IROS 2018 Papers

Our fingernails help us to accomplish a variety of manual tasks, but surprisingly only a few robotic hands are equipped with nails. In this paper, we present a sensorized fingernail for mechatronic hands that can capture static and dynamic interaction forces with the nail. Over the course of several iterations, we have developed a very compact working prototype that fits together with our previously developed multi-cell tactile fingertip sensor into the cavity of the distal phalange of a human-sized robotic hand. We present the construction details, list the key performance characteristics and demonstrate an example application of finding the end of an adhesive tape roll using the signals captured by the sensors integrated in the nail. We conclude with a discussion about improvement ideas for future versions.

The aim of this study is to analyse the modular mechanical design and integration of all three low-level modules in UX-1 (pendulum, ballast system and propulsion unit). The components of the perception and navigation systems have position and orientation requirements that dictate the shape of the hull. A structural strength analysis using Finite Element method (FEM) was made to study the hull strength during deep dives. The results are presented here, which indicates that the hull endures pressures related to deep dives. Also for validation, strain gauge locations were defined.

Previously, we presented the concept of a novel haptic display device that could generate lateral localized displacement on a human fingertip. This device is characterized by a bundle of haptic pins whose ends gently make contact with a human fingertip. In this paper, we proposed a dynamic model of interaction between haptic pins and finger for investigation of mechanical response of stress or strain on human fingertip under operation of the proposed haptic device. We also conducted preliminary experiment to determine the possible setups that maximizes the sense of partial slippage. The results presented in this paper may help assess human slip perception for the development of haptic display devices.

In this paper we present an approach for material recognition using capacitive tactile and proximity sensors. By variating the spatial resolution and the exciter frequency during the measurement in mutual capacitive mode, information about the dielectrical properties of different objects was captured and provided as data frames. For material recognition an artificial neural network was set up and fed with various data sets of different electrode combinations and exciter frequencies. The influence of the electrode combinations and shapes on the recognition accuracy was investigated. It is shown that seven objects of conductive and non-conductive dielectric materials have been ranged with an overall accuracy of about 71%-94%.

Programming robots for tasks involving force interaction is difficult, since both the knowledge of the task and the dynamics of the robots are necessary. An immersive haptic interface for task demonstration is proposed, where the operator can sense and act through the robot. This is achieved by coupling two robotic systems with virtual constraints such that they have the same coordinates in the operational space disregarding a fixed offset. Limitations caused by the singular configurations or the reach of the robots are naturally reflected to either side as haptic feedback.

Currently, the manufacturing of self-actuating and self-sensing robots requires non-standard manufacturing techniques and assembly steps to integrate electrical and mechanical systems. In this work, we developed a novel manufacturing technique, where such robots can be produced at a flexible electronics factory. We developed the technique using standard industrial machines, processes, and materials. Using a lamination process, we were able to integrate air pouches or shape memory alloy (SMA) inside a polyamide-based flexible circuit to produce bending actuators. The bend angle of the actuators is sensed with a chain of inertial measurement units integrated on the actuator. Air-pouch actuators can produce a force of a 2.24N, and a maximum bend angle of 74 degrees. To demonstrate, we manufactured a five-legged robot with the developed actuators and bend sensors, with all the supporting electronics (e.g., microcontrollers, radio) directly integrated into the flexible printed circuit. Such robots are flat and lightweight (15 grams) and thus conveniently compact for transportation and storage. We believe that our technique can allow inexpensive and fast prototyping and deployment of self-actuating and self-sensing robots.

Making an optimal decision to avoid obstacles while heading to the goal is one of the fundamental challenges for mobile robots equipped with limited computational resources. In this paper, we present a deep imitation learning algorithm that develops a computationally efficient obstacle avoidance policy based on egocentric local occupancy maps. The trained model embedded with a variant of the value iteration networks is able to provide near-optimal continuous action commands through fast feed-forward inferences and generalize well to unseen planning-based scenarios. To improve the policy robustness, we augment the training data set with artificially generated maps, which effectively alleviates the shortage of catastrophic samples in normal demonstrations. Extensive experiments on a Segway robot show the effectiveness of the proposed approach in terms of solution optimality, robustness as well as computation time.

