ICRA 2020 Papers

This paper tackles the problem of data abstraction in the context of 3D point sets. Our method classifies points into different geometric primitives, such as planes and cones, leading to a compact representation of the data. Being based on a semi-global Hough voting scheme, the method does not need initialization and is robust, accurate, and efficient. We use a local, low-dimensional parameterization of primitives to determine type, shape and pose of the object that a point belongs to. This makes our algorithm suitable to run on devices with low computational power, as often required in robotics applications. The evaluation shows that our method outperforms state-of-the-art methods both in terms of accuracy and robustness.

The high deformability and compliance of soft robots allow safer interaction with the environment. On the other hand, these advantages bring along controllability and predictability challenges which result in loss of force and stiffness output. Such challenges should be addressed in order to improve the overall functional performance and to meet the requirements of real-scenario applications. In this paper, we present a bidirectional in-plane manipulator which consists of two unidirectional fiber-reinforced actuators (FRAs) and a hybrid soft-rigid stiffness control structure (SCS), all of them controlled by air pressure. Both controllability and predictability of the manipulator are enhanced by the hybrid soft-rigid structure. While the FRAs provide positioning and position dependent stiffness, the SCS increases the stiffness of the manipulator without position dependency. The SCS is able to increase the manipulator stiffness by 35%, 30%, and 18%, when one FRA is pressurized at 150 kPa, 75 kPa, and 0 kPa, respectively. Experiments are carried out to present the feasibility of the proposed manipulator.

This paper presents a preliminary study of an Aerial Manipulator suspended by a spring to a robotic carrier. The suspended aerial manipulator is actuated by six pairs of contra-rotating propellers generating a 6-DoF wrench. Simulations show path following results using a computed torque (feedback linearization) control strategy. Active vibration canceling is validated experimentally on a first prototype.

In this paper, we present a novel factor graph formulation to estimate the pose and velocity of a quadruped robot on slippery and deformable terrain. The factor graph introduces a preintegrated velocity factor that incorporates velocity inputs from leg odometry and also estimates related biases. From our experimentation we have seen that it is difficult to model uncertainties at the contact point such as slip or deforming terrain, as well as leg flexibility. To accommodate for these effects and to minimize leg odometry drift, we extend the robot's state vector with a bias term for this preintegrated velocity factor. The bias term can be accurately estimated thanks to the tight fusion of the preintegrated velocity factor with stereo vision and IMU factors, without which it would be unobservable. The system has been validated on several scenarios that involve dynamic motions of the ANYmal robot on loose rocks, slopes and muddy ground. We demonstrate a 26% improvement of relative pose error compared to our previous work and 52% compared to a state-of-the-art proprioceptive state estimator.

This paper presents a personalized gait optimization framework for lower-body exoskeletons. Rather than optimizing numerical objectives such as the mechanical cost of transport, our approach directly learns from user prefer-ences, e.g., for comfort. Building upon work in preference-based interactive learning, we present the CoSpar algorithm. CoSpar prompts the user to give pairwise preferences between trials and suggest improvements; as exoskeleton walking is a non-intuitive behavior, users can provide preferences more easily and reliably than numerical feedback. We show that CoSpar performs competitively in simulation and demonstrate a prototype implementation of CoSpar on a lower-body exoskeleton to optimize human walking trajectory features. In the experiments, CoSpar consistently found user-preferred parameters of the exoskeleton’s walking gait, which suggests that it is a promising starting point for adapting and personalizing exoskeletons (or other assistive devices) to individual users.

We propose in this paper Periodic Interaction Primitives - a probabilistic framework that can be used to learn compact models of periodic behavior. Our approach extends existing formulations of Interaction Primitives to periodic movement regimes, i.e., walking. We show that this model is particularly well-suited for learning data-driven, customized models of human walking, which can then be used for generating predictions over future states or for inferring latent, biomechanical variables. We also demonstrate how the same framework can be used to learn controllers for a robotic prosthesis using an imitation learning approach. Results in experiments with human participants indicate that Periodic Interaction Primitives efficiently generate predictions and ankle angle control signals for a robotic prosthetic ankle, with MAE of 2.21° in 0.0008s per inference. Performance degrades gracefully in the presence of noise or sensor fall outs. Compared to alternatives, this algorithm functions 20 times faster and performed 4.5 times more accurately on test subjects.

Human movement prediction is difficult as humans naturally exhibit complex behaviors that can change drastically from one environment to the next. In order to alleviate this issue, we propose a prediction framework that decouples short-term prediction, linked to internal body dynamics, and long-term prediction, linked to the environment and task constraints. In this work we investigate encoding short-term dynamics in a recurrent neural network, while we account for environmental constraints, such as obstacle avoidance, using gradient-based trajectory optimization. Experiments on real motion data demonstrate that our framework improves the prediction with respect to state-of-the-art motion prediction methods, as it accounts to beforehand unseen environmental structures. Moreover we demonstrate on an example, how this framework can be used to plan robot trajectories that are optimized to coordinate with a human partner.

