IROS 2018 Papers

Precise actuation of continuum surfaces in combination with continuum robotic arms that undergo large deformation is of high interest in soft robotics but of limited model-based study to date. This work develops this area towards enabling the robust design and control of large deformation continuum surfaces (LDCS) across multiple industrial applications in the healthcare, aerospace, manufacturing, and automotive domains. It introduces an actuation based dynamic model of LDCSs to accurately determine their deflection due to application of concentrated external forces while maintaining many physical characteristics and constraints on actuation elements and surface structure such as gravity, inertia, damping, elasticity, and interactive forces between actuators and LDCS. Using the lumped-mass methodology, a 3D integrated surface-arm model is developed, simulated and then validated experimentally where a pair of parallel arms are attached to the surface to actuate and deform it. The surface is then simultaneously subjected to a concentrated constant external force at its top center between the two arms. Comparing measured displacements between the experimental and modelling results over actuation time yielded the maximum error is less than 1% of the length of the surface's side at its final deflected profile despite the limited number of nodes (masses) used in the LDCS model while it is exposed to a significant external force.

This paper presents modeling of robotic fish propelled by a hybrid tail with Servo and IPMC actuated two joints. The first joint is driven by a servo motor, which generates flapping motion for main propulsion. The second joint is actuated by a soft actuator, or ionic polymer-metal composite (IPMC) artificial muscle, which directs the propelled fluid for steering. A dynamic model is developed to capture the 2D motion dynamics of the robotic fish. The model fully captures the actuation dynamics of the IPMC soft actuator, two-link tail motion dynamics, and body motion dynamics. Experimental results have shown that the robotic fish is capable of swimming forward (up to 0.45 body length/second) and turning left and right (up to 40 degree/sec) with a small turning radius (less than half a body length). Finally, the dynamic model has been validated with experimental data, in terms of steady-state forward speed and turning speed versus the flapping frequency.

A completely new compliant lightweight robot is presented with a kinematic loop and a highly flexible link. It is explained how to model such parallel robots accurately but still computationally efficient. The elastic deformations are described with the floating frame of reference approach. For the flexible components this allows to use linear finite element models, which can represent arbitrary geometries. These models are further reduced by modal truncation and a Component Mode Synthesis minimizing the number of elastic degrees of freedom, which is necessary for real-time control purposes. The obtained model of the underactuated robot is non-minimum phase for the end-effector as output. Thus, for the applied trajectory tracking controller which is based on servo constraints, the concept of stable inversion is used. The performance is compared to a relocated minimum phase output. Corresponding simulations are validated by first experimental results showing the need for and high accuracy of the flexible model and the trajectory tracking control.

Research on legged robots has developed rapidly in the recent decades. One-legged robots, unlike multi-legged ones, have only one type of motion, called hopping. Hopping motion is generally divided into stance and flight phases. Switching between these two phases represents a hybrid dynamic model. Dynamic stabilization of hopping motion is a challenging control issue. Most of one-legged hopping robots studied in the past are able to hop with their one springy leg. In this paper, a novel one-legged robot is introduced and studied with two springs on the two sides. The one-legged somersaulting robot is able to hop with both springy sides. This ability causes lower energy consumption in passing obstacles and a longer step length in comparison with well-known SLIP robots with hopping motion stemming from the fact that it only has one rotary actuator. Fuzzy logic control is applied to achieve a stable limit cycle in the robot's somersaulting motion.

This paper presents dynamic modeling and control of an articulated robotic tail to maneuver and stabilize a reduced degree-of-freedom (DOF) quadruped robot. Conventional legged robotic systems consist of leg mechanisms that provide simultaneous propulsion, maneuvering and stabilization. However, in nature animals have been observed to utilize their tails to assist the legs in multiple tasks. Similarly, by incorporating an articulated tail onboard a quadruped robot, dynamic tail motions can be used to aid maneuvering. Therefore, tail implementation can potentially lead to simplifications in design and control of the legged robot since the legs will be responsible for only propulsion tasks. In this paper, a robotic system design consisting of an articulated tail and quadruped robot system is presented. Dynamic models are derived to analyze an optimal tail mass and length ratio to enhance inertial adjustment applications and develop an outer loop controller to plan tail trajectories for desired maneuvering applications. Results of analytical optimization are corroborated with measured data from biological animals. To decouple the dynamics of the articulated tail mechanism an inner loop controller using feedback linearization maps the desired behavior to the actuator inputs. This approach is validated using hardware-in-the-loop experiments with tail prototype in conjunction with simulated quadruped platform. Results demonstrate the capabilities of the articulated tail in enabling precise left and right turning (maneuvering).