We present a manipulation planning algorithm for a robot to keep an object stable under changing external forces. We particularly focus on the case where a human may be applying forceful operations, e.g. cutting or drilling, on an object that the robot is holding. The planner produces an efficient plan by intelligently deciding when the robot should change its grasp on the object as the human applies the forces. The planner also tries to choose subsequent grasps such that they will minimize the number of regrasps that will be required in the long-term. Furthermore, as it switches from one grasp to the other, the planner solves the problem of bimanual regrasp planning, where the object is not placed on a support surface, but instead it is held by a single gripper until the second gripper moves to a new position on the object. This requires the planner to also reason about the stability of the object under gravity. We provide an implementation on a bimanual robot and present experiments to show the performance of our planner.

In this paper, we examine the reduced order pendular dynamic climbing model with the addition of attachment windows based on prescribed body roll. With this model and on the new dynamic climbing platform, TAILS, we demonstrate dynamic downward climbing as well as identify distinct dynamic gaits within downward climbing. This, combined with the application of an asymmetric configuration of the rear legs enables strafing motions and thus dynamic maneuverability on walls in the vertical domain.

This article studies imitation learning policies that encode task compliance to provide teleoperation assistance for remote manufacturing. The central challenge is how to handle off-nominal situations, such as out-of-sequence work or unplanned obstacles, since the assistance has not been trained to handle such scenarios. In such cases, there is potential for the assistance to degrade-rather than improve-operator performance. This work proposes a method that exploits the learned task compliance to classify persistent human actions as off-nominal, and attenuate assistance in these regions. Applied to a hole-cleaning task with n = 11 subjects, the proposed method shows up to 17% reduction in task completion time and up to 68% reduction in forces in off-nominal situations as compared to assistance without the method. Additionally, the method retains the performance improvements of assistance in nominal operating regimes.

In this paper we present the design and control strategies of a novel quadruped climbing robot (named Magneto) with three degrees of freedom (3-DOF) actuated limbs and a 3-DOF compliant magnetic foot. By exploiting its high degrees of freedom, Magneto is able to deform its body shape to squeeze through gaps of 23cm, which is smaller than standard human entry portholes of industrial confined spaces. Its compact foot design of footprint 4cm allows Magneto to walk on narrow beams of thickness less than 5cm, even at varying separation. The inherent high dimensional system design enables the body to be positioned in a wide range of orientations and seamlessly switch a limb function from locomotion to manipulation mode mid-climb. This capability enables access to confined space openings and occluded pockets and navigation through complex 3-D structures previously not demonstrated on legged climbing robots.

Taking inspiration from the properties of magnetic fields, we propose a reactive navigation method for soft continuum manipulators operating in unknown environments. The proposed navigation method outperforms previous works since it is able to successfully achieve collision-free movements towards the goal in environments with convex obstacles without relying on a priori information of the obstacles' shapes and locations. Simulations for the kinematic model of a soft continuum manipulator and preliminary experiments with a 2-segments soft continuum arm are performed, showing promising results and the potential for our approach to be applied widely.

Reliable and real-time 3D reconstruction and localization functionality is a crucial prerequisite for the navigation of actively controlled capsule endoscopic robots as an emerging, minimally invasive diagnostic and therapeutic technology for use in the gastrointestinal (GI) tract. In this study, we propose a fully dense, non-rigidly deformable, strictly real-time, intraoperative map fusion approach for actively controlled endoscopic capsule robot applications which combines magnetic and vision-based localization, with non-rigid deformations based frame-to-model map fusion. The performance of the proposed method is evaluated using four different ex-vivo porcine stomach models. Across different trajectories of varying speed and complexity, and four different endoscopic cameras, the root mean square surface reconstruction errors vary from 1.58 to 2.17 cm.

Microrobots are considered as promising tools for biomedical applications. However, the imaging of them becomes challenges in order to be further applied on in vivo environments. Here we report the magnetic navigation of a paramagnetic nanoparticle-based swarm using ultrasound images. The swarm can be generated using simple rotating magnetic fields, resulting in a region containing particles with a high area density. Ultrasound images of the swarm shows a periodic changing of imaging contrast. The reason for such dynamic contrast has been analyzed and experimental results are presented. Moreover, this swarm exhibits enhanced ultrasound imaging in comparison to that formed by individual nanoparticles with a low area density, and the relationship between imaging contrast and area density is testified. Furthermore, the microrobotic swarm can be navigated near a solid surface at different velocities, and the imaging contrast show negligible changes. This method allows us to localize and navigate a microrobotic swarm with enhanced ultrasound imaging indicating a promising approach for imaging of microrobots.

