IROS 2016 论文列表

Optimal control approaches in combination with trajectory optimization have recently proven to be a promising control strategy for legged robots. Computationally efficient and robust algorithms were derived using simplified models of the contact interaction between robot and environment such as the linear inverted pendulum model (LIPM). However, as humanoid robots enter more complex environments, less restrictive models become increasingly important. As we leave the regime of linear models, we need to build dedicated solvers that can compute interaction forces together with consistent kinematic plans for the whole-body. In this paper, we address the problem of planning robot motion and interaction forces for legged robots given predefined contact surfaces. The motion generation process is decomposed into two alternating parts computing force and motion plans in coherence. We focus on the properties of the momentum computation leading to sparse optimal control formulations to be exploited by a dedicated solver. In our experiments, we demonstrate that our motion generation algorithm computes consistent contact forces and joint trajectories for our humanoid robot. We also demonstrate the favorable time complexity due to our formulation and composition of the momentum equations.

Persistent merging of maps created by different sensor modalities is an insufficiently addressed problem. Current approaches either rely on appearance-based features which may suffer from lighting and viewpoint changes or require pre-registration between all sensor modalities used. This work presents a framework using structural descriptors for matching LIDAR point-cloud maps and sparse vision keypoint maps. The matching algorithm works independently of the sensors' viewpoint and varying lighting and does not require pre-registration between the sensors used. Furthermore, we employ the approach in a novel vision-laser map-merging algorithm. We analyse a range of structural descriptors and present results of the method integrated within a full mapping framework. Despite the fact that we match between the visual and laser domains, we can successfully perform map-merging using structural descriptors. The effectiveness of the presented structure-based vision-laser matching is evaluated on the public KITTI dataset and furthermore demonstrated on a map merging problem in an industrial site.

Automating item picking in the e-commerce warehouse is pressing but challenging, due to a massive variety of items, tight environmental constraints and item location uncertainty. In this paper, we present an effective and efficient strategy-based planning approach to implement the robotic picking from shelves for e-commerce. Making full advantage of a gripper with multiple securing methods, differentiated strategies are modeled as picking primitives with different securing methods. A strategy generator is proposed to produce feasible potential pickings as quickly and as successfully as possible. A strategy evaluator considering reachability, collision, object-bias preference and the securing performance is also presented for ranking the picking strategies. Experiments were conducted to validate that the robotic picker is able to plan a picking strategy within 2 ms and pick daily items from the shelves with an average success rate of 68%.

This paper presents a graph-based model of periodic migrations of tagged fish populations and two multi-AUV stochastic controllers developed to track these fish from the model. The model presented in this paper characterizes patterns in the historical movement of tagged fish and is used to develop stochastic tracking by a “model based control” and a “feedback control”. These two controllers permit swarms of AUVs to track the transition probabilities of the tagged population between vertices of the model. To validate these controllers, a periodic model is developed for a simulated population based on three months of geolocation data from a kelp bass (Paralabrax clathratus), and AUV teams utilizing both controllers are simulated in tracking this population. Results show the viability of stochastic controls for multi-AUV tracking of populations whose behavior is well-approximated by the graph-based model. Preliminary trials with physical AUV systems indicate the plausibility of hardware implementation.

This work investigates the applicability of stiffness-based modelling in soft robotics manipulation. The methodology is introduced and applied to model a soft robotics manipulator as single 3d Timoshenko beam element. The model is then utilized to solve the forward kinematics problem for the manipulator. The algorithm is validated comparing the simulated deflection with the deflection of the physical manipulator for two defined pressure sequences. It is shown that the model behaves in a highly similar fashion in comparison to the manipulator. For both trajectories the maximum position error is close to 6 mm while the error in orientation not more than 18°. The methodology as described in this work reveals great applicability to the field of soft robots being limited only by the stiffness matrix assembly for the given system. Implementations of inverse kinematics and the effects of external force applications are effectively integrable in the described theory.

