ICRA 2017 Papers

Robots that are operating for extended periods of time need to be able to deal with changes in their environment and represent them adequately in their maps. In this paper, we present a novel 3D reconstruction algorithm based on an extended Truncated Signed Distance Function (TSDF) that enables to continuously refine the static map while simultaneously obtaining 3D reconstructions of dynamic objects in the scene. This is a challenging problem because map updates happen incrementally and are often incomplete. Previous work typically performs change detection on point clouds, surfels or maps, which are not able to distinguish between unexplored and empty space. In contrast, our TSDF-based representation naturally contains this information and thus allows us to more robustly solve the scene differencing problem. We demonstrate the algorithms performance as part of a system for unsupervised object discovery and class recognition. We evaluated our algorithm on challenging datasets that we recorded over several days with RGB-D enabled tablets. To stimulate further research in this area, all of our datasets are publicly available3.

In this paper, we present the mechatronic design of our Tactile Omnidirectional Robot Manipulator (TOMM), which is a dual arm wheeled humanoid robot with 6DoF on each arm, 4 omnidirectional wheels and 2 switchable end-effectors (1 DoF grippers and 12 DoF Hands). The main feature of TOMM is its arms and hands which are covered with robot skin. We exploit the multi-modal tactile information of our robot skin to provide a rich tactile interaction system for robots. In particular, for the robot TOMM, we provide a general control framework, capable of modifying the dynamic behavior of the entire robot, e.g., producing compliance in a non-compliant system. We present the hardware, software and middleware components of the robot and provide a compendium of the base technologies deployed in it. Furthermore, we show some applications and results that we have obtained using this robot.

Planning under motion and observation uncertainties requires the solution of a stochastic control problem in the space of feedback policies. In this paper, by restricting the policy class to the linear feedback polices, we reduce the general (n2 + n)-dimensional belief space planning problem to an (n)-dimensional problem. As opposed to the previous literature that search in the space of open-loop optimal control policies, we obtain this reduction in the space of closed-loop policies by obtaining a Linear Quadratic Gaussian (LQG) design with the best nominal performance. Then, by taking the entire underlying trajectory of the LQG controller as the decision variable, we pose a coupled design of the trajectory and estimator (while keeping the design of the controller separate) as a NonLinear Program (NLP) that can be solved by a general NLP solver. We prove that under a first-order approximation and a careful usage of the separation principle, our approximations are valid. We provide an analysis on the existing major belief space planning methods and show that our algorithm keeps the lowest computational burden while searching in the policy space. Finally, we extend our solution to contain general state and control constraints. Our simulation results support our design.

We present a general approach for online learning and optimal control of manipulation tasks in a supervisory teleoperation context, targeted to underwater remotely operated vehicles (ROVs). We use an online Bayesian nonparametric learning algorithm to build models of manipulation motions as task-parametrized hidden semi-Markov models (TP-HSMM) that capture the spatiotemporal characteristics of demonstrated motions in a probabilistic representation. Motions are then executed autonomously using an optimal controller, namely a model predictive control (MPC) approach in a receding horizon fashion. This way the remote system locally closes a high-frequency control loop that robustly handles noise and dynamically changing environments. Our system automates common and recurring tasks, allowing the operator to focus only on the tasks that genuinely require human intervention. We demonstrate how our solution can be used for a hot-stabbing motion in an underwater teleoperation scenario. We evaluate the performance of the system over multiple trials and compare with a state-of-the-art approach. We report that our approach generalizes well with only a few demonstrations, accurately performs the learned task and adapts online to dynamically changing task conditions.

There has been a recent paradigm shift in robotics to data-driven learning for planning and control. Due to large number of experiences required for training, most of these approaches use a self-supervised paradigm: using sensors to measure success/failure. However, in most cases, these sensors provide weak supervision at best. In this work, we propose an adversarial learning framework that pits an adversary against the robot learning the task. In an effort to defeat the adversary, the original robot learns to perform the task with more robustness leading to overall improved performance. We show that this adversarial framework forces the robot to learn a better grasping model in order to overcome the adversary. By grasping 82% of presented novel objects compared to 68% without an adversary, we demonstrate the utility of creating adversaries. We also demonstrate via experiments that having robots in adversarial setting might be a better learning strategy as compared to having collaborative multiple robots. For supplementary video see: youtu.be/QfK3Bqhc6Sk

Supervised learning is used to build a control policy for robust, stable, dynamic walking of an underactuated bipedal robot. The training and testing sets consist of controllers based on a full dynamic model, virtual constraints, and parameter optimization to meet torque limits, friction cone, and environmental conditions. The controllers are designed to induce locally exponentially stable periodic walking gaits at various speeds, both forward and backward, and for various constant ground slopes. They are also designed to induce aperiodic gaits that transition among a subset of the periodic gaits in a fixed number of steps. In experiments, the learned policy allows a 3D bipedal robot to recover from a significant kick. It also enables the robot to walk down a 22 degree slope and walk on sinusoidally varying terrain, all without using a camera. During the development of these results, it is demonstrated that supervised learning of locally exponentially stable controllers can result in a loss of stability and a means to avoid this is suggested.

