ICRA 2020 Papers

Robot grasping of deformable hollow objects such as plastic bottles and cups is challenging, as the grasp should resist disturbances while minimally deforming the object so as not to damage it or dislodge liquids. We propose minimal work as a novel grasp quality metric that combines wrench resistance and object deformation. We introduce an efficient algorithm to compute the work required to resist an external wrench for a manipulation task by solving a linear program. The algorithm first computes the minimum required grasp force and an estimation of the gripper jaw displacements based on the object's empirical stiffness at different locations. The work done by the jaws is the product of the grasp force and the displacements. Grasps requiring minimal work are considered to be of high quality. We collect 460 physical grasps with a UR5 robot and a Robotiq gripper. We consider a grasp to be successful if it completes the task without damaging the object or dislodging the content. Physical experiments suggest that the minimal work quality metric reaches 74.2% balanced accuracy, a metric that is the raw accuracy normalized by the number of successful and failed real-world grasps, and is up to 24.2% higher than classical wrench-based quality metrics.

In this paper, we address the problem of finding cost-efficient three-dimensional paths that satisfy the maximum allowed curvature and the pitch angle of the vehicle. For any given initial and final configurations, the problem is decoupled into finding the horizontal and vertical parts of the path separately. Although the individual paths are modeled as two-dimensional Dubins curves using closed-form solutions, the final 3D path is constructed using the proposed local optimization to find a cost-efficient solution. Moreover, based on the decoupled approach, we provide a lower bound estimation of the optimal path that enables us to determine the quality of the found heuristic solution. The proposed solution has been evaluated using existing benchmark instances and compared with state-of-the-art approaches. Based on the reported results and lower bounds, the proposed approach provides paths close to the optimal solution while the computational requirements are in hundreds of microseconds. Besides, the proposed method provides paths with fewer turns than others, which make them easier to be followed by the vehicle's controller.

We propose an online incremental learning approach through teleoperation which allows an operator to partially modify a learned model, whenever it is necessary, during task execution. Compared to conventional incremental learning approaches, the proposed approach is applicable for teleoperation-based teaching and it needs only partial demonstration without any need to obstruct the task execution. Dynamic authority distribution and kinesthetic coupling between the operator and the agent helps the operator to correctly perceive the exact instance where modification needs to be asserted in the agent's behaviour online using partial trajectory. For this, we propose a variation of the Expectation-Maximization algorithm for updating original model through mini batches of the modified partial trajectory. The proposed approach reduces human workload and latency for a rhythmic peg-in-hole teleoperation task where online partial modification is required during the task operation.

Clear Corneal Incision, a challenging step in cataract surgery, and important to the overall quality of the surgery. New surgeons usually spend one full year trying to perfect their incision, but even after such rigorous training deficient incisions can still occur. This paper proposes an autonomous robotic system for this self-sealing incision. A conventional ophthalmic microscope system with a monocular camera is utilized to capture the surgical scene, ascertain the robot's position, and estimate depth information. Kinematics with a remote centre of motion (RCM) is designed for a multi-axes robot to perform the incision route. The experimental results on ex-vivo porcine eyes show the autonomous Clear Corneal Incision has a stricter three-plane structure than a surgeon-made incision, which is closer to the ideal incision.

We propose an efficient method for monocular simultaneous localization and mapping (SLAM) that is capable of estimating metrically-scaled motion without additional sensors or hardware acceleration by integrating metric depth predictions from a neural network into a geometric SLAM factor graph. Unlike learned end-to-end SLAM systems, ours does not ignore the relative geometry directly observable in the images. Unlike existing learned depth estimation approaches, ours leverages the insight that when used to estimate scale, learned depth predictions need only be coarse in image space. This allows us to shrink our network to the point that performing inference on a standard CPU becomes computationally tractable.We make several improvements to our network architecture and training procedure to address the lack of depth observability when using coarse images, which allows us to estimate spatially coarse, but depth-accurate predictions in only 30 ms per frame without GPU acceleration. At runtime we incorporate the learned metric data as unary scale factors in a Sim(3) pose graph. Our method is able to generate accurate, scaled poses without additional sensors, hardware accelerators, or special maneuvers and does not ignore or corrupt the observable epipolar geometry. We show compelling results on the KITTI benchmark dataset in addition to real-world experiments with a handheld camera.

Modern reinforcement learning methods suffer from low sample efficiency and unsafe exploration, making it infeasible to train robotic policies entirely on real hardware. In this work, we propose to address the problem of sim-to-real domain transfer by using meta learning to train a policy that can adapt to a variety of dynamic conditions, and using a task-specific trajectory generation model to provide an action space that facilitates quick exploration. We evaluate the method by performing domain adaptation in simulation and analyzing the structure of the latent space during adaptation. We then deploy this policy on a KUKA LBR 4+ robot and evaluate its performance on a task of hitting a hockey puck to a target. Our method shows more consistent and stable domain adaptation than the baseline, resulting in better overall performance.