Gait training is widely used to treat gait abnormalities. Traditional gait measurement systems are limited to instrumented laboratories. Even though gait measurements can be made in these settings, it is challenging to estimate gait parameters robustly in real-time for gait rehabilitation, especially when walking over-ground. In this paper, we present a novel approach to track the continuous gait cycle during overground walking outside the laboratory. In this approach, we instrument standard footwear with a sensorized insole and an inertial measurement unit. Artificial neural networks are used on the raw data obtained from the insoles and IMUs to compute the continuous percentage of the gait cycle for the entire walking session. We show in this paper that when tested with novel subjects, we can predict the gait cycle with a Root Mean Square Error (RMSE) of 7.2%. The onset of each cycle can be detected within an RMSE time of 41.5 ms with a 99% detection rate. The algorithm was tested with 18840 strides collected from 24 adults. In this paper, we tested a combination of fully-connected layers, an Encoder-Decoder using convolutional layers, and recurrent layers to identify an architecture that provided the best performance.

A common approach for defining a reward function for multi-objective reinforcement learning (MORL) problems is the weighted sum of the multiple objectives. The weights are then treated as design parameters dependent on the expertise (and preference) of the person performing the learning, with the typical result that a new solution is required for any change in these settings. This paper investigates the relationship between the reward function and the optimal value function for MORL; specifically addressing the question of how to approximate the optimal value function well beyond the set of weights for which the optimization problem was actually solved, thereby avoiding the need to recompute for any particular choice. We prove that the value function transforms smoothly given a transformation of weights of the reward function (and thus a smooth interpolation in the policy space). A Gaussian process is used to obtain a smooth interpolation over the reward function weights of the optimal value function for three well-known examples: Gridworld, Objectworld and Pendulum. The results show that the interpolation can provide robust values for sample states and actions in both discrete and continuous domain problems. Significant advantages arise from utilizing this interpolation technique in the domain of autonomous vehicles: easy, instant adaptation of user preferences while driving and true randomization of obstacle vehicle behavior preferences during training.

This paper presents a method to incorporate ergonomics into the optimization of action sequences for bi-manual human-robot cooperation tasks with continuous physical interaction. Our first contribution is a novel computational model of the human that allows prediction of an ergonomics assessment corresponding to each step in a task. The model is learned from human motion capture data in order to predict the human pose as realistically as possible. The second contribution is a combination of this prediction model with an informed graph search algorithm, which allows computation of human-robot cooperative plans with improved ergonomics according to the incorporated method for ergonomic assessment. The concepts have been evaluated in simulation and in a small user study in which the subjects manipulate a large object with a 32 DoF bimanual mobile robot as partner. For all subjects, the ergonomic-enhanced planner shows their reduced ergonomic cost compared to a baseline planner.

Understanding the effects of actions is essential for planning and executing robot tasks. By imagining possible action consequences, a robot can choose specific action parameters to achieve desired goal states. We present an approach for parametrizing pushing actions based on learning internal prediction models. These pushing actions must fulfill constraints given by a high-level planner, e. g., after the push the brown box must be to the right of the orange box. In this work, we represent the perceived scenes as object-centric graphs and learn an internal model, which predicts object pose changes due to pushing actions. We train this internal model on a large synthetic data set, which was generated in simulation, and record a smaller data set on the real robot for evaluation. For a given scene and goal state, the robot generates a set of possible pushing action candidates by sampling the parameter space and then evaluating the candidates by internal simulation, i. e., by comparing the predicted effect resulting from the internal model with the desired effect provided by the high-level planner. In the evaluation, we show that our model achieves high prediction accuracy in scenes with a varying number of objects and, in contrast to state-of-the-art approaches, is able to generalize to scenes with more objects than seen during training. In experiments on the humanoid robot ARMAR-6, we validate the transfer from simulation and show that the learned internal model can be used to manipulate scenes into desired states effectively.

Horizontally scanning 2D laser rangefinders are a popular approach for indoor robot localization because of the high accuracy of the sensors and the compactness of the required 2D maps. As the scanners in this configuration only provide information about one slice of the environment, the measurements typically do not capture the full extent of a large variety of obstacles, including chairs or tables. Accordingly, obstacle avoidance based on laser scanners mounted in such a fashion is likely to fail. In this paper, we propose a learning-based approach to predict collisions in 2D occupancy maps. Our approach is based on a convolutional neural network which is trained on a 2D occupancy map and collision events recorded with a bumper while the robot is navigating in its environment. As the network operates on local structures only, it can generalize to new environments. In addition, the robot can collect and integrate new collision examples after an initial training phase. Extensive experiments carried out in simulation and a realistic real-world environment confirm that our approach allows robots to learn from collision events to avoid collisions in the future.

