IROS 2016 Papers

This work presents the development of iT-Knee, a novel modular knee exoskeleton that makes use of a torque transmission interface to deliver pure assistive torque to the knee articulation, with the objective to act as rehabilitation tool, or a single joint add-on exoskeleton device for power augmentation. The specific kinematics features employed by iT-Knee can accommodate not only the translational of the knee instantaneous center of rotation in the sagittal plane and on its orthogonal direction, but also the knee varus/valgus angle as well as the internal/external rotation tibia movements delivering pure torque only around the flexion/extension movement. This functionality results in superior transparency and comfortability which allow prolonged period of use without generating significant discomfort. The iT-Knee kinematics, its mechanism, implementation, and actuation are introduced. The Self-aligning feature combined with a fast lock/release mounting system on the human leg shortens the setup time required to wear it, increasing its user-friendly perception. iT-Knee uses a torque controlled actuation unit to generate torque around the knee flexion extension enabling the implementation of several interaction control schemes that combine assistive functionality on demand with high back drivability and transparency during unloaded free motions of the human knee. Experimental results are presented and effectively demonstrate the functionality of the iT-Knee device.

The Wolverine is a mobile, wearable haptic device designed for simulating the grasping of rigid objects in a virtual reality interface. In contrast to prior work on wearable force feedback gloves, we focus on creating a low cost and lightweight device that renders a force directly between the thumb and three fingers to simulate objects held in pad opposition (precision) type grasps. Leveraging low-power brake-based locking sliders, the system can withstand over 100N of force between each finger and the thumb, and only consumes 0.24 mWh (0.87 joules) for each braking interaction. Integrated sensors are used both for feedback control and user input: time-of-flight sensors provide the position of each finger and an IMU provides overall orientation tracking. This paper describes the mechanical design, control strategy, and performance analysis of the Wolverine system and provides a comparison with several existing wearable haptic devices.

This paper presents a new method for estimation and identification of shear wind and discrete gusts of a previously unknown wind field by using an Unmanned Aircraft System (UAS). Wind estimation and identification is key in energy-efficient trajectory planning and dynamic soaring applications. The research proposes an approach for mapping a complete wind field from the collected data. Therefore, the generated map also describes areas where UAS has not passed through. The proposed method consists of the next steps: 1) the wind vector is estimated in each UAS position; 2) wind data are fitted into a Weibull probability density function and meeting the Prandtl's power law relationship; 3) scale factor of the Weibull distribution and the power law coefficient are computed; 4) wind feature detection such as shear layer and gusts is performed from the relation wind magnitude vs. altitude obtained; and finally 5), data could be extrapolated to generate the complete wind field. Novel aspects and advantages include the optimization of the scale factor from the estimated wind data by using a genetic algorithm, the identification of wind features separately, and the possibility to apply the method online. Real data of flight have been used to validate the method and many simulations and studies have been performed to test and analyze the proposed method in different scenarios.

Localization of mobile robots can be challenging in highly dynamic environments (airport terminals, stores, hospitals) yet these areas offer great potential for robotics applications. HighWiFi coverage provides a way to localize even low-cost robots that do not need to be equipped with expensive exteroceptive sensors or excessive computational power to run visual-based SLAM algorithms. Since current trend in robotics is in mass-deployment of low-cost, replaceable robots, we focus our efforts in this direction as well. Possible solution is combination of high-tech robots deployed as support assuring reliable localization for the low-cost robots that are actually doing the job. The proposed scenario is to use one SLAM robot to map WiFi signal strength in the working area and provide it to a low-cost robot to correct drift of its 6-DOF gyro-odometry localization system. For this purpose, we extend Gaussian-processes-based WiFi localization algorithms to full 3D and experimentally evaluate the proposed approach in 2 mapping and 7 localization sorties performed on different dates spanning four months.

