This paper addresses the recognition of daily-life objects by a robot equipped with tactile sensors. The main contribution is a deep learning framework that can recognize objects already touched as well as objects never touched before. To this end, we train a deconvolutional neural network that generates synthetic tactile data for novel classes. Then, we use both these synthetic data and the real data collected by touching objects, to train a convolutional neural network to recognize both known (trained) objects and novel objects. Furthermore, we propose a method for integrating newly encountered data into novel classes. Finally, we evaluate the framework using the largest available dataset of tactile objects descriptions.

This article presents a new kinematic control paradigm for redundant robots based on optimization. The general approach takes into account convex objective functions with inequality constraints and a specific equality constraint resulting from a Lyapunov function, which ensures closed-loop stability by design. Furthermore, we tackle an important particular case by using a convex combination of quadratic and l1-norm objective functions, making possible for the designer to choose different degrees of sparseness and smoothness in the control inputs. We provide a pseudoanalytical solution to this optimization problem and validate the approach by controlling the center of mass of the humanoid robot HOAP3.

The paper describes a control framework using a capacitive skin for advanced and safe human-robot interactions. The skin consists of high density capacitive sensors which are able to detect objects at close distance. We propose a multitask kinematic control scheme which manages at once the control task of the end-effector and the obstacle avoidance task by using the capacitive skin measurements. To solve the multiobjective problem, a Weight-Prioritized solution based on a QP formalism is adopted. The robustness of the kinematic controller is enhanced by adding a local joint reconfiguration when the robot moves close to a singularity. Experimental results on manipulator arm equipped with the capacitive sensors are presented in realistic situations of human-robot interactions.

This study proposes a visuo-tactile Zero-Shot object recognition framework. The proposed framework recognizes a set of novel objects for which no tactile or visual training data are available. It uses visuo-tactile training data collected from known objects to recognize the novel ones, given their attributes. This framework extends the haptic Zero-Shot Learning framework that we proposed in [1] with vision, which enables a multimodal recognition system. In our test with the PHAC-2 dataset, the system was able to get a recognition accuracy of 72% among 6 objects that were never touched or seen during the training phase.

Knowledge of the mass and inertial parameters of a humanoid robot is crucial for the development of model-based controller and motion planning in dynamics situation. Parameters are usually provided from Computer Aided Design (CAD) data and thus inaccurate specially if the robot is modified over time. In this paper, a practical method consisting of hanging a humanoid robot to a fix force sensor to perform its dynamic identification is proposed. This allows, contrary to the literature, to generate very exciting and dynamic motions to identify most of the elements of the inertia tensors in a reduced amount of time. This procedure transforms an instable floating base legged humanoid robot to a safe fix base tree structure robot which makes easier to generate optimal exciting motions. Because of a better excitation the overall trajectory lasts for less than a minute. The method was experimentally validated with a HOAP3 humanoid robot and using a 6-axis force sensor. A reduction of 3 times in average of the RMS difference between measured external reaction forces and moments and their estimates from CAD data was obtained with a single minute of optimal exciting motions.

To make production lines more flexible, dual-arm robots are good candidates to be deployed in autonomous assembly units. In this paper, we propose a sparse kinematic control strategy, that minimizes the number of joints actuated for a coordinated task between two arms. The control strategy is based on a hierarchical sparse QP architecture. We present experimental results that highlight the capability of this architecture to produce sparser motions (for an assembly task) than those obtained with standard controllers.

Deforming a cable to a desired (reachable) shape is a trivial task for a human to do without even knowing the internal dynamics of the cable. This paper proposes a framework for cable shapes manipulation with multiple robot manipulators. The shape is parameterized by a Fourier series. A local deformation model of the cable is estimated on-line with the shape parameters. Using the deformation model, a velocity control law is applied on the robot to deform the cable into the desired shape. Experiments on a dual-arm manipulator are conducted to validate the framework.

Knowledge of the mass and inertial parameters of a humanoid robot or a human being is crucial for the development of model-based control, as well as for monitoring the rehabilitation process. These parameters are also important for obtaining realistic simulations in the field of motion planning and human motor control. For robots, they are often provided by computer-aided design data, while averaged anthropometric table values are often used for human subjects. The unit/subject-specific inertial parameters can be identified by using the external wrench caused by the ground reaction. However, the identification accuracy intrinsically depends on the excitation properties of the recorded motion. In this paper, a new method for obtaining optimal excitation motions is proposed. This method is based on the identification model of legged systems and on optimization processes to generate excitation motions while handling mechanical constraints. A pragmatic decomposition of this problem, the use of a new excitation criterion, and a quadratic program to identify inertial parameters are proposed. The method has been experimentally validated onto an HOAP-3 humanoid robot and with one human subject.