Haptic rendering of soft bodies is essential in medical simulations of procedures such as surgery or palpation. The most commonly used approach is to recreate the sense of touch using a specific design and control of a robotic arm. In this paper, we propose a new approach, based on soft-robotics technology. We create a tangible deformable device that allows users to “touch” soft tissues and perceive mechanical material properties, in a realistic manner. The device is able to dynamically provide user touch with different stiffness perceptions, thanks to actuators placed at the boundaries. We introduce a control algorithm, based on inverse Finite Element Analysis, which controls the actuators in order to recreate a desired stiffness that corresponds to the contact with soft tissues in the virtual environment. The approach uses antagonistic actuation principle to create a wide range of stiffness. We validate our algorithm and demonstrate the method using prototypes based on simple mechanisms.

This paper concerns the self-localization problem of a hexapod walking robot operating in rough terrains. Given that legged robots exhibit higher terrain passability than wheeled or tracked platforms when operating in harsh environments, they constitute a challenge for the localization techniques because the camera motion between consecutive frames can be arbitrary due to the motion gait and terrain irregularities. In this paper, we present and evaluate an inertially assisted Stereo Parallel Tracking and Mapping (S-PTAM) method deployed on a hexapod crawling robot in a rough terrain. The considered deployment scenario is motivated by autonomous navigation in an unknown environment in an open loop fashion. The reported results and comparison with an existing RGB-D SLAM technique show the feasibility of the proposed approach and its suitability for navigation of crawlers in harsh environments.

Many highly dynamic novel mobile robots have been developed being inspired by animals. In this study, we are inspired by a basilisk lizard's ability to run and steer on water surface for a hexapedal robot. The robot has an active tail with a circular plate, which the robot rotates to steer on water. We dynamically modeled the platform and conducted simulations and experiments on steering locomotion with a bang-bang controller. The robot can steer on water by rotating the tail, and the controlled steering locomotion is stable. The dynamic modelling approximates the robot's steering locomotion and the trends of the simulations and experiments are similar, although there are errors between the desired and actual angles. The robot's maneuverability on water can be improved through further research.

For many tasks, tactile or visual feedback is helpful or even crucial. However, designing controllers that take such high-dimensional feedback into account is non-trivial. Therefore, robots should be able to learn tactile skills through trial and error by using reinforcement learning algorithms. The input domain for such tasks, however, might include strongly correlated or non-relevant dimensions, making it hard to specify a suitable metric on such domains. Auto-encoders specialize in finding compact representations, where defining such a metric is likely to be easier. Therefore, we propose a reinforcement learning algorithm that can learn non-linear policies in continuous state spaces, which leverages representations learned using auto-encoders. We first evaluate this method on a simulated toy-task with visual input. Then, we validate our approach on a real-robot tactile stabilization task.

To guarantee its balance, a humanoid robot has to respect some contact force constraints. Therefore, traditional controllers generate motions complying with these constraints, but they usually consider the robot as stiff and the joint position perfectly known. However, several robots contain compliant parts in their structure. This flexibility modifies the forces at contacts and endangers balance. However, most solutions to stabilize the robot rely on force sensors. But several humanoid robots aren't equipped with these sensors. This paper has two aims. The first one is to develop a compliance stabilizer using the center of mass position and upper-body orientation through a viscoelastic reaction mass pendulum model. The second objective is to show the performances of such a stabilizer when relying only on an IMU-based state observer. Experimental results on HRP-2 robot show that the stabilization successfully rejects perturbations with high gains using only these IMU signals. Moreover, the actuation of the upper-body orientation provides redundancy, robustness and finally improved performances to the stabilizer.

