ICRA 2019 论文列表

We consider a problem minimizing the total duration of accomplishing missions performed by heterogeneous vehicles. The problem respects constraints related to vehicles' capabilities and energy capacities. The goal is to determine the best routes of each vehicle deployed by choosing which waypoints to pass and which observations to perform. Each vehicle has a particular distance matrix and a limited energy. In order to provide high quality solutions within reasonable computational time, two decomposition-based approximate methods were implemented: (i) the Multiphase heuristic, and (ii) the Two-Phase iterative heuristic. The performance of the methods is evaluated against the Branch-and-Cut algorithm using generated instances.

Planetary exploration rovers are expensive, weight constrained, and cannot be serviced once deployed. Here, we explore one way to increase their capabilities while avoiding the cost, mass, and complexity leading to these issues. We propose to re-use the large wheel actuators for trenching and other digging operations, which will enable a range of missions such as sampling deeper layers of soil. We present a new, closed-form model of the soil displaced by an angled, spinning wheel to analyze the trenching potential of a driving strategy and inform the control of the wheel. The model is demonstrated with single wheel experiments under different driving conditions. The model suggests: that a deep trench does not require large tractive efforts; that the shape of the trench can be controlled; and that a rear wheel has a lower risk of entrapment when trenching than a front wheel. Ultimately this model could be used in a nonprehensile manipulation planning or learning algorithm to enable autonomous trenching.

Intrinsic Tactile Sensing (ITS) is a well-established technique, relying on force/torque and geometric surface description to find contact centroids. The method works well for rigid surfaces. However, finding a solution for deformable surfaces is an open issue. This work presents two solutions to extend ITS to deformable surfaces, relying on force-deformation characteristics of the surface under exploration: (i) a closed-form approach that calculates the contact centroid using standard ITS, but on a shrunk geometry approximating the deformed surface; (ii) an iterative procedure that takes into account soft surface deformation, and force/torque equilibrium to minimize a cost function. We have tested both using ellipsoid silicone specimens, with different softness levels and indented along different directions. Both linear and quadratic fitting for the force-indentation behavior were employed. The two methods have distinct advantages and limitations. However, a combination of two methods, using one to produce the initial guess for the other, turns out to be very effective. Indeed, in our validation this solution showed convergence under 1ms, attaining errors lower than 1 mm. The proposed approaches were implemented in a ROS-based toolbox, integrating both solutions.

This paper presents a fully-integrated soft robotic glove with multi-articular textile actuators, custom soft sensors, and an intuitive state machine intent detection controller. We demonstrate that the pressurized actuators can generate motion and force comparable to natural human fingers through bench-top testing. We apply textile-elastomer capacitive sensors to the glove to track finger flexion via strain and detect contact with objects via force. Intuitive user control is achieved via a state machine controller based on signals from the integrated sensors to detect relative changes in hand-object interactions. Results from an initial evaluation with 3 participants with spinal cord injury (SCI), of varied injury levels and years since injury, wearing and controlling the glove show an average of 87% improvement in grasping force, and improvements in functional assessments for participants with recent injuries. A significant variation in response suggests further investigation is required to understand the adaptation needed across different injury levels and durations since injury. Additionally, we evaluate the controller and find an average of 3 seconds from user initiations to completed grasps, and 10% inadvertent grasp triggers and no false releases when objects are held.

The design of robotic grippers requires the accomplishment of several contrasting requirements. Research in under actuated soft hands is a lively topic, with several potentialities and challenges. Soft hands are simple, robust and able of adapting to uncertain environment and operative conditions, however their intrinsic compliance and underactuation reduce control capabilities and precision. Recent studies attempted to compensate this limitation by wisely exploiting environmental constraints and considering them as supports to accomplish the task rather than obstacle to avoid. The development of grasp primitives taking into account environment features leaded to interesting and encouraging results. In this paper, we propose to embed on the hand the positive aspects of studies on environmental constraints exploitation. We present a modular under actuated soft hand in which we added a scoop as a feature of the palm, which simplify object grasping. The scoop allows to grasp objects in narrow spaces, augments the possible contact areas, allows to obtain more robust grasps, with lower forces. The paper illustrates the main design principles, a prototype and experimental results.