This work provides a framework for the 3D calibration of wafers and a wafer handling robot by monocular vision. The proposed method precisely reconstructs the 3D poses of wafers from a set of images captured by the camera mounted on the robot. In addition, it calibrates the robot kinematics simultaneously. A robust ellipse detection and tracking algorithm based on the edge arcs is developed to recognize wafers among images. Then a joint optimization is constructed from a multi-object pose graph to solve the 3D poses of wafers and other calibration parameters of the robot-camera system. The proposed tracking method is able to associate multiple incomplete elliptic segments using a Gaussian Mixture Model-based registration algorithm. The algorithm is point-based where no feature descriptor is required. The proposed 3D pose optimization incorporates shape constraints, and is more accurate than the point-wise reconstruction produced by classic bundle adjustment methods.

Many existing SLAM approaches rely on the assumption of static environments for accurate performance. However, several robot applications require them to traverse repeatedly in semi-static or dynamic environments. There has been some recent research interest in designing persistence filters to reason about persistence in such scenarios. Our goal in this work is to incorporate such persistence reasoning in visual SLAM. To this end, we incorporate persistence filters [1] into ORB-SLAM, a well-known visual SLAM algorithm. We observe that the simple integration of their proposal results in inefficient persistence reasoning. Through a series of modifications and using two locally collected datasets, we demonstrate the utility of such persistence filtering as well as our customizations in ORB-SLAM. Overall, incorporating persistence filtering could result in a significant reduction in map size (about 30% in the best case) and a corresponding reduction in run-time while retaining similar accuracy to methods that use much larger maps.

Posture control for legged robots has been widely developed on custom-designed robotic platforms, with little work being done on commercially-available robots despite their potential as low-cost research platforms. This paper presents the implementation of a Walking Posture Control system on a commercially-available hexapod robot which utilizes low-cost joint actuators without torque control capabilities. The hierarchical control system employs Virtual Model Control with simple foot force distribution and a novel, position-based Foot Force Controller that enables direct force control during the leg's stance phase and active compliance control during the swing phase. Ground truth measurements in experimental tests, obtained with a Vicon motion capture system, demonstrate the improvement to posture made by the control system on uneven terrain, with the results comparing favorably to those obtained in similar tests on more sophisticated, custom-designed platforms.

Collision checking is a computational bottleneck in motion planning, requiring lazy algorithms that explicitly reason about when to perform this computation. Optimism in the face of collision uncertainty minimizes the number of checks before finding the shortest path. However, this may take a prohibitively long time to compute, with no other feasible paths discovered during this period. For many real-time applications, we instead demand strong anytime performance, defined as minimizing the cumulative lengths of the feasible paths yielded over time. We introduce Posterior Sampling for Motion Planning (PSMP), an anytime lazy motion planning algorithm that leverages learned posteriors on edge collisions to quickly discover an initial feasible path and progressively yield shorter paths. PSMP obtains an expected regret bound of Õ(√(SAT)) and outperforms comparative baselines on a set of 2D and 7D planning problems.

This paper presents the design of a compliant manipulator using a series elastic actuator (SEA) and a mechanism for precisely measuring the force acting on the contact part of the manipulator without using a force sensor. It is important to maintain a constant contact force between the compliant manipulator and the wall in order to guarantee cleaning performance, and the ball screw mechanism is used to adapt to changes in the distance and the angle. Position-based impedance control is used to maintain a constant contact force when the manipulator interacts with the wall of the building, and the results confirm that the system stability is guaranteed when using SEA, regardless of the variation in the actual stiffness of the manipulator. The results of extensive experimentation using the test bench demonstrate the force tracking performance against various types of wall changes using the stiff wet-type cleaning manipulator. The results indicate that the stiffness of SEA affects the force tracking performance and system stability under the condition of the manipulator and environment interaction, and that the system stability and control performance can be improved by applying a robust force measurement mechanism to noise.

Accurately estimating the 6-DoF object pose from a single RGB image is a challenging task in computer vision. Though pose regression approaches have achieved great progress, the performance is still limited. In this work, we propose Pose-guided Auto-Encoder (PAE), which can distill better pose-related features from the image by utilizing a suitable pose representation, 3D Location Field (3DLF), to guide the encoding process. The features from PAE show strong robustness to pose-irrelevant factors. Compared with traditional auto-encoder, PAE can not only improve the pose estimation performance but also handle the ambiguity viewpoints problem. Further, we propose Feature-based Pose Refiner (FPR), which refines the pose from the extracted features without rendering. Combining PAE with FPR, our approach achieved state-of-the-art performance on the widely used LINEMOD dataset. Our approach not only outperforms the direct regression-based approaches with a large margin but also thrillingly surpasses current state-of-the-art indirect PnP-based approach.

In this paper, we present a method for pose-estimate-based target tracking (PBTT) that enables the performance of autonomous aerial manipulation operations in unstructured environments using fully on-board computation for both UAV localization and target tracking. The PBTT method does not depend on extracting traditional visual features (e.g. using SIFT, SURF, ORB, etc.) on or near the target. Instead, the algorithm combines input from an RGB-D camera and the UAV's position estimator (which utilizes a downward-facing optical flow camera for horizontal localization) to track a target point selected by a human operator. The effectiveness of the PBTT method is evaluated through several autonomous flight tests performed with the Interacting-Boomcopter (I-BC) UAV platform in unstructured environments and in the presence of light wind disturbances.