ICRA 2017 论文列表

We present a novel powered knee prosthesis with an active variable transmission (AVT) that adapts motor torque and speed output based on the requirements of different ambulation activities. The AVT works in combination with a spring/damper system to allow for both active and passive operation modes. Preliminary tests performed by an able-bodied subject wearing a bypass orthosis show that the proposed knee can support walking on level ground in passive mode, as well as ambulating on stairs with a reciprocal gait pattern in active mode.

We study the problem of reducing the amount of communication in a distributed target tracking problem. We focus on the scenario where a team of robots are allowed to move on the boundary of the environment. Their goal is to seek a formation so as to best track a target moving in the interior of the environment. The robots are capable of measuring distances to the target. Decentralized control strategies have been proposed in the past that guarantee that the robots asymptotically converge to the optimal formation. However, existing methods require that the robots exchange information with their neighbors at all time steps. Instead, we focus on reducing the amount of communication among robots. We propose a self-triggered communication strategy that decides when a particular robot should seek up-to-date information from its neighbors and when it is safe to operate with possibly outdated information from the neighbor. We prove that this strategy converges to an optimal formation. We compare the two approaches (constant communication and self-triggered communication) through simulations of tracking stationary and mobile targets.

This paper considers the modelling of scalar fields exhibiting non-stationary noise in the context of Gaussian Process (GP) regression. We show how a Heteroscedastic GP produces more accurate predictions of the variance of a process of this type compared to the standard Homoscedastic model. We present a parametric model for the noise process and derive analytical solutions to the Log Marginal Likelihood of the data and its gradients with respect to Hyper Parameters of the kernel and the noise process. We compare our parametric model to one which estimates a full GP for the noise and show analogous predictive performance with a model which has greater computational efficiency and is less complex to implement. We also discuss active sample selection in this framework and show through the numerical simulation of an arrested bathymetric front in an estuary, the superiority of using Mutual Information to Fisher Information, Entropy or Random sampling in terms of errors in the first two moments of the predicted distributions.

We propose an active exposure control method to improve the robustness of visual odometry in HDR (high dynamic range) environments. Our method evaluates the proper exposure time by maximizing a robust gradient-based image quality metric. The optimization is achieved by exploiting the photometric response function of the camera. Our exposure control method is evaluated in different real world environments and outperforms the built-in auto-exposure function of the camera. To validate the benefit of our approach, we adapt a state-of-the-art visual odometry pipeline (SVO) to work with varying exposure time and demonstrate improved performance using our exposure control method in challenging HDR environments.

This work considers estimation of mass parameters for multi-robot coordinated lifting in the context of coordinated aerial manipulation, and develops strategies for active parameter estimation for cooperative manipulation tasks through an information-theoretic framework. The active sensing problem is formulated based on application of increasing forces to the object and detection of small motions that occur when the center of pressure exits the convex hull formed by existing contacts. In order to enable identification of informative actions, we develop and employ a closed-form solution of Cauchy-Schwarz quadratic mutual information (Ics) for non-parametric filters. The evaluation considers iterative selection from a finite set of measurements and demonstrates that choosing measurements to maximize Ics significantly improves the convergence rate of the parameter estimates compared to random and cyclic selection methods. This approach is extended to consider actuator constraints and feasible lifting configurations and achieves an 80% success rate in formation of feasible lifting configurations compared to a 53% baseline performance.

Deep learning models have achieved state-of-the-art performance in recognizing human activities, but often rely on utilizing background cues present in typical computer vision datasets that predominantly have a stationary camera. If these models are to be employed by autonomous robots in real world environments, they must be adapted to perform independently of background cues and camera motion effects. To address these challenges, we propose a new method that firstly generates generic action region proposals with good potential to locate one human action in unconstrained videos regardless of camera motion and then uses action proposals to extract and classify effective shape and motion features by a ConvNet framework. In a range of experiments, we demonstrate that by actively proposing action regions during both training and testing, state-of-the-art or better performance is achieved on benchmarks. We show the outperformance of our approach compared to the state-of-the-art in two new datasets; one emphasizes on irrelevant background, the other highlights the camera motion. We also validate our action recognition method in an abnormal behavior detection scenario to improve workplace safety. The results verify a higher success rate for our method due to the ability of our system to recognize human actions regardless of environment and camera motion.

