IROS 2018 Papers

In this paper, we present sensor-controlled antagonistic pneumatic actuators (SCAPAs) that integrate proven soft robotic actuators and sensors into a simplified, controllable design. The antagonistic actuators together compose a bidirectional bending actuator with embedded capacitive strain sensors. By designing the SCAPAs from the ground-up for closed-loop control, we are able to minimize both the number of constituent components and the types of materials used, and further streamline the manufacturing processes. These improvements are embodied in the multipurpose use of a single conductive fabric sheet for both actuation and sensing, integrated into an otherwise all-silicone device. Such reduced material complexity allows us to use simple finite element analysis (FEA) models to predict the performance of a given design. We compare various designs to maximize sensor effectiveness using FEA and experimentally verify the suitability of select designs for state reconstruction. After converging on our final design, we demonstrate that this design evaluation process enables the use of simple control strategies to achieve closed-loop control.

In this paper we tackle the problem of finding the source of a gaseous leak with a robot in a three-dimensional (3-D) physical space. The proposed method extends the operational range of the probabilistic Infotaxis algorithm [1] into 3-D and makes multiple improvements in order to increase its performance in such settings. The method has been tested systematically through high-fidelity simulations and in a wind tunnel emulating realistic conditions. The impact of multiple algorithmic and environmental parameters has been studied in the experiments. The algorithm shows good performance in various environmental conditions, particularly in high wind speeds and different source release rates.

This work presents the design and development of a semi-active hybrid orthotic system for support and facilitation of unilateral pathological human walking. The system is based on a novel lower limb orthosis with mechanical control and its combination with non-invasive muscle electrostimulation. The paper presents the concept design and realization of a novel Stance Control Knee Ankle Foot Orthoses (SCKAFO) and the design of Functional Electrical Stimulation (FES) strategies for artificial activation of ankle joint muscles in this hybrid scheme for gait support. In particular, we present the investigation of the effects of electrical stimulation patterns of ankle muscles synchronized in real-time with gait events., on the possible biomechanical alterations to gait in able-bodied individuals (n=8). Bilateral 3D ground reaction forces (GRF) were analyzed between barefoot overground walking with and without FES of ankle muscles. The observed effects of the tested FES strategy are coherent with a physiological gait strategy and compatible with the proposed SCKAFO.

This paper proposes a novel aerial manipulator with tandem ducted fans, which takes both trafficability and effective loading into account. The aerial manipulator is particularly suitable for grasping in complex and narrow environment, in which traditional multi-rotor and helicopter would be inaccessible. The comprehensive integrated dynamic model is established by taking the aerial vehicle dynamics and manipulator dynamics as a whole. On this basis, a multilayer composite controller with feedforward compensation is designed, considering the mutual reactive influence between the aerial vehicle and the manipulator to improve the stability of the system under the motion of the manipulator. The simulation and actual flight tests verify the effectiveness of the design and show good stability and tracking performance of the system.

This paper presents the design and implementation of a preliminary version of a programmable drawing automaton (PDA-0) that draws a user-specified 3D trajectory. PDA-0 is strongly inspired by Jaquet Droz's a programmable drawing automaton built in the 1770s using 6,000 moving parts, which was hand-coded. PDA-0 consists of RSSR (Revolute-Spherical-S-R) linkage and cam mechanisms with three interchangeable cams. Interchangeable cams make PDA-0 programmable because a user-specified 2D/3D trajectory is encoded into the set of three cams. The user programs PDA-0 by specifying a trajectory via a GUI or 3D animation. Subsequently, the compiler estimates a 3D trajectory mathematically from the user-specified 2D/3D trajectory and calculates the shape of the three cams, i.e., a code for PDA-0 by solving kinematic constraints. Finally, PDA-0 with the 3D-printed cams executes the code to draw the user-specified trajectory. The current PDA-0 with three cams demonstrates drawing simple trajectories such as letters and symbols.

This study presents a new drone, called S-CLOUD, developed for safe and long flight time. It provides 3-axial (x, y, and z)translational motion and stable hovering for more than an hour after takeoff. S-CLOUD consists of two parts; soft blimp part and driving one. The soft blimp is a center-pierced torus-shaped part filled with Helium gas. Thus, it is safe to fly near people because it is light and soft, and all its rotating parts are at the center of the vehicle, which does not get damaged on collision. The driving part is plugged into the center of the soft blimp and includes the flow control mechanism, which consists of co-axial rotors and 2-axis crossed flaps. It controls the altitude, attitude, and translational movements of the vehicle. Its dynamic and reaction features against disturbances are derived using Newton-Euler formulation, and the simulation results are discussed. Finally, a prototype of S-CLOUD is fabricated and its feasibility is experimentally validated with practical applications.