We present a parallel robot mechanism and the constitutive laws that govern the deformation of its constituent soft actuators. Our ultimate goal is the real-time motion-correction of a patient's head deviation from a target pose where the soft actuators control the position of the patient's cranial region on a treatment machine. We describe the mechanism, derive the stress-strain constitutive laws for the individual actuators and the inverse kinematics that prescribes a given deformation, and then present simulation results that validate our mathematical formulation. Our results demonstrate deformations consistent with our radially symmetric displacement formulation under a finite elastic deformation framework.

Tendon actuated multisection continuum arms have high potential for inspection applications in highly constrained spaces. They generate motion by axial and bending deformations. However, because of the high mechanical coupling between continuum sections, variable length-based kinematic models produce poor results. A new mechanics model for tendon actuated multisection continuum arms is proposed in this paper. The model combines the continuum arm curve parameter kinematics and concentric tube kinematics to correctly account for the large axial and bending deformations observed in the robot. Also, the model is computationally efficient and utilizes tendon tensions as the joint space variables thus eliminating the actuator length related problems such as slack and backlash. A recursive generalization of the model is also presented. Despite the high coupling between continuum sections, numerical results show that the model can be used for generating correct forward and inverse kinematic results. The model is then tested on a thin and long multisection continuum arm. The results show that the model can be used to successfully model the deformation.

This paper presents a robotic gripper design that can perform customizable grasping tasks at the millimeter scale. The design is based on the origami string, a mechanism with a single degree of freedom that can be mechanically programmed to approximate arbitrary paths in space. By using this concept, we create miniature fingers that bend at multiple joints with a single actuator input. The shape and stiffness of these fingers can be varied to fit different grasping tasks by changing the crease pattern of the string. We show that the experimental behavior of these strings follows their analytical models and that they can perform a variety of tasks including pinching, wrapping, and twisting common objects such as pencils, bottle caps, and blueberries.

This paper presents a study that evaluates the effects on the walking pattern of a full-sized humanoid robot as it pushes different carts. Furthermore, it discuss a modified Zero Moment Point (ZMP) pattern based on a capture point method, and a friction compensation method for the arms. Humanoid researchers have demonstrated that robots can perform a wide range of tasks including handling tools, climbing ladders, and patrolling rough terrain. However, when it comes to handling objects while walking, humanoids are relatively limited; it becomes more apparent when humanoids have to push a cart. Many challenges become evident under such circumstances; for example, the walking pattern will be severely affected by the external force opposed by the cart. Therefore, an appropriate walking pattern dynamic model and arm compliance are needed to mitigate external forces. This becomes crucial in order to ensure the robot's self-balance and minimize external disturbances.

We consider the problem of mapless collision-avoidance navigation where humans are present using 2D laser scans. Our proposed method uses ego-safety to measure collision from the robot's perspective and social-safety to measure the impact of robot's actions on surrounding pedestrians. Specifically, the social-safety part predicts the intrusion impact of the robot's action into the interaction area with surrounding humans. We train the policy using reinforcement learning on a simple simulator and directly evaluate the learned policy in Gazebo and real robot tests. Experiments show the learned policy smoothly transferred to different scenarios without any fine tuning. We observe that our method demonstrates time-efficient path planning behavior with high success rate in the mapless navigation task. Furthermore, we test our method in a navigation task among dynamic crowds, considering both low and high volume traffic. Our learned policy demonstrates cooperative behavior that actively drives our robot into traffic flows while showing respect to nearby pedestrians. Evaluation videos are at https://sites.google.com/view/ssw-batman.

Algorithms for motion planning in unknown environments are generally limited in their ability to reason about the structure of the unobserved environment. As such, current methods generally navigate unknown environments by relying on heuristic methods to choose intermediate objectives along frontiers. We present a unified method that combines map prediction and motion planning for safe, time-efficient au-tonomous navigation of unknown environments by dynamically-constrained robots. We propose a data-driven method for predicting the map of the unobserved environment, using the robot's observations of its surroundings as context. These map predictions are then used to plan trajectories from the robot's position to the goal without requiring frontier selection. We applied this map-predictive motion planning strategy to randomly generated winding hallway environments, yielding substantial improvement in trajectory duration over a naïve frontier pursuit method. We also experimentally demonstrate similar performance to methods using more sophisticated fron-tier selection heuristics while significantly reducing computation time.

This work presents a semantic map management approach for various environments by triggering multiple maps with different simultaneous localization and mapping (SLAM) configurations. A modular map structure allows to add, modify or delete maps without influencing other maps of different areas. The hierarchy level of our algorithm is above the utilized SLAM method. Evaluating laser scan data (e.g. the detection of passing a doorway) triggers a new map, automatically choosing the appropriate SLAM configuration from a manually predefined list. Single independent maps are connected by link-points, which are located in an overlapping zone of both maps, enabling global navigation over several maps. Loop- closures between maps are detected by an appearance-based method, using feature matching and iterative closest point (ICP) registration between point clouds. Based on the arrangement of maps and link-points, a topological graph is extracted for navigation purpose and tracking the global robot's position over several maps. Our approach is evaluated by mapping a university campus with multiple indoor and outdoor areas and abstracting a metrical-topological graph. It is compared to a single map running with different SLAM configurations. Our approach enhances the overall map quality compared to the single map approaches by automatically choosing predefined SLAM configurations for different environmental setups.