In this paper, we introduce a novel acoustic-inertial navigation system (AINS) for Autonomous Underwater Vehicles (AUVs). We are aiming to reduce the cost and latency of current underwater navigation systems that typically employ high-accuracy and thus high-cost inertial sensors. In particular, the proposed approach efficiently fuses the acoustic observations from a 2D imaging sonar and the inertial measurements from a MEMS inertial measurement unit (IMU) within a tightly-coupled EKF framework, while having no need to keep the acoustic features in the state vector. As a result, the computational complexity of the proposed AINS is independent from the scale of the operating environment. Moreover, we develop an acoustic feature linear triangulation to provide accurate initial estimates for iterative solvers, and perform an in-depth observability analysis to investigate the effects of sensor motion on the triangulation. Additionally, since it is challenging to perform a priori sensor extrinsic calibration underwater, we advocate to calibrate IMU-sonar online. The proposed AINS has been validated extensively in Monte-Carlo simulations.

In this paper we consider the problem of controlling the dynamic behavior of a multi-robot system while interacting with the environment. In particular, we propose a general methodology that, by means of locally scaling inter-robot coupling relationships, leads to achieving a desired interactive behavior. The proposed method is shown to guarantee passivity preservation, which ensures a safe interaction. The performance of the proposed methodology is evaluated in simulation, over large-scale multi-robot systems.

The most prominent criterion for learning of manipulation skills is the optimization of task success, modeled as expected reward or probability of success. This is sensible if we only want to optimize a single controller. But if learned manipulation primitives are used as modules in a larger system, then it is also important that their generated sensor traces facilitate recognition of action-outcomes. Optimization solely for expected success of a primitive does not guarantee this. We demonstrate a simple example for optimization of actions towards observability, combined with optimization for expected success. Our experiment is a manipulation task with a soft manipulator, where an action primitive is learned such that its generated sensor trace helps a classifier to distinguish task success and task failure. The experimental results indicate that adding auxiliary forces to the original manipulation primitive can indeed facilitate outcome recognition for manipulation tasks.

Point-based stereo visual odometry systems typically estimate the camera motion by minimizing a cost function of the projection residuals between consecutive frames. Under some mild assumptions, such minimization is equivalent to maximizing the probability of the measured residuals given a certain pose change, for which a suitable model of the error distribution (sensor model) becomes of capital importance in order to obtain accurate results. This paper proposes a robust probabilistic model for projection errors, based on real world data. For that, we argue that projection distances follow Gamma distributions, and hence, the introduction of these models in a probabilistic formulation of the motion estimation process increases both precision and accuracy. Our approach has been validated through a series of experiments with both synthetic and real data, revealing an improvement in accuracy while not increasing the computational burden.

Traditional methods to achieve high localization accuracy with tactile sensors usually use a matrix of miniaturized individual sensors distributed on the area of interest. This approach usually comes at a price of increased complexity in fabrication and circuitry, and can be hard to adapt for non planar geometries. We propose to use low cost optic components mounted on the edges of the sensing area to measure how light traveling through an elastomer is affected by touch. Multiple light emitters and receivers provide us with a rich signal set that contains the necessary information to pinpoint both the location and depth of an indentation with high accuracy. We demonstrate sub-millimeter accuracy on location and depth on a 20mm by 20mm active sensing area. Our sensor provides high depth sensitivity as a result of two different modalities in how light is guided through our elastomer. This method results in a low cost, easy to manufacture sensor. We believe this approach can be adapted to cover non-planar surfaces, simplifying future integration in robot skin applications.

Future robotic missions to the poles of the Moon and Mercury will face challenges not encountered by current and prior planetary rover missions. Careful energy-aware spatiotemporal path planning will be required to accomplish mission objectives at high cadence under changing illumination conditions. With attention to landing site and time, such spatiotemporal path planning may enable extended missions on order of months that would not otherwise be possible. This work presents improvements in energy-aware spatiotemporal path planning for multiple waypoints to significantly reduce planning time. A time-compression technique is used to simplify planning in areas where changes occur infrequently. Consideration of end-goal reachability reduces the search space. Finally, heuristics that use pre-computation with static obstacles speed up the search.

This paper presents a method for constructing 3D maps of marine archaeological sites using deployments of Autonomous Underwater Vehicles (AUV) equipped with sonar and cameras. The method requires multiple AUV missions in which the first mission directs the AUV to conduct a high altitude lawnmower scan over the area to create a course bathymetry map using sonar. Subsequent AUV missions then direct the AUV to make low altitude fly-overs just above the wreck with the goal of obtaining camera images from multiple viewpoints of the wreck to enable offboard 3D mapping via photogrammetric reconstruction. This approach uses a coarse map generated after the first mission to construct AUV paths that attempt to maximize information gain, i.e. maximize the number of viewpoints of the wreck within a time limit. Presented is a motion planner derived from Rapidly-Exploring Random Trees (RRT) that have sampling strategies modified for this problem. Specifically, the random node selection and new node generation are designed to consider the kinematics of an AUV and the information gain associated with each flyover. Simulation results demonstrate improvements of up to 152% when these sampling strategies are used. Experiment results, involving deployments for mapping two known wrecks located along the coast of Malta, validate the system's ability to construct 3D maps and associated visualizations.