In this paper, we present the design and performance evaluation of a bipedal robot that utilizes the Hybrid Leg mechanism. It is a leg mechanism that achieves 6 DOF with a combined structure of serial and parallel mechanism. It is designed to have a light structural inertia and large workspace for agile bipedal locomotion. A new version of Hybrid Leg is fabricated with carbon fiber tubes and bearings to improve its structural rigidity and accuracy while supporting its weight. A pair of Hybrid Legs is assembled together for bipedal locomotion. In the assembly, we adopt a pelvis structure with an yaw angle offset to enlarge the feet workspace, inspired by the toe-out angle of the human feet. The workspace and range of velocity are presented in simulation and verified with hardware experiments. We also demonstrate a simple forward walking motion with the developed robot.

Wheel-legged hybrid robot with multi-modal locomotion can efficiently adapt to different terrain environments, as well as realize rapid maneuver on flat ground. We have developed a novel hydraulic wheel-legged robot (WLR) combined with a humanoid structural design. This robot can assist to emergency scenarios where the high mobility, adaptability and robustness are required. The paper introduces the details of the WLR, highlighting the innovative design and optimization of physical construction which is considered to maximize the mobile abilities, enhance the environmental adaptability and improve the reliability of hydraulic system. Firstly, maximizing the mobile abilities includes optimizing the configuration of each actuator and integrating them with the structure, so as to achieve a large range of movement and also reduce the mass and inertia of the legs. Secondly, the environmental adaptability can be ensured with a magnetorheological (MR) fluid-based damper and direct-drive wheels. Thirdly, improving the reliability of hydraulic system involves using the selective laser melting (SLM) technology to integrate hydraulic system and reducing the number of exposed tubes. The maneuverability of the WLR is demonstrated with a series of experiments. At present, the WLR can perform the following operations, including moving on the flat ground, squatting, and picking up a heavy load.

This paper introduces a novel hydrostatic air-liquid torque transmission system for an upper limb exoskeleton. The proposed design is based on remote electrical actuation, with grounded motors, combined with high performance fluid power transmission employed to deliver the power to the joints of the exoskeleton. The fluid transmission is based on rolling membrane cylinders that guarantee leakage-free operation, no backlash, and virtually zero stick-friction. This solution makes it possible to obtain easy controllability, good efficiency, intrinsic backdrivable operation, and reduced mass/inertia of the links of the robot. Additionally, the proposed system can be potentially implemented at relatively low-costs thanks to the employment of standard components and an architecture based on a modular approach. A test bench of the fluid transmission system is developed and a campaign of experiments is conducted to characterize its static/dynamic response for different choice of design parameters. In addition, we present a preliminary complete integrated arrangement of an upper limb exoskeleton equipped with the proposed transmission system. Results confirm the feasibility of the proposed actuation approach for the envisaged application.

In this paper, we develop a bipedal robot control system that has an ability to perform instantaneous high power and flexibility to absorb an impulsive disturbance. We utilize a sensor-less whole body torque control method executed in a high responsive realtime distributed system. This system also includes a robust online walking controller that can avoid fall over caused by a strong collision with the robot's legs. We evaluated the proposed control system by hitting a rubber ball or adding a leg sweep disturbance and verified the functionality of the absorbing motion and the balance restoring motion.

In-situ aeroengine maintenance works (e.g. inspection, repair) are highly beneficial as it can significantly reduce currently accepted maintenance cycle which is extensive and costly due to the need to remove engines from the wing of an aircraft. However, feeding in/out via inspection ports and performing a multi-axis movement of an end-effector in a very constrained environment such as aeroengine combustion chamber is a fairly challenging task. This paper presents the design and development of a highly slender (i.e., low diameter-to-length ratio) dual-structure continuum robot with 16 degrees of freedom (DoFs) to provide the feeding motion needed to navigate into confined environments and then perform a required configuration shape for further repair operation. This continuum robot is a compact system and presents a set of innovative mechatronic solutions such as: (i) two-stage tendon-driven structure with bevelled disk design to perform required configuration shape and to provide selective stiffness for the ability of taking high payloads; (ii) various compliant joints to enable different flexibility requirement in each stage; (iii) three commanding cables for each 2-DoF section to minimise the number of actuators with a precise actuation. To be able to achieve the desired configuration shape, a kinematic model has been established and the configuration-cable kinematics has been implemented. Finally, the continuum robot has been built and tested for performing the predefined configuration shape.