Accurate and robust global localization is essential to robotics applications. We propose a novel global localization method that employs the map traversability as a hidden observation. The resulting map-corrected odometry localization is able to provide an accurate belief tensor of the robot state. Our method can be used for blind robots in dark or highly reflective areas. In contrast to odometry drift in the long-term, our method using only odometry and the map converges in long-term. Our method can also be integrated with other sensors to boost the localization performance. The algorithm does not have any initial state assumption and tracks all possible robot states at all times. Therefore, our method is global and is robust in the event of ambiguous observations. We parallel each step of our algorithm such that it can be performed in real-time (up to ~300 Hz) using GPU. We validate our algorithm in different publicly available floor-plans and show that it is able to converge to the ground truth fast while being robust to ambiguities.

This paper presents a novel steering mechanism, which leads to frequency-controlled locomotion demonstrated for the first time in micro bristle robots. The miniaturized robots are 3D-printed, 12 mm × 8 mm × 6 mm in size, with bristle feature sizes down to 400 µm. The robots can be steered by utilizing the distinct resonance behaviors of the asymmetrical bristle sets. The left and right sets of the bristles have different diameters, and thus different stiffnesses and resonant frequencies. The unique response of each bristle side to the vertical vibrations of a single on-board piezoelectric actuator causes differential steering of the robot. The robot can be modeled as two coupled uniform bristle robots, representing the left and the right sides. At distinct frequencies, the robots can move in all four principal directions: forward, backward, left and right. Furthermore, the full 360◦ 2D plane can be covered by superimposing the principal actuation frequency components with desired amplitudes. In addition to miniaturized robots, the presented resonance-based steering mechanism can be applied over multiple scales and to other mechanical systems.

In this paper, we introduce novel miniature swimmers with multiple rigid tails based on spherical helices. The tail distribution of these prototypes enhances its swimming features as well as allowing to carry objects with it. The proposed swimmers are actuated by a rotating magnetic field, generating the robot rotation and thus producing a considerable thrust to start self-propelling. These prototypes achieved propulsion speeds up to 6 mm/s at 3.5 Hz for a 6-mm in size prototypes. We study the efficiency of different tail distribution for a 2-tailed swimmer by varying the angular position between both tails. Moreover, it is demonstrated that these swimmers experience great sensibility when changing their tail height. Besides, these swimmers demonstrate to be effective for cargo carrying tasks since they can displace objects up to 3.5 times their weight. Finally, wall effect is studied with multi-tailed swimmer robots considering 2 containers with 20 and 50-mm in width. Results showed speeds' increments up to 59% when swimmers are actuated in the smaller container.

A button battery accidentally ingested by a toddler or small child can cause severe damage to the stomach within a short period of time. Once a battery lands on the surface of the esophagus or stomach, it can run a current in the tissue and induce a chemical reaction resulting in injury. Following our previous work where we presented an ingestible magnetic robot for button battery retrieval, this study presents a remotely achieved novel localization method of a button battery with commonly available magnetic sensors (Hall-effect sensors). By applying a direct magnetic field to the button battery using an electromagnetic coil, the battery is magnetized, and hence it becomes able to be sensed by Hall-effect sensors. Using a trilateration method, we were able to detect the locations of an LR44 button battery and other ferromagnetic materials at variable distances. Additional four electromagnetic coils were used to autonomously navigate a magnet-containing capsule to dislocate the battery from the affected site.

The autonomous operation of robots in the International Space Station (ISS) is required to maximize the use of limited resources and enable astronauts to concentrate more on valuable tasks. To achieve these goals, we are developing a station where the robot approaches, docks, charges, and then undocks. In this paper, the authors propose a magnetic docking mechanism for free-flying robots with spherical surfaces that makes it possible for a robot to dock securely without requiring highly precise guidance, navigation and control capability. By making use of a slider guide and repelling pairs of magnets, this mechanism can achieve tolerance for larger robot position error as compared with the conventional fixed guide mechanism. The experimental results show that the proposed mechanism can effectively enlarge the acceptable error range of poses, and also reduce acceleration at the moment of impact. We also introduce a model to predict the success or failure of docking from the contact condition of the robot and the guide by using a machine learning technique - Gaussian Process Regression (GPR). The prediction results shows that the learnt model can express the contact condition of successful docking.

Understanding of characteristics of objects such as fragility, rigidity, texture and dimensions facilitates and innovates robotic grasping. In this paper, we propose a context- aware anthropomorphic robotic hand (MagicHand) grasping system which is able to gather various information about its target object and generate grasping strategies based on the perceived information. In this work, NIR spectra of target objects are perceived to recognize materials on a molecular level and RGB-D images are collected to estimate dimensions of the objects. We selected six most used grasping poses and our system is able to decide the most suitable grasp strategies based on the characteristics of an object. Through multiple experiments, the performance of the MagicHand system is demonstrated.