The MAGMaS (Multiple Aerial-Ground Manipulator System) was proposed in [1] as a heterogeneous system composed of multiple ground (mobile) manipulators and aerial robots to collaboratively manipulate a long/large-sized object and demonstrated therein for rigid load manipulation. Here, we extend this result of [1] to the case of load manipulation with flexibility, which is crucial for long/slender object manipulation, yet, not considered in [1]. We first provide a rigorous modeling of the load flexibility and its effects on the MAGMaS dynamics. We then propose a novel collaborative control framework for flexible load-tip pose tracking, where the ground manipulator provides slower nominal pose tracking with overall load weight holding, whereas the aerial robot allows for faster vibration suppression with some load weight sharing. We also discuss the issue of controllability stemming from that the aerial robot provides less number of actuation than the modes of the load flexibility; and elucidate some peculiar conditions for this vibration suppression controllability. Simulations are also performed to demonstrate the effectiveness of the proposed theory.

When a human supervisor collaborates with a team of robots, the human's attention is divided, and cognitive resources are at a premium. We aim to optimize the distribution of these resources and the flow of attention. To this end, we propose the model of an idealized supervisor to describe human behavior. Such a supervisor employs a potentially inaccurate internal model of the the robots' dynamics to judge safety. We represent these safety judgements by constructing a safe set from this internal model using reachability theory. When a robot leaves this safe set, the idealized supervisor will intervene to assist, regardless of whether or not the robot remains objectively safe. False positives, where a human supervisor incorrectly judges a robot to be in danger, needlessly consume supervisor attention. In this work, we propose a method that decreases false positives by learning the supervisor's safe set and using that information to govern robot behavior. We prove that robots behaving according to our approach will reduce the occurrence of false positives for our idealized supervisor model. Furthermore, we empirically validate our approach with a user study that demonstrates a significant (p = 0.0328) reduction in false positives for our method compared to a baseline safety controller.

In this paper, we propose a method to model social interaction between a human and a virtual avatar. To this end, two human performers fist perform social interactions according to the Learning from Demonstration paradigm. Then, the relative relevance of all joints of both performers should be reasonably modeled based on human demonstrations. However, among all possible combinations of relative joints, it is necessary to select only some of the combinations that play key roles in social interaction. We select such significant features based on the joint motion significance, which is a metric to measure the significance degree by calculating both temporal entropy and spatial entropy of all human joints from a Gaussian mixture model. To evaluate our proposed method, we performed experiments on five social interactions: hand shaking, hand slapping, shoulder holding, object passing, and target kicking. In addition, we compared our method to existing modeling methods using different metrics, such as principal component analysis and information gain.

This paper addresses the problem of grasping arbitrarily shaped objects, observed as partial point-clouds, without requiring: models of the objects, physics parameters, training data, or other a-priori knowledge. A grasp metric is proposed based on Local Contact Moment (LoCoMo). LoCoMo combines zero-moment shift features, of both hand and object surface patches, to determine local similarity. This metric is then used to search for a set of feasible grasp poses with associated grasp likelihoods. LoCoMo overcomes some limitations of both classical grasp planners and learning-based approaches. Unlike force-closure analysis, LoCoMo does not require knowledge of physical parameters such as friction coefficients, and avoids assumptions about fingertip contacts, instead enabling robust contacts of large areas of hand and object surface. Unlike more recent learning-based approaches, LoCoMo does not require training data, and does not need any prototype grasp configurations to be taught by kinesthetic demonstration. We present results of real-robot experiments grasping 21 different objects, observed by a wrist-mounted depth camera. All objects are grasped successfully when presented to the robot individually. The robot also successfully clears cluttered heaps of objects by sequentially grasping and lifting objects until none remain.