Envisioned applications for humanoid robots call for the design of balancing and walking controllers. While promising results have been recently achieved, robust and reliable controllers are still a challenge for the control community dealing with humanoid robotics. Momentum-based strategies have proven their effectiveness for controlling humanoids balancing, but the stability analysis of these controllers is still missing. The contribution of this paper is twofold. First, we numerically show that the application of state-of-the-art momentum-based control strategies may lead to unstable zero dynamics. Secondly, we propose simple modifications to the control architecture that avoid instabilities at the zero-dynamics level. Asymptotic stability of the closed loop system is shown by means of a Lyapunov analysis on the linearized system's joint space. The theoretical results are validated with both simulations and experiments on the iCub humanoid robot.

In this paper, we propose a framework to map stationary sound sources while simultaneously localise a moving robot. Conventional methods for localisation and sound source mapping rely on a microphone array and either, 1) a proprioceptive sensor only (such as wheel odometry) or 2) an additional exteroceptive sensor (such as cameras or lasers) to get accurately the robot locations. Since odometry drifts over time and sound observations are bearing-only, sparse and extremely noisy, the former can only deal with relatively short trajectories before the whole map drifts. In comparison, the latter can get more accurate trajectory estimation over long distances and a better estimation of the sound source map as a result. However, in most of the work in the literature, trajectory estimation and sound source mapping are treated as uncorrelated, which means an update on the robot trajectory does not propagate properly to the sound source map. In this paper, we proposed an efficient method to correlate robot trajectory with sound source mapping by exploiting the conditional independence property between two maps estimated by two different Simultaneous Localisation and Mapping (SLAM) algorithms running in parallel. In our approach, the first map has the flexibility that can be built with any SLAM algorithm (filtering or optimisation) to estimate robot poses with an exteroceptive sensor. The second map is built by using a filtering-based SLAM algorithm locating all stationary sound sources parametrised with Inverse Depth Parametrisation (IDP). Robot locations used during IDP initialisation are the common features shared between the two SLAM maps, which allow to propagate information accordingly. Comprehensive simulations and experimental results show the effectiveness of the proposed method.

This paper presents a spatial layout recovery approach from single omnidirectional images. Vertical structures in the scene are extracted via classification from heterogeneous features computed at the superpixel level. Vertical surfaces are further classified according to their main orientation by fusing oriented line features, floor-wall boundary features and histogram of oriented gradients (HOG) with a Random Forest classifier. Oriented line features are used to build an orientation map that considers the main vanishing points. The floor-wall boundary feature attempts to reconstruct the scene shape as if it were observed from a bird's-eye view. Finally, the HOG descriptors are aggregated per superpixel and summarize the gradient distribution at homogeneous appearance regions. Compared to existing methods in the literature which rely only on corners or lines, our method gains statistical support from multiple cues aggregated per superpixel which provide more robustness against noise, occlusion, and clutter.

We address the following question: is it possible to reconstruct the geometry of an unknown environment using sparse and incomplete depth measurements? This problem is relevant for a resource-constrained robot that has to navigate and map an environment, but does not have enough on-board power or payload to carry a traditional depth sensor (e.g., a 3D lidar) and can only acquire few (point-wise) depth measurements. In general, reconstruction from incomplete data is not possible, but when the robot operates in man-made environments, the depth exhibits some regularity (e.g., many planar surfaces with few edges); we leverage this regularity to infer depth from incomplete measurements. Our formulation bridges robotic perception with the compressive sensing literature in signal processing. We exploit this connection to provide formal results on exact depth recovery in 2D and 3D problems. Taking advantage of our specific sensing modality, we also prove novel and more powerful results to completely characterize the geometry of the signals that we can reconstruct. Our results directly translate to practical algorithms for depth reconstruction; these algorithms are simple (they reduce to solving a linear program), and robust to noise. We test our algorithms on real and simulated data, and show that they enable accurate depth reconstruction from a handful of measurements, and perform well even when the assumption of structured environment is violated.