Inspired from Hogan's myoelectric processor in 1980, this study presents a stochastic sEMG processing method to estimate the muscle activation level for manipulator control. Hogan's previous study showed the feasibility to estimate the muscle activation level with multi-channel sEMG under static force condition. However, it is difficult to continuously estimate muscle activation during dynamic contraction because of the nonlinear effects by the time-varying nature of sEMG. To enhance the performance of high SNR and rapid response in force-varying contraction, we propose a new method with statistical analysis extended from a whitening method of Hogan's study. The signals from eight sEMG channels were used to estimate the muscle activation level during isometric force-varying contractions. Experimentally, a two-DoF manipulator was controlled by input signals from the estimated muscle activation signal.

Safe path planning is a crucial component in autonomous robotics. The many approaches to find a collision free path can be categorically divided into trajectory optimizers and sampling-based methods. When planning using occupancy maps, the sampling-based approach is the prevalent method. The main drawback of such techniques is that the reasoning about the expected cost of a plan is limited to the search heuristic used by each method. We introduce a novel planning method based on trajectory optimization to plan safe and efficient paths in continuous occupancy maps. We extend the expressiveness of the state-of-the-art functional gradient optimization methods by devising a stochastic gradient update rule to optimize a path represented as a Gaussian process. This approach avoids the need to commit to a specific resolution of the path representation, whether spatial or parametric. We utilize a continuous occupancy map representation in order to define our optimization objective, which enables fast computation of occupancy gradients. We show that this approach is essential in order to ensure convergence to the optimal path, and present results and comparisons to other planning methods in both simulation and with real laser data. The experiments demonstrate the benefits of using this technique when planning for safe and efficient paths in continuous occupancy maps.

In small mobile robots, decreasing the number of actuators is usually desirable to reduce the size and weight of the robot, but it is usually at the expense of the robot's degree of freedom (DOF). This work presents the development and preliminary experimental testing of a novel Legged Piezoelectric Miniature Robot (LPMR) driven only by a single piezoelectric unimorph actuator and yet fully capable of being maneuvered to move forward, turn right, or turn left. The underactuated motion is achieved by exploiting the bending vibration modes disparity of the piezoelectric actuator at different driving frequencies and designing specific positions of the robot's legs to generate a differential-drive-like mechanism. The speed of the robot can be controlled through regulating the magnitude of the applied voltage. The proposed underactuated system is experimentally verified and a preliminary characterization of the LPMR in terms of its forward and turning speed versus applied voltage and payload is investigated and reported.

Although supervisor simplification in the framework of automated manufacturing systems has been studied in many literatures, there is still an intense demand for essential and general techniques. Basically, supervisors can be synthesized by specifications which are expressed by generalized mutual exclusion constraints (GMECs). In this paper, we propose static and dynamic partitions on GMECs to remove redundant ones while retain necessary ones. These two partitions are distinctly different in the utilization of system information. Static partition separates inequalities into independent and dependent ones without doing any structure analysis. While dynamic partition divides inequalities into active and inactive ones necessarily with the aid of system information. Nevertheless, mathematical analysis shows that statically dependent inequalities are essentially dynamically inactive while dynamically active ones are substantially statically independent. Thus, these partitions are contradictory in theory whereas compatible in practice. Since static partition is more general while its dynamic counterpart is more precise, they together complementarily explain many simplification principles. Furthermore, they are applicable to both ordinary and general systems. Experimental results show the effectiveness and efficiency of supervisor simplification based on static and dynamic inequality partitions.

Based on the liner inverted pendulum model (LIPM) for the dynamics of biped humanoid robots, an analytical method for computing stability regions relevant to standing balance of the biped humanoid robots is introduced in this paper. More precisely, two types of the stability regions are discussed in this paper with the consideration of that the zero moment point (ZMP) should be located in the supporting region to guarantee stable standing of the biped humanoid robots. First, assuming no external disturbances affecting the motion of the biped humanoid robots, the set of the initial values of the center of mass (CoM) position and velocity with which the location of the ZMP is limited to be inside the supporting region can be explicitly obtained by solving a finite number of linear inequalities. Second, two admissible sets of external force disturbances (impulse and finite energy) with which the ZMP does not deviate from the supporting region are characterized by solving finite number of linear inequalities or the discrete-time Lyapunov equation, respectively. The validity and effectiveness of the analytical method proposed in this paper are verified through a simulation result.

This paper presents quantification of multidimensional ankle stability in relation to mechanical environments having different levels of stability. This study, for the first time, explores the range of stiffness-defined haptic environments over which young healthy individuals can maintain stability despite aggressive perturbation. Ankle stability was quantified in 2 degree-of-freedom (DOF) of the ankle, in both the sagittal and frontal planes. Importantly, the magnitude of negative environmental stiffness that the subjects could maintain stability is 4 times as great in the sagittal plane as in the frontal plane. In addition to managing a wider range of unstable environments in the sagittal plane, subjects were also more efficient at regaining stability after perturbation and less sensitive to changes in the environmental stiffness. Outcomes of this study would be beneficial to the design and control of robots physically interacting with human lower extremities, such as lower-limb exoskeletons and powered ankle-foot orthoses.