The task of whole-body motion planning for humanoid robots is challenging due to its high-DOF nature, stability constraints, and the need for obstacle avoidance and movements that are efficient. Over the years, various approaches have been adopted to solve this problem such as bounding-box models and jacobian-based techniques. More commonly though, sampling-based algorithms are employed for this task since they perform admirably well in high-dimensional spaces. As an alternative, search-based planners offer improvements in terms of optimality and consistency of the solution. However, they are normally considered impractical for high-dimensional motion planning. In this paper, we present a heuristic search-based motion planning framework for humanoid robots that circumvents the drawbacks traditionally associated with search-based planners while catering to the specific requirements of humanoid motion planning. This is achieved primarily through a combination of informative yet computationally inexpensive heuristics, carefully crafted motion primitives as atomic actions, and a whole body inverse kinematics solver for achieving desired end effector orientations. The experimental results show the ability of our framework to perform complex motion planning tasks quickly and efficiently.

In this work, we present an approach to learn cost maps for driving in complex urban environments from a large number of demonstrations of human driving behaviour. The learned cost maps are constructed directly from raw sensor measurements, bypassing the effort of manually designing cost maps as well as features. When deploying the cost maps, the trajectories generated not only replicate human-like driving behaviour but are also demonstrably robust against systematic errors in putative robot configuration. To achieve this we deploy a Maximum Entropy based, non-linear IRL framework which uses Fully Convolutional Neural Networks (FCNs) to represent the cost model underlying expert driving behaviour. Using a deep, parametric approach enables us to scale efficiently to large datasets and complex behaviours while being run-time independent of dataset extent during deployment. We demonstrate scalability and performance on an ambitious dataset collected over the course of one year including more than 25k demonstration trajectories extracted from over 120km of driving and 13 different drivers. We evaluate against a carefully designed cost map and, in addition, demonstrate robustness to systematic errors by learning precise cost-maps even in the presence of system calibration perturbations.

In this paper we present motion optimization algorithms for computing manipulation motions in presence of obstacles. Our approach builds a geometric representation of the workspace by constructing Riemannian metrics using electric potentials emanating from the workspace obstacles. Velocity of the robot's body is measured with respect to this metric instead of traditional Cartesian velocity. Empirical results demonstrate that Riemannian metrics are better handled by optimizers that leverage objective and constraint functions' second order information. This information encodes how the Riemannian geometry of the modeled workspace pulls back into the configuration space. We also show that despite the additional computational burden of computing the electric-potential based metric, it results in faster overall convergence than reasoning on Euclidean geometry alone. Evaluation is made efficient by cashing the electric potential in voxel grids and using Tri-cubic spline interpolation to assess the potentials gradient.

For robots, a flexible approach is very important. Even when the robot is a humanoid, it does not have to perfectly mimic human motion. If necessary, the humanoid robot can adapt to other technologies even when these technologies make it different from humans. In a real environment or a disaster environment, humanoid robots are a good solution to assist humans. However, the stability problem of biped locomotion remains. In this paper, we introduce a new strategy of attaching a wheel to our humanoid robot in order to achieve mobility efficiency with a bipedal walking and wheel driven mode. Additionally, this strategy can enhance the stability. Through the proposed method, the humanoid robot can traverse uneven terrain and climb stairs with a walking motion, and can move safely with wheels on the ground. In addition, a method will be presented that can enhance the performance of wheel movement by using redundant joints and sensors on the humanoid robot. Various experiments and the DARPA Robotics Challenge results validate the efficiency and the feasibility of the proposed strategy.

In this paper, we develop a waterproof suit for humanoid robots and propose an underwater walking control method. Although very few life-sized humanoid robots are completely waterproof, we can easily make these humanoid robots watertight by putting a waterproof suit on them. In water, humanoid robots are influenced by the two forces due to the water: buoyancy and drag force. We take buoyancy into account when generating a walking pattern because the force is large and easy to estimate before walking. However, drag force is small and difficult to precisely predict and therefore, we treat the force as an unknown disturbance. In our method, we modify footsteps based on the Capture Point in order to deal with large disturbances. We verify the effectiveness of the proposed methods through an experiment in which a life-sized humanoid robot walks on a floor, stairs and debris in water.