Interactions between robots and their environment give rise to external wrenches acting on the robot structure. The estimation of the resulting torques in the joints is fundamental in human-robot interaction to detect/identify collisions and perform suitable reaction strategies. Other applications may require to use the estimation for compensating the effects of the external torques within the control loop. The well-established momentum observer, which relies on proprioceptive sensors only, is usually used for these purposes. In this work, the momentum dynamics is used to derive new observers. While the classic momentum observer provides a first-order filtered version of the external torques, here a (theoretically) finite-time convergence is achieved. Simulations and experiments are used to validate the performance of the proposed methods.

Pneumatic artificial muscles (PAMs) have been researched for applications in powered exoskeletons, orthosis and robotics. Their high force to mass ratio, low cost and inherent compliance are particularly advantageous for systems requiring physical interaction with humans.Sleeve PAMs, which introduce an internal structure to the actuator, offer improved force capacity, contraction ratio, efficiency and operating bandwidth. In this paper sleeve PAMs are applied to a popular muscle configuration; that of a joint operated antagonistically by two muscles. It is shown that the sleeve PAM can increases the range of joint rotation by 14% or load capacity by over 50% of that of a comparable joint actuated with traditional PAMs, depending on the joint configuration. The stiffness of joints actuated with both PAM types is also studied, particularly the case of closed system operation (mass of air in the PAMs is constant), where the reduced volume of the sleeve PAM significantly increases the observed stiffness. Finally energy consumption is considered, showing substantial savings in the case of joints actuated with sleeve PAMs.

We propose a learning framework, named Multi-Coordinate Cost Balancing (MCCB), to address the problem of acquiring point-to-point movement skills from demonstrations. MCCB encodes demonstrations simultaneously in multiple differential coordinates that specify local geometric properties. MCCB generates reproductions by solving a convex optimization problem with a multi-coordinate cost function and linear constraints on the reproductions, such as initial, target, and via points. Further, since the relative importance of each coordinate system in the cost function might be unknown for a given skill, MCCB learns optimal weighting factors that balance the cost function. We demonstrate the effectiveness of MCCB via detailed experiments conducted on one handwriting dataset and three complex skill datasets.

Transcatheter aortic valve implantation (TAVI) is a minimally invasive surgical technique to treat aortic heart valve diseases. According to current clinical guidelines, the implanted prosthetic valve replacing the native one is selected based on pre-operative size assessment of the aortic annulus through different imaging techniques. That very often leads to suboptimal device selection resulting in major complications, such as prosthetic valve leakage or interruption of the cardiac electrical signal. In this paper, we propose a new, intra-operative approach to determine the diameter of theaortic annulus exploiting intra-balloon pressure and volume data, acquired from a robotised valvuloplasty balloon catheter. An inflation device, capable of collecting real-time intra-balloon pressure and volume data, was designed and interfaced with a commercially available valvuloplasty balloon catheter. A sizing algorithm allowing to precisely estimate the annular diameter was integrated. The algorithm relies on a characterised analytical model of the balloon free inflation and an iterative method based on linear regression. In vitro tests were performed on idealised aortic phantoms. Experimental results show that pressure-volume data can be used to determine annular diameters bigger than the unstretched diameter of the balloon catheter. For these cases, the proposed approach exhibited good precision (maximum average error 0.93%) and good repeatability (maximum standard deviation ±0.11 mm).

