of robots, global search in configuration space is time consuming although it is sometimes indispensable (e.g. collision avoidance in complex environment). Our idea is to use global sampling algorithm not in the state space but in the task nullspace, which is the redundant degrees of freedom of the state space with respect to the task space. Because the task nullspace is smaller than the original search space (state space), fast global sampling is possible. For embedding this hidden task nullspace parameters, we propose a new deep-auto-encoder-based neural network structure. Our approach learns the map from task and task nullspace towards robot's state (Task-State Map, TSM). As the demonstration, the relationship between 28-dof joint angles (state) and the end-effector coordinates of all limbs (task) is learned, and egress postures and reaching postures are generated.

In most 6 DOF robot arms, considerable amounts of gravitational torques due to the robot's own weight are applied to pitch joints of the robot, which causes most arms to use high capacity motors and speed reducers. A spring-based counterbalance mechanism can compensate for this gravitational torque, thus leading to a significant reduction in the effect of gravity. However, a simple installation of counterbalance mechanisms at each pitch joint does not work properly because the gravitational torque at each joint is dependent also on the other joints. To achieve multi-DOF counterbalancing, we propose a parallelogram linkage combined with dual counterbalance mechanisms, each being composed of a slider-crank mechanism and springs. Simulations and experimental results showed that the counterbalance robot arm based on the proposed counterbalance mechanisms effectively reduced the torques required to support the robot mass, thus allowing the prospective use of much smaller motors and speed reducers than traditional industrial robots.

Multi-concept visual classification is emerging as a common environment perception technique, with applications in autonomous mobile robot navigation. Supervised visual classifiers are typically trained with large sets of images, hand annotated by humans with region boundary outlines followed by label assignment. This annotation is time consuming, and unfortunately, a change in environment requires new or additional labeling to adapt visual perception. The time is takes for a human to label new data is what we call adaptation latency. High adaptation latency is not simply undesirable but may be infeasible for scenarios with limited labeling time and resources. In this paper, we introduce a labeling framework to the environment perception domain that significantly reduces adaptation latency using unsupervised learning in exchange for a small amount of label noise. Using two real-world datasets we demonstrate the speed of our labeling framework, and its ability to collect environment labels that train high performing multi-concept classifiers. Finally, we demonstrate the relevance of this label collection process for visual perception as it applies to navigation in outdoor environments.