Based on the integrated design concept of using the robot's mechanical arm for both locomotion and manipulation, we proposed a compact and transformable robot, i.e. the WALS-Robot, which utilizes combination and switch among four components: wheels, arms, legs and suckers. The WALS-Robot could transform among five locomotion modes to fulfill locomotion requirements in indoor unstructured environments, and have dexterous manipulation, and low power consumption. To reduce the redundancy of mechanical structure and the moving payload for the robot, the mobile platform and the mechanical arm could work as a unity, i.e. the mechanical arm can switch into a leg of the robot to reduce payload during moving modes. Experimental results validated key performance of the robot, including locomotion accuracy and efficiency, compact volume and low power consumption. The integrated design concept illustrated the adaptability in diverse unstructured indoor environments, enlarging the manipulation workspace, and enhancing the energy efficiency.

In order to define trajectory-tracking strategies for nano-robotic systems using piezoelectric stick-slip actuators, the dynamics of the elementary actuator must be studied and well modeled. The modeling of this class of actuators is complex because several nonlinear parameters are involved. In this paper, we propose a systematic modeling methodology of piezoelectric stick-slip actuators for nano-robotic systems control. The main idea is the proposition of an augmented voltage/frequency rate dependent modeling of the friction force based on a multi-state elasto-plastic formulation. Experimental and simulation results demonstrate the efficiency of the model in the time and the frequency domains. As a case of study, the proposed model is used to define a control strategy in order to detect collisions when the nano-robotic system is operating inside a Scanning Electron Microscope (SEM). This application demonstrates the need of a voltage/frequency rate dependent modeling.

Disaster response robots for searching in a narrow area have limited space to mount tactile sensors, but the operators require sensory feedback to recognize contact situations with the surrounding environment. This study proposes a new approach to transmitting contact information of a remote-operated snake-like robot called Active Scope Camera (ASC) to the operator using simple configurations for the sensing and display methods. For the sensing side, we develop a contact estimation method with a limited number of tactile sensors. We establish the method to localize the contact position and the magnitude by sensing multiple propagated vibrations based on experiments and formulations. Preliminary experiments show that the developed method estimates a collision angle with high probability (93.8% at the worst condition) at several collisional situations. For the display side, we combine visual and vibrotactile feedback to provide the operator both directional and temporal cues to perceive contact events. The proposed visualization method uses colored bars, peripherally superposed on the video image, to show the estimated contact location and magnitude. A single DoF vibrotactile feedback is used for a joystick interface to control the head movement of the ASC. The effect of vibrotactile feedback on the response time to contact events is evaluated. Finally, we investigate the performance of the operation by identifying the contact behavior at simulated scenarios. Experimental results show that collision times per operation time is decreased by the developed feedback system compared with a simple video-based operation.

We present a simultaneous localization and mapping (SLAM) algorithm that uses Bézier curves as static landmark primitives rather than sparse feature points. Our approach allows us to estimate the full 6-DOF pose of a robot while providing a structured map which can be used to assist a robot in motion planning and control. We demonstrate how to reconstruct the 3-D location of curve landmarks from a stereo pair without searching for point-based stereo correspondences and how to compare the 3-D shape of curve landmarks between chronologically sequential stereo frames to solve the data association problem. We present a method to combine curve landmarks for mapping purposes, resulting in a map with a continuous set of curves that contain fewer landmark states than conventional sparse point-based SLAM algorithms. Note, to combine curves, we assume the curved landmarks are fixed to a larger curved object naturally occurring in the scene. While our algorithm is less accurate than point-based SLAM algorithms, we are able to create maps with considerably less landmark states and our algorithm can operate in settings lacking texture.

We present a general framework for accurate positioning of sensors and end effectors in farm settings using a camera mounted on a robotic manipulator. Our main contribution is a visual servoing approach based on a new and robust feature tracking algorithm. Results from field experiments performed at an apple orchard demonstrate that our approach converges to a given termination criterion even under environmental influences such as strong winds, varying illumination conditions and partial occlusion of the target object. Further, we show experimentally that the system converges to the desired view for a wide range of initial conditions. This approach opens possibilities for new applications such as automated fruit inspection, fruit picking or precise pesticide application.