Long flexible continuum robots have a high potential for search and rescue operations that explore deep layered debris. A general problem of these robots is in the control of the head motion because their thin bodies limit the space available to mount multiple actuators. This paper develops a biaxial active nozzle which can rotate the air jet direction along a roll and pitch axis in order to control the direction of reaction force and the head motion of a long flexible robot. A major challenge is how to change the air jet direction without a large resistance to the flow, which reduces the reaction force induced by the air jet. We propose a nozzle whose outlet is connected with a flexible air tube. The direction of the air jet is controlled by the smooth shape deformation of the tube. The nozzle should be compact enough to be installed on a thin robot, although the shape deformation of the tube may cause buckling. The flexible tube is modeled and simulated by a multiple link model used to derive the geometric parameters of the nozzle so that the nozzle is compact and the tube does not buckle. Based on the derived parameters, the biaxial active nozzle was developed. A basic performance experiment shows that the nozzle can change the reaction force direction by deforming the tube shape, while the magnitude of the reaction force is almost constant. We integrated the proposed nozzle with a conventional Active Scope Camera (ASC). The range where the robot can look around in a vertical exploration was significantly improved, which was three times larger than the previous ASC whose head was controlled by pneumatic actuators. The rubble field test demonstrates that the integrated ASC could move over rubble (maximum height of 200 mm) and steer the course.

Estimating the three-dimensional (3D) position of microrobots is necessary in order to develop closed-loop control techniques and to improve the user's 3D perception in the micro-scale. This paper describes a depth estimation method based on supervised learning for optically transparent microrobots of known geometry. The proposed methodology uses Convolutional Neural Networks (CNNs) combined with a Recurrent Network, in particular a Long Short-Term Memory (LSTM) cell for depth regression. The proposed depth regression model is independent of the 3D orientation of the microrobot and is robust to varying illumination levels while it uses learned data-specific features. The model is trained and validated using microscope images and ground truth data generated from 3D-printed microrobots imaged in an Optical Tweezers (OT) setup. The validation results demonstrate that the proposed trained model can reconstruct the depth of the microrobot independently of its 3D orientation with submicron accuracy for the test set.

Motivated by the need for tools to aid in the design of effective robots, we examine how to determine the role that particular sensing and actuator resources play in enabling a robot to achieve useful ends. Rather than merely asking “will this sensor suffice?” we classify general modifications to the set of sensors and actuators based on the feasibility of accomplishing given tasks using these sets. The goal is to probe the boundary between modifications that are destructive on a given planning problem, and modifications that are not. Since this boundary itself can be impractically large, classic search methods are of no avail to summarize discriminatory features on this boundary. Instead, we propose a decision tree learning method to efficiently construct a compact implicit representation of the boundary. The idea is to allow the designer to use prior knowledge to constrain the search, then use the tool to probe the boundary subject to those constraints, gaining insight into the information necessary for a robot to ensure task achievement. Ultimately we envision a interactive process where additional constraints are repeatedly included as new light is shed. We aim to pave the way for interactive tools that help the roboticist navigate the complexities of the design space. We describe an implementation of this approach along with experimental results that show that the method can construct decision trees with explanatory value. Our experiments suggest that some domain knowledge (specifically picking features that emphasize monotonicity) substantially improves running-time with only negligible reduction in accuracy.

In this video we present soft self-sensing capacitive arrays and demonstrate their use in capturing dense surface deformations without requiring line of sight. The capacitive arrays are made of two electrode patterns embedded into a single silicone compound. The overlaps of the electrode strip patterns form local capacitors. As the sensor is stretched the local capacitance measurements change. We introduce a fabrication technique that allows to produce such sensors while only requiring hardware readily available in modern fablabs. The resulting sensors are able to densely capture area changes as they deform. Since they do not directly measure bend, a prior is required to fully reconstruct the underlying motion. We propose a deep neural network regressing the sensor geometry from the local area measurements. A motion capture system is used for training data acquisition. At runtime vertex positions are predicted and used as positional constraints to deform a mesh using a state-of-the-art elastic surface energy. The flexibility and accuracy of the introduced sensors is demonstrated in a series of controlled experiments and by fabricating a prototype sensor and applying it to capture deforming skeletal and non-skeletal objects.