This paper addresses for the first time the singularity analysis of cable-driven parallel robot (CDPR) with sagging cables using the Irvine model. We present the mathematical framework of singularity analysis of CDPR using this cable model. We then show that, besides a cable model representation singularity, both the inverse and forward kinematics (IK and FK) have a singularity type, called parallel robot singularity, which correspond to the singularity of an equivalent parallel robot with rigid legs. We then show that both the IK and FK have also full singularities, that are not parallel robot singularity and are obtained when two of the IK or FK solution branches intersect. IK singularity will usually lie on the border of the CDPR workspace. We then exhibit an algorithm that allow one to prove that a singularity exist in the neighborhood of a given pose and to estimate its location with an arbitrary accuracy. Examples are provided for parallel robot, IK and FK singularities. However we have not been able to determine examples of combined singularity where both the IK and FK are singular (besides parallel robot singularity).

In this paper, we propose a method for selecting the optimal footholds for legged systems. The goal of the proposed method is to find the best foothold for the swing leg on a local elevation map. First, we evaluate the geometrical characteristics of each cell on the elevation map, checks kinematic constraints and collisions. Then, we apply the Convolutional Neural Network to learn the relationship between the local elevation map and the quality of potential footholds. During execution time, the controller obtains the qualitative measurement of each potential foothold from the neural model. This method evaluates hundreds of potential footholds and checks multiple constraints in a single step which takes 10 ms on a standard computer without GPU. The experiments were carried out on a quadruped robot walking over rough terrain in both simulation and real robotic platforms.

Recent developments in multi-agent imitation learning have shown promising results for modeling the behavior of human drivers. However, it is challenging to capture emergent traffic behaviors that are observed in real-world datasets. Such behaviors arise due to the many local interactions between agents that are not commonly accounted for in imitation learning. This paper proposes Reward Augmented Imitation Learning (RAIL), which integrates reward augmentation into the multi-agent imitation learning framework and allows the designer to specify prior knowledge in a principled fashion. We prove that convergence guarantees for the imitation learning process are preserved under the application of reward augmentation. This method is validated in a driving scenario, where an entire traffic scene is controlled by driving policies learned using our proposed algorithm. Further, we demonstrate improved performance in comparison to traditional imitation learning algorithms both in terms of the local actions of a single agent and the behavior of emergent properties in complex, multi-agent settings.

This article considers the problem of near-optimal trajectory generation for autonomous vehicles on highways. The goal is to select a predictive reference trajectory in the free evolution space, while avoiding both generating a pre-calculated set of candidate trajectories and decoupling path and velocity optimizations. Moreover, this trajectory aims at optimizing a decision process based on multi-criteria functions, which are not straightforward to design and can have a blackbox formulation. The main idea of this article is to use the decision evaluation function in the trajectory generator with a Simulated Annealing (SA) approach. The parameters of a sigmoid trajectory are optimized within Non-Collision Nominal Intervals (NCNI), which are defined as collision-free intervals under nominal conditions using a velocity-space representation.

Learning robot tasks or controllers using deep reinforcement learning has been proven effective in simulations. Learning in simulation has several advantages. For example, one can fully control the simulated environment, including halting motions while performing computations. Another advantage when robots are involved, is that the amount of time a robot is occupied learning a task-rather than being productive-can be reduced by transferring the learned task to the real robot. Transfer learning requires some amount of fine-tuning on the real robot. For tasks which involve complex (non-linear) dynamics, the fine-tuning itself may take a substantial amount of time. In order to reduce the amount of fine-tuning we propose to learn robustified controllers in simulation. Robustified controllers are learned by exploiting the ability to change simulation parameters (both appearance and dynamics) for successive training episodes. An additional benefit for this approach is that it alleviates the precise determination of physics parameters for the simulator, which is a non-trivial task. We demonstrate our proposed approach on a real setup in which a robot aims to solve a maze game, which involves complex dynamics due to static friction and potentially large accelerations. We show that the amount of fine-tuning in transfer learning for a robustified controller is substantially reduced compared to a non-robustified controller.