Some electric-powered wheelchairs are recently redefined as personal mobility devices. Their users are not only elderly or handicapped people, but also passengers with large baggage or pedestrians going from station to destination, i.e., last-mile transport. Consequently, people with different operation skills and expectations on personal mobility would become new users of this kind of devices. Safe and comfort travel in human co-existing environment such as sidewalks and airports is a social expectation for personal mobility. In order to realize this, understanding the operation skill of each user by a practical and simple method is essential. This paper thus introduced a skill level classification method by machine learning using only joystick data as input. In order to determine the number of skill level clusters, basic 26 features of joystick operation data are used for unsupervised clustering (single-linkage). We then made evaluation indexes by using speed, speed control, and direction control. For a five-level classification by using gradient boosting as supervised learning, we achieved a 67% accuracy (tolerance: 0) and a 98% accuracy (tolerance: 1). Further analysis of the feature importance of gradient boosting revealed key features to a good operation. Results also show that skill level differed among people with different driving experiences.

This paper presents a methodology for height control of a quadrotor that transports a constant unknown load, given the estimates on both weight and state variables, based on measurements from motion sensors installed on-board. The proposed control and estimation framework is a Multi-Model Adaptive Controller using LQR with integrative action and Kalman filter with integrative component. The control system obtained is validated both in simulation and experimentally, resorting to an off-the-shelf commercially available quadrotor equipped with an IMU, an ultrasound height sensor, and a barometer, among other sensors.

This paper introduces a new robust, dynamic quadruped, the MIT Cheetah 3. Like its predecessor, the Cheetah 3 exploits tailored mechanical design to enable simple control strategies for dynamic locomotion and features high-bandwidth proprioceptive actuators to manage physical interaction with the environment. A new leg design is presented that includes proprioceptive actuation on the abduction/adduction degrees of freedom in addition to an expanded range of motion on the hips and knees. To make full use of these new capabilities, general balance and locomotion controllers for Cheetah 3 are presented. These controllers are embedded into a modular control architecture that allows the robot to handle unexpected terrain disturbances through reactive gait modification and without the need for external sensors or prior environment knowledge. The efficiency of the robot is demonstrated by a low Cost of Transport (CoT) over multiple gaits at moderate speeds, with the lowest CoT of 0.45 found during trotting. Experiments showcase the ability to blindly climb up stairs as a result of the full system integration. These results collectively represent a promising step toward a platform capable of generalized dynamic legged locomotion.

In this paper, we present a Mobile Agile Printer (MAP) construction robot; a highly agile, 4-legged, omnidirectional robot capable of 3D printing large structures. To overcome dynamic challenges when operating within an outdoors construction site, MAP incorporates a high-DoF 3D printing system connected to a mobile platform with novel features designed to enable disturbance rejection and live adaption to the robot's pose. In doing so, we demonstrate the benefits of designing construction robots with a focus on agility, a compact working volume and ability to operate within a potentially unlimited workspace. Performance tests were conducted showing smooth omni-directional motion as a key requirement for maintaining low 3D printing trajectory deviations over a large volume. In doing so, we show that MAP has the ability to construct in new ways more sensitive to its environment, context and concurrent on-site operations.

Many unmanned aerial vehicle surveillance and monitoring applications require observations at precise locations over long periods of time, ideally days or weeks at a time (e.g. ecosystem monitoring), which has been impractical due to limited endurance and the requirement of humans in the loop for operation. To overcome these limitations, we propose a fully autonomous small rotorcraft UAS that is capable of performing repeated sorties for long-term observation missions without any human intervention. We address two key technologies that are critical for such a system: full platform autonomy including emergency response to enable mission execution independently from human operators, and the ability of vision-based precision landing on a recharging station for automated energy replenishment. Experimental results of up to 11 hours of fully autonomous operation in indoor and outdoor environments illustrate the capability of our system.