A concept consisting of a new configurable vacuum gripper system and a corresponding method for determining optimal grasp configurations solely based on 3D vision is introduced. The robot system consists of a dynamically configurable vacuum gripper, a visual sensor, and a robot arm that are used in combination with a new grasp planner to robustly grasp unknown objects in arbitrary positions. For this purpose, formalized aspects of selecting contact surfaces for arbitrary suction cups are described; the concept involves visual detection of the objects, segmentation, iterative grasp planning, and action execution. The approach allows for a fast and efficient, yet precise execution of grasps. The core idea is a two-step 3D data acquisition approach and grasp point computation that takes advantage of the fact that the suction cups of the gripper can all be aligned axis-parallel. Therefore, an adequate sensor-based surface acquisition is done from a single viewpoint with respect to the gripper. Results of realworld experiments show that the proposed concept is suitable for a wide range of different and unknown objects in our setup.

In this paper, we propose a method and associate platform to couple model-based system engineering and safety analysis at the early phases of robotic system (RS) life-cycle. The method is compatible with IEC12100 and ISO13482. The platform is based on Papyrus UML modeler and supports RobotML, a domain specific language for RSs, as well as tools for safety analysis and risk assessment, Sophia and Safety Architect. It includes an ability (a) to model architecture of RSs; (b) to automatically run safety analysis (e.g. failure mode and effects analysis, fault tree analysis, etc.); (c) to save and reuse safety artefacts; (d) to represent safety analysis results in the modeling environment. We illustrate the proposed method by considering a humanoid personal care robot from SoftBank Robotics developed in the scope of the ROMEO2 project.

Deep reinforcement learning has achieved great strides in solving challenging motion control tasks. Recently, there has been significant work on methods for exploiting the data gathered during training, but there has been less work on how to best generate the data to learn from. For continuous action domains, the most common method for generating exploratory actions involves sampling from a Gaussian distribution centred around the mean action output by a policy. Although these methods can be quite capable, they do not scale well with the dimensionality of the action space, and can be dangerous to apply on hardware. We consider learning a forward dynamics model to predict the result, (xt+1), of taking a particular action, (u), given a specific observation of the state, (xt). With this model we perform internal lookahead predictions of outcomes and seek actions we believe have a reasonable chance of success. This method alters the exploratory action space, thereby increasing learning speed and enables higher quality solutions to difficult problems, such as robotic locomotion and juggling.

We propose a novel approach for optimal trajectory tracking for unmanned aerial vehicles (UAV), using a linear model predictive controller (MPC) in combination with non-linear state feedback. The solution relies on fast onboard simulation of the translational dynamics of the UAV, which is guided by a linear MPC. By sampling the states of the virtual UAV, we create a control command for fast non-linear feedback, which is capable of performing agile maneuvers with high precision. In addition, the proposed pipeline provides an interface for a decentralized collision avoidance system for multi-UAY scenarios. Our solution makes use of the long prediction horizon of the linear MPC and allows safe outdoors execution of multi-UAV experiments without the need for in-advance collision-free planning. The practicality of the tracking mechanism is shown in combination with priority-based collision resolution strategy, which performs sufficiently in experiments with up to 5 UAVs. We present a statistical and experimental evaluation of the platform in both simulation and real-world examples, demonstrating the usability of the approach.

Localization robustness against environment dynamics is significant for robots to achieve autonomous navigation in unmodified environments. A basic method of improving the robustness of a robot is considering the sensor observations obtained from mapped obstacles and using them for localizing the robot's pose. This study proposes an observation model that considers the class of sensor observations, where “class” categorizes the sensor observations as those obtained from mapped and unmapped obstacles. In the proposed approach, the robot's pose and the class are estimated simultaneously. As a result, the robot's pose can be localized using the sensor observations obtained only from mapped obstacles. First, we evaluated the performance of the proposed approach using simulations. Further, we tested the proposed approach in a real-world mobile robot navigation competition, called “Tsukuba Challenge,” held in Japan. The robustness and effectiveness of the proposed approach against environment dynamics were verified from the experimental results.