Manipulation tasks often require robots to be continuously in contact with an object. Therefore tactile perception systems need to handle continuous contact data. Shear deformation causes the tactile sensor to output path-dependent readings in contrast to discrete contact readings. As such, in some continuous-contact tasks, sliding can be regarded as a disturbance over the sensor signal. Here we present a shear-invariant perception method based on principal component analysis (PCA) which outputs the required information about the environment despite sliding motion. A compliant tactile sensor (the TacTip) is used to investigate continuous tactile contact. First, we evaluate the method offline using test data collected whilst the sensor slides over an edge. Then, the method is used within a contour-following task applied to 6 objects with varying curvatures; all contours are successfully traced. The method demonstrates generalisation capabilities and could underlie a more sophisticated controller for challenging manipulation or exploration tasks in unstructured environments.

Seamless communication of desired motions and goals is essential for enabling effective physical human-robot collaboration. In such cases, muscle activity measured via surface electromyography (EMG) can provide insight into a person's intentions while minimally distracting from the task. The presented system uses two muscle signals to create a control framework for team lifting tasks in which a human and robot lift an object together. A continuous setpoint algorithm uses biceps activity to estimate changes in the user's hand height, and also allows the user to explicitly adjust the robot by stiffening or relaxing their arm. In addition to this pipeline, a neural network trained only on previous users classifies biceps and triceps activity to detect up or down gestures on a rolling basis; this enables finer control over the robot and expands the feasible workspace. The resulting system is evaluated by 10 untrained subjects performing a variety of team lifting and assembly tasks with rigid and flexible objects.

Soft robotic systems offer benefits over traditional rigid systems through reduced contact trauma with soft tissues and by enabling access through tortuous paths in minimally invasive surgery. However, the inherent deformability of soft robots places both a greater onus on accurate modelling of their shape, and greater challenges in realising intraoperative shape sensing. Herein we present a proprioceptive (self-sensing) soft actuator, with an electrically conductive working fluid. Electrical impedance measurements from up to six electrodes enabled tomographic reconstructions using Electrical Impedance Tomography (EIT). A new Frequency Division Multiplexed (FDM) EIT system was developed capable of measurements of 66 dB SNR with 20 ms temporal resolution. The concept was examined in two two-degree-of-freedom designs: a hydraulic hinged actuator and a pneumatic finger actuator with hydraulic beams. Both cases demonstrated that impedance measurements could be used to infer shape changes, and EIT images reconstructed during actuation showed distinct patterns with respect to each degree of freedom (DOF). Whilst there was some mechanical hysteresis observed, the repeatability of the measurements and resultant images was high. The results show the potential of FDM-EIT as a low-cost, low profile shape sensor in soft robots.

The particularly compliant and adaptable properties of soft robotic systems and structures offer enormous potential for use in unpredictable environments as well as for safe and interactive work in human environments. Therefore, new approaches for the design of soft robotic systems are constantly being introduced in this still emerging field of research. Through the use of additive manufacturing methods, it is possible to design systems specifically for an individual task. In this paper we present an approach for an automated design process for monolithic soft robotic structures that can assume pre-defined end poses. The idea thereby is to design simple individualizable systems that are 3D-printable and require a minimum number of actuators. Using the automated design process, we could already generate soft robotic systems for different applications which show promising properties.

The paper presents a multiple task-priority inverse kinematics algorithm for a dual-arm aerial manipulator. Both tasks defined as equality constraints and inequality constraints are handled by means of a singularity robust method based on the Null-Space based Behavioral control. The proposed schema is constituted by the inverse kinematics control, that receives the desired behavior of the system and outputs the reference values for the motion variables, i.e. the UAV pose and the arm joints position, and a motion control, that computes the vehicle thrusts and the joint torques. The method has been experimentally validated on a system composed by an underactuated aerial hexarotor vehicle equipped with two lightweight 4-DOF manipulators, involved in operations requiring the coordination of the two arms and the vehicle.