We propose a visual programming system that allows users to specify navigation tasks for mobile robots using high-level landmarks in a virtual reality (VR) environment constructed from the output of visual simultaneous localization and mapping (vSLAM). The VR environment provides a Google Street View-like interface for users to familiarize themselves with the robot's working environment, specify high-level landmarks, and determine task-level motion commands related to each landmark. Our system builds a roadmap by using the pose graph from the vSLAM outputs. Based on the roadmap, the high-level landmarks, and task-level motion commands, our system generates an output path for the robot to accomplish the navigation task. We present data structures, architecture, interface, and algorithms for our system and show that, given ns search-type motion commands, our system generates a path in O(ns(nr lognr+mr)) time, where nr and mr are the number of roadmap nodes and edges, respectively. We have implemented our system and tested it on real world data.

In this paper we present an approach for efficiently retrieving the most similar image, based on point-to-point correspondences, within a sequence that has been acquired through continuous camera movement. Our approach is entailed to the use of standardized binary feature descriptors and exploits the temporal form of the input data to dynamically adapt the search structure. While being straightforward to implement, our method exhibits very fast response times and its Precision/Recall rates compete with state of the art approaches. Our claims are supported by multiple large scale experiments on publicly available datasets.

This paper presents a vision-based approach for estimating the configuration of, and providing control signals for, an under-sensored robot manipulator using a single monocular camera. Some remote manipulators, used for decommissioning tasks in the nuclear industry, lack proprioceptive sensors because electronics are vulnerable to radiation. Additionally, even if proprioceptive joint sensors could be retrofitted, such heavy-duty manipulators are often deployed on mobile vehicle platforms, which are significantly and erratically perturbed when powerful hydraulic drilling or cutting tools are deployed at the end-effector. In these scenarios, it would be beneficial to use external sensory information, e.g. vision, for estimating the robot configuration with respect to the scene or task. Conventional visual servoing methods typically rely on joint encoder values for controlling the robot. In contrast, our framework assumes that no joint encoders are available, and estimates the robot configuration by visually tracking several parts of the robot, and then enforcing equality between a set of transformation matrices which relate the frames of the camera, world and tracked robot parts. To accomplish this, we propose two alternative methods based on optimisation. We evaluate the performance of our developed framework by visually tracking the pose of a conventional robot arm, where the joint encoders are used to provide ground-truth for evaluating the precision of the vision system. Additionally, we evaluate the precision with which visual feedback can be used to control the robot's end-effector to follow a desired trajectory.

Real-time 3D mapping with MAV (Micro Aerial Vehicle) in GPS-denied environment is a challenging problem. In this paper, we present an effective vision-based 3D mapping system with 2D laser-scanner. All algorithms necessary for this system are on-board. In this system, two cameras work together with the laser-scanner for motion estimation. The distance of the points detected by laser-scanner are transformed and treated as the depth of image features, which improves the robustness and accuracy of the pose estimation. The output of visual odometry is used as an initial pose in the Iterative Closest Point (ICP) algorithm and the motion trajectory is optimized by the registration result. We finally get the MAV's state by fusing IMU with the pose estimation from mapping process. This method maximizes the utility of the point clouds information and overcomes the scale problem of lacking depth information in the monocular visual odometry. The results of the experiments prove that this method has good characteristics in real-time and accuracy.

In this paper, a method is proposed for vision-based within-hand precision manipulation with underactuated grippers. The method combines the advantages of adaptive underactuation with the robustness of visual servoing algorithms by employing simple action sets in actuator space, called precision manipulation primitives (PMPs). It is shown that, with this approach, reliable precision manipulation is possible even without joint and force sensors by using only minimal gripper kinematics information. An adaptation method is also utilized in the vision loop to enhance the system's transient performance. The proposed methods are analyzed with experiments using various target objects and reference signals. The results indicate that underactuated hands, even with minimalistic sensing and control via visual servoing, can provide a simple and inexpensive solution to allow low-fidelity precision manipulation.