This paper presents the design of a small aerial robot for inhabited microgravity environments, such as orbiting space stations (e.g., ISS). In particular, we target a fleet of robots, called Space CoBots, for collaborative tasks with humans, such as telepresence and cooperative mobile manipulation. The design is modular, comprising an hexrotor based propulsion system, and a stack of modules including batteries, cameras for navigation, a screen for telepresence, a robotic arm, space for extension modules, and a pair of docking ports. These ports can be used for docking and for mechanically attaching two Space CoBots together. The kinematics is holonomic, and thus the translational and the rotational components can be fully decoupled. We employ a multi-criteria optimization approach to determine the best geometric configuration for maximum thrust and torque across all directions. We also tackle the problem of motion control: we use separate converging controllers for position and attitude control. Finally, we present simulation results using a realistic physics simulator. These experiments include a sensitivity evaluation to sensor noise and to unmodeled dynamics, namely a load transportation.

Soft robotics methods such as impedance control and reflexive collision handling have proven to be a valuable tool to robots acting in partially unknown and potentially unstructured environments. Mainly, the schemes were developed with focus on classical electromechanically driven, torque controlled robots. There, joint friction, mostly coming from high gearing, is typically decoupled from link-side control via suitable rigid or elastic joint torque feedback. Extending and applying these algorithms to stiff hydraulically actuated robots poses problems regarding the strong influence of friction on joint torque estimation from pressure sensing, i.e. link-side friction is typically significantly higher than in electromechanical soft robots. In order to improve the performance of such systems, we apply state-of-the-art fault detection and estimation methods together with observer-based disturbance compensation control to the humanoid robot Atlas. With this it is possible to achieve higher tracking accuracy despite facing significant modeling errors. Compliant end-effector behavior can also be ensured by including an additional force/torque sensor into the generalized momentum-based disturbance observer algorithm from [1].

Social activity based on body motion is a key feature for non-verbal and physical behavior defined as function for communicative signal and social interaction between individuals. Social activity recognition is important to study human-human communication and also human-robot interaction. Based on that, this research has threefold goals: (1) recognition of social behavior (e.g. human-human interaction) using a probabilistic approach that merges spatio-temporal features from individual bodies and social features from the relationship between two individuals; (2) learn priors based on physical proximity between individuals during an interaction using proxemics theory to feed a probabilistic ensemble of activity classifiers; and (3) provide a public dataset with RGB-D data of social daily activities including risk situations useful to test approaches for assisted living, since this type of dataset is still missing. Results show that using the proposed approach designed to merge features with different semantics and proximity priors improves the classification performance in terms of precision, recall and accuracy when compared with other approaches that employ alternative strategies.

Robots capable of both locomotion and interaction with the environment are necessary for robots to move from ideal laboratory situations to real applications. Snake robots have been researched for locomotion in unstructured environments due to its unique and adaptable gaits. However, they have not been used to interact with the environment in a dexterous manner, for example to grasp or push an object. In this paper, the model of a snake robot in contact with external objects (or the environment) is derived. Metrics that are coordinateindependent are derived in order to quantify the wrench that a snake robot could exert into these objects. In particular, we show that the configuration of the robot plays a significant role in these metrics. The model and metrics are tested in a study case.

We propose a new optimal trajectory generation technique on SE(3) which avoids known obstacles. We leverage techniques from differential geometry and Lie algebra to formulate a cost functional which is intrinsic to the geometric structure of this space and makes physical sense. We propose an approximation technique to generate trajectories on the subgroup SO(3) and use Semidefinite Programming (SDP) to approximate an NP-Hard problem with one which is tractable to compute. From this trajectory on the subgroup, the trajectory generation on the other dimensions of the group becomes a Quadratic Program (QP). For obstacle avoidance, we use a computational geometric technique to decompose the environment into overlapping convex regions to confine the trajectory. We show how this motion planning technique can be used to generate feasible trajectories for a space robot in SE(3) and describe controllers that enable the execution of the generated trajectory. We compare our method to other geometric techniques for calculating trajectories on SO(3) and SE(3), but in an obstacle-free environment.