We propose a control law for stabilization of a quadrotor-load system, and provide conditions on the control law's gains that guarantee exponential stability of the equilibrium. The system is composed of a load and an unmanned aerial vehicle (UAV) attached to each other by a cable of fixed length, which behaves as a rigid link under tensile forces; and the control input is composed of a three dimensional force requested to the UAV, which the UAV provides with or without delay. Given the proposed control law, we analyze the stability of the equilibrium in two separate parts. In the first, the system is modeled assuming that the UAV provides the requested control input without delay, and we verify that the equilibrium is exponentially stable. In the second part, the UAV is modeled as possessing an attitude inner loop, and we provide a lower bound on the attitude gain for which exponential stability of the equilibrium is preserved. An integral action term is also included in the control law, which compensates for battery drainage or model mismatches, such as an unknown load mass. We present experiments for different scenarios that demonstrate and validate the robustness of the proposed control law.

Provides an abstract for each of the presentations and may include a brief professional biography of each presenter. The complete presentations were not made available for publication as part of the conference proceedings.

In this paper, we propose an epoch-making soft sheet actuator called “Wavy-sheet”. Inspired by gastropod's locomotion, Wavy-sheet can generate continuous traveling waves on the whole soft body. It aims to be applied to a mobile soft mat capable of moving and transporting without damaging the object and the ground. The actuator, driven by pneumatics, is mainly composed of a couple of flexible rubber tubes and fabrics. The advantages are: i) many traveling waves can be generated by just three tubes, ii) the whole structure can adapt its own shape to the outer environment passively, and iii) only 10 mm in thickness and can generate waves with larger than 10mm in amplitude. In this paper, first, we describe the basic concept of Wavy-sheet, and then show the configuration and the principle of wave propagation. Next, fabrication methods are illustrated and the design methods are addressed. By using a prototype actuator, several experiments are conducted. Finally, we verify the effectiveness of the proposed actuator and its design methods.

This paper extends the longitudinal walking controller of biped robots proposed by one of the authors to walking along a smooth curved path. The previous controller defined in a fixed inertial frame is redesigned with respect to a moving frame attached to the center of mass (COM) of the robot in order to achieve walking along an arc. As a result, it is found that a desired position of the zero-moment point (ZMP) can be determined only from the local curvature of the referential path besides the state of the robot with respect to the moving frame. Thus, it is directly applicable to walking along an arbitrary smooth path including straight walk. In addition to the extension of the ZMP manipulation, techniques of foot control consistent with COM movement and automatic update of the referential COM position are fully extended to curved walk. The idea was examined through computer simulations to show responsiveness to updates of motion references and robustness against external disturbances.

The accuracy of reconfigurable robot manipulators is a critical aspect which prevents their diffused industrial adoption. This work presents a novel model for designing joint motion profiles, particularly suitable for the motion planning of modular robots with high accuracy requirements. The model generates smooth motion profiles, without sacrificing execution time and taking into account the different characteristics of each joint and the requirements of each production task. The proposed method has been tested across the most performing motion planning approaches found in the literature, providing up to 39% faster jerk-bounded trajectories. Moreover, the model is flexible to the generation of adapted trajectories when degrading phenomena occur over the time.

In robotics, it is often practically and theoretically convenient to design motion planners for approximate low-order (e.g., position-or velocity-controlled) robot models first, and then adapt such reference planners to more accurate high-order (e.g., force/torque-controlled) robot models. In this paper, we introduce a novel provably correct approach to extend the applicability of low-order feedback motion planners to high-order robot models, while retaining stability and collision avoidance properties, as well as enforcing additional constraints that are specific to the high-order models. Our smooth extension framework leverages the idea of reference governors to separate the issues of stability and constraint satisfaction, affording a bidirectionally coupled robot-governor system where the robot ensures stability with respect to the governor and the governor enforces state (e.g., collision avoidance) and control (e.g., actuator limits) constraints. We demonstrate example applications of our framework for augmenting path planners and vector field planners to the second-order robot dynamics.

This paper presents a proof of concept to calibrate iCub's skin using vacuum bags. The main idea of the method consists of inserting the skin, made of distributed capacitive sensors, in a vacuum bag and then decreasing the pressure in the bag to create a uniform pressure distribution on the skin surface. The capacitance changes of each sensor were then related to the applied differential pressure using a least square fitting with a fifth order polynomial model. After calibration, integration of the pressure distribution over the skin geometry provides us with the net normal force applied to the skin. Experiments were conducted using the forearm skin of the iCub humanoid robot and the calibration results were validated using standard weights. The validation results indicate acceptable average errors in force prediction.