Typical continuum robots, such as pneumatic and tendon-driven robots, have a restricted section length and require a large external component for pulleys and a compressor, making them unsuitable for locomotion. This paper presents a new mobile continuum robot design with virtually unlimited extensible sections. A driving unit, which has a mechanism similar to the rack-and-pinion, consists of three DC motors with gears, each of which moves each flexible tube. The rotation of the motor translates the flexible tube, which has a helical groove on the surface that meshes with the gear. The long flexible tube provides a large traveling distance as long as it does not buckle. The elongation and bending motion of each section may be controlled during operation by varying the speed of each flexible tube. This design not only allows the expansion of the robot to otherwise unreachable work areas but also improves the locomotion velocity by generating a large traveling distance of the flexible tubes. The most important point in this paper is to use multiple driving units for locomotion. Since all the driving units can be mounted on the same tubes, by increasing the number of them, the robot can take various forms without expanding its diameter. A preliminary prototype was built, and its crawling locomotion performance was tested using two operating sequences. The results indicate that earthworm-like locomotion can be achieved with good performance by elongating the sections even when the ground is slippery. The proposed design can be easily be rebuilt by anyone with access to a basic 3D printer.

Multistage decision policies provide useful control strategies in high-dimensional state spaces, particularly in complex control tasks. However, they exhibit weak performance guarantees in the presence of disturbance, model mismatch, or model uncertainties. This brittleness limits their use in high-risk scenarios. We present how to quantify the sensitivity of such policies in order to inform of their robustness capacity. We also propose a minimax iterative dynamic game framework for designing robust policies in the presence of disturbance/uncertainties. We test the quantification hypothesis on a carefully designed deep neural network policy; we then pose a minimax iterative dynamic game (iDG) framework for improving policy robustness in the presence of adversarial disturbances. We evaluate our iDG framework on a mecanum-wheeled robot, whose goal is to find a ocally robust optimal multistage policy that achieve a given goal-reaching task. The algorithm is simple and adaptable for designing meta-learning/deep policies that are robust against disturbances, model mismatch, or model uncertainties, up to a disturbance bound. Videos of the results are on the author's website: https://goo.gl/JhshTB, while the codes for reproducing our experiments are on github: https://goo.gl/3G2VBy. A self-contained environment for reproducing our results is on docker: https://goo.gl/Bo7MBe.

With the advent of polygonal maps finding their way into the navigational software of mobile robots, the Visibility Graph can be used to search for the shortest collision-free path. The nature of the Visibility Graph-based shortest path algorithms is such that first the entire graph is computed in a relatively time-consuming manner. Then, the graph can be searched efficiently any number of times for varying start and target state combinations with the A* or the Dijkstra algorithm. However, real-world environments are typically too dynamic for a map to remain valid for a long time. With the goal of obtaining the shortest path quickly in an ever changing environment, we introduce a rapid path finding algorithm-Minimal Construct-that discovers only a necessary portion of the Visibility Graph around the obstacles that actually get in the way. Collision tests are computed only for lines that seem heuristically promising. This way, shortest paths can be found much faster than with a state-of-the-art Visibility Graph algorithm and as our experiments show, even grid-based A* searches are outperformed in most cases with the added benefit of smoother and shorter paths.

To prioritize miniaturization, the actuators of micro robot hands are placed far from the end effectors, but such mechanisms restrict controllability and dexterity. We propose a miniature robot finger driven by a new micro linear ultrasonic motor as a key component for micro robot hands. It enables dexterous and multiple motions for micro hands used in limited spaces. In this paper, we build a new micro linear ultrasonic motor involving a cuboid stator with a side length of approximately 2 mm, making it one of the smallest linear motors. The micro linear ultrasonic motor prototype shows an output torque of approximately 7.75 mN at low voltage operation, which is sufficient force to handle tiny objects. The miniature finger, a closed-loop six-bar-linkage mechanism, is built by micro fabrication and connected to the motor prototype. The first demonstration of the miniature finger is shown under a high-speed camera with a high power lens.

We present a 70mg, 3cm wing-span, flapping wing aerial vehicle capable of generating up to 60mg of lift using an electromagnetic actuator with low-voltage input (≈5.5V). Its design is novel, with the actuation and transmission integrated into a single resonant mechanism, thus not requiring any small-linear-displacement amplifying stages seen in other works. It can produce ±45° wing strokes and ±45° wing plane rotations at 98Hz operation mimicking relevant insects at this size scale. With required input power of only 250mW, it is, to the best of our knowledge, the most energy efficient electromagnetic design at the sub-100mg scale reported to date, and an order of magnitude more efficient than all other electromagnetic works.