For intuitive human-robot collaboration, the robot must quickly adapt to the human behavior. To this end, we propose a multimodal sensor-based control framework, enabling a robot to recognize human intention, and consequently adapt its control strategy. Our approach is marker-less, relies on a Kinect and on an on-board camera, and is based on a unified task formalism. Moreover, we validate it in a mock-up industrial scenario, where human and robot must collaborate to insert screws in a flank.

In this paper, we propose a control scheme that allows a humanoid robot to perform a transportation task jointly with a human partner. From the study of how human dyads achieve such a task, we have developed a control law for physical interaction that unifies standalone and collaborative (leader and follower) modes for trajectory-based tasks. We present it in the case of a linear impedance controller but it can be generalized to more complex impedances. Desired trajectories are decomposed into sequences of elementary motion primitives. We implemented this model with a Finite State Machine associated with a reactive pattern generator. First experiments conducted on a real HRP-2 humanoid robot assess the overall approach.

In this paper, we propose a control scheme that allows a humanoid robot to perform a complex transportation scenario jointly with a human partner. At first, the robot guesses the human partner's intentions to proactively participate to the task. In a second phase, the human-robot dyad switches roles: the robot takes over the leadership of the task to complete the scenario. During this last phase, the robot is remotely controlled with a joystick. The scenario is realized on a real HRP-2 humanoid robot to assess the overall approach.

In this paper, we present a new algorithm for the alignment of two 3D scans. The approach uses a region-based matching technique. We make no assumptions about the initial positions of the scans. Regions are described by a probability density function (pdf) computed from low dimensional surface descriptors (curvature or normal cone). The algorithm allows registering directly raw noisy data, possibly with the presence of outliers, without any pre-processing, such as filtering, denoising, or reconstruction. Region correspondence is found using similarity function based on the comparison of regions pdf and under geometry constraints. Results on raw scan data sets are presented to illustrate and evaluate the algorithm.

We present a Graph-based method for low-level segmentation of unfiltered 3D data. The core of this approach is based on the construction of a local neighborhood structure and its recursive subdivision. The Minimum Spanning Tree (MST) is the graph support used to measure the attribute variation through the region. The subdivision criterion relies on the evidence for a boundary between two partitions, which is detected through MST edge analysis. Although our algorithm converges to a local minimum, our experiments show that it produces segments that satisfy global properties. We assume that the 3D image is composed of regions homogeneous according to some criterion (color, curvature, etc.), but no assumption about noise, nor spatial repartition/shape of the regions or points is made. Robustness is achieved by choosing the appropriate neighborhood and the analysis of noise impact on the MST construction. We demonstrate the performance of our algorithm with experimental results on real images.

The ICP (iterative closest point) algorithm remains a very popular method for the registration of 3D data sets, when an initial guess of the relative pose between them is known. The purpose of the work presented in this paper is to improve performance of classical ICP. We address, here, the problem of pair-wise registration of color range images. Many variants of ICP have been proposed for the registration of 3D data sets. However, there are only a few solutions dealing with color range images. In this paper, we have adapted some variants of ICP to take into account color information in the closest point research. Two approaches of color data integration have been evaluated with real 3D/color data sets

The paper presents a technical solution for manually making replication of sculptures in real material from a numerical representation of the original work. This solution is based on a human centered system that provides an alternative to current techniques used in the art domain. This paper is focused on the presentation of the main capabilities of the system: functional and hardware characteristics, techniques for computer aided gesture.

The paper suggests a representation of swept volumes generated by the motion of deformable objects. This representation is based on the topological properties of n-dimensional manifolds. The algorithm developed for the evaluation of the boundary exploits these properties in such a way that the n-dimensional object (swept volume) is generated only from the (n-1)-dimensional components resulting from a family of deformations. The deformation, analytically expressed, integrates physical properties, especially the nonlinearities of deformable objects. The representation is applied to the modeling of the deformations of a milling cutter being fed into a workpiece and it constitutes an efficient tool for NC verification. Many other applications may be concerned by this work, especially in robotics (modeling of flexible robots, collision detection) or in mechanical engineering.