Sampling-based Motion Planners (SMPs) have become increasingly popular as they provide collision-free path solutions regardless of obstacle geometry in a given environment. However, their computational complexity increases significantly with the dimensionality of the motion planning problem. Adaptive sampling is one of the ways to speed up SMPs by sampling a particular region of a configuration space that is more likely to contain an optimal path solution. Although there are a wide variety of algorithms for adaptive sampling, they rely on hand-crafted heuristics; furthermore, their performance decreases significantly in high-dimensional spaces. In this paper, we present a neural network-based adaptive sampler for motion planning called Deep Sampling-based Motion Planner (DeepSMP). DeepSMP generates samples for SMPs and enhances their overall speed significantly while exhibiting efficient scalability to higher-dimensional problems. DeepSMP's neural architecture comprises of a Contractive AutoEncoder which encodes given workspaces directly from a raw point cloud data, and a Dropout-based stochastic deep feedforward neural network which takes the workspace encoding, start and goal configuration, and iteratively generates feasible samples for SMPs to compute end-to-end collision-free optimal paths. DeepSMP is not only consistently computationally efficient in all tested environments but has also shown remarkable generalization to completely unseen environments. We evaluate DeepSMP on multiple planning problems including planning of a point-mass robot, rigid-body, 6-link robotic manipulator in various 2D and 3D environments. The results show that on average our method is at least 7 times faster in point-mass and rigid-body case and about 28 times faster in 6-link robot case than the existing state-of-the-art.

We demonstrate how a sequence model and a sampling-based planner can influence each other to produce efficient plans and how such a model can automatically learn to take advantage of observations of the environment. Sampling-based planners such as RRT generally know nothing of their environments even if they have traversed similar spaces many times. A sequence model, such as an HMM or LSTM, guides the search for good paths. The resulting model, called DeRRT*, observes the state of the planner and the local environment to bias the next move and next planner state. The neural-network-based models avoid manual feature engineering by co-training a convolutional network which processes map features and observations from sensors. We incorporate this sequence model in a manner that combines its likelihood with the existing bias for searching large unexplored Voronoi regions. This leads to more efficient trajectories with fewer rejected samples even in difficult domains such as when escaping bug traps. This model can also be used for dimensionality reduction in multi-agent environments with dynamic obstacles. Instead of planning in a high-dimensional space that includes the configurations of the other agents, we plan in a low-dimensional subspace relying on the sequence model to bias samples using the observed behavior of the other agents. The techniques presented here are general, include both graphical models and deep learning approaches, and can be adapted to a range of planners.

In autonomous driving systems a strong relation to highly accurate maps is taken to be inevitable, although street scenes change frequently. However, a preferable system would be to equip the automated cars with a sensor system that is able to navigate urban scenarios without an accurate map. We present a novel pipeline using a deep neural network to detect lane semantics and topology given RGB images. On the basis of this classification, the information about the road scene can be extracted just from the sensor setup supporting mapless autonomous driving. In addition to superseding the huge effort of creating and maintaining highly accurate maps, our system reduces the need for precise localization. Using an extended Cityscapes dataset, we show accurate ego lane detection including lane semantics on challenging scenarios for autonomous driving.

Mobile robot navigation in complex and dynamic environments is a challenging but important problem. Reinforcement learning approaches fail to solve these tasks efficiently due to reward sparsities, temporal complexities and high-dimensionality of sensorimotor spaces which are inherent in such problems. We present a novel approach to train action policies to acquire navigation skills for wheel-legged robots using deep reinforcement learning. The policy maps height-map image observations to motor commands to navigate to a target position while avoiding obstacles. We propose to acquire the multifaceted navigation skill by learning and exploiting a number of manageable navigation behaviors. We also introduce a domain randomization technique to improve the versatility of the training samples. We demonstrate experimentally a significant improvement in terms of data-efficiency, success rate, robustness against irrelevant sensory data, and also the quality of the maneuver skills.