A significant challenge for unmanned aerial vehicles capable of flying long distances is planning in a wind field. Although there has been a plethora of work on the individual topics of planning long routes, smooth trajectory optimization and planning in a wind field, it is difficult for these methods to scale to solve the combined problem. In this paper, we address the problem of planning long, dynamically feasible, time-optimal trajectories in the presence of wind (which creates a moving reference frame). We present an algorithm, κITE, that elegantly decouples the joint trajectory optimization problem into individual path optimization in a fixed ground frame and a velocity profile optimization in a moving reference frame. The key idea is to derive a decoupling framework that guarantees feasibility of the final fused trajectory. Our results show that κITE is able to produce high-quality solutions for planning with a helicopter flying at speeds of 50 m/s, handling winds up to 20 m/s and missions over 200 km. We validate our approach with real-world experiments on a full-scale helicopter with a pilot in the loop. Our approach paves the way forward for autonomous systems to exhibit pilot-like behavior when flying missions in winds aloft.

A method is proposed to predict the shape of a paper model of a three piece brassiere cup, which consists of several cloth and wire parts. Currently, the shape of each part is determined by creating a paper model and then refining the model. This process is repeated until the desired shape is achieved. However, predicting the 3D shape with a simulation would improve design efficiency. As a model is made of paper, each part is assumed to be inextensible and its surface is represented as combination of developable surfaces. The potential energy of a cup and geometric constraints imposed on the cup are formulated by use of the normal curvature and the direction of a generatrix of each developable surface. Minimizing the potential energy under geometric constraints derives a stable shape of the cup model. The computed and measured cup shapes coincided qualitatively.

This work deals with a Conducting-Interpenetrating Polymer Network (C-IPN). The C-IPN exhibits very interesting and promising properties which can make it suitable for applications in robotics as a tool to perform tasks in the fields of manipulation, grasping or force measurement. It is known in the literature that such C-IPN may be actuated and bended to interact with other objects. Some of them can also be used as sensors to characterize the interaction. In this paper, we show that actuation and sensing can be performed at the same time. Moreover, we propose analytical models which can be useful for future work to process the C-IPN output and to control them. All results are verified with experimental data.

This paper studies the problem of image-based leader-follower formation control for mobile robots, where the controller is designed independently of the leader's motion. An adaptive control scheme, which is suitable for both omnidirectional and perspective cameras, is proposed. The proposed approach avoids the need for accurate calibration of the extrinsic parameters of the omnidirectional camera as well as the intrinsic and extrinsic parameters of perspective camera. Additionally, the coefficients of the plane where the feature point moves relative to the camera frame can be uncertain. These uncertain constant parameters are estimated using an adaptive estimator. Uniform Semi-global Practical Asymptotic Stability (USPAS) of the system is shown using the Lyapunov approach. Experimental results are presented to demonstrate the effectiveness of the proposed control scheme.

Soft compliant materials and novel actuation mechanisms ensure flexible motions and high adaptability for soft robots, but also increase the difficulty and complexity of constructing control systems. In this work, we provide an efficient control algorithm for a multi-segment extensible soft arm in 2D plane. The algorithm separate the inverse kinematics into two levels. The first level employs gradient descent to select optimized arm's pose (from task space to configuration space) according to designed cost functions. With consideration of viscoelasticity, the second level utilizes neural networks to figure out the pressures from each segment's pose (from configuration space to actuation space). In experiments with a physical prototype, the control accuracy and effectiveness are validated, where the control algorithm is further improved by an optional feedback strategy.

It is very challenging for a robotic gripper to achieve large reorientations with grasped objects without accidental object ejection. This paper presents a simple gripper that can repeatedly achieve large reorientations over π/2 rad through the kinematics of the hand-object system alone, without the use of high fidelity contact sensors, complex control of active finger surfaces, or highly actuated fingers. This gripper is the result of two kinematic parameter search optimizations connected in cascade. Besides the large range of reorientation attained, the obtained gripper also corresponds to a novel topology since ternary joints in the palm are presented. The in-hand planar reorientation capabilities of the proposed gripper are experimentally tested with success.