ICRA 2019 论文列表

In this paper, we propose a novel method for underwater robot-to-human communication using the motion of the robot as “body language”. To evaluate this system, we develop simulated examples of the system's body language gestures, called kinemes, and compare them to a baseline system using flashing colored lights through a user study. Our work shows evidence that motion can be used as a successful communication vector which is accurate, easy to learn, and quick enough to be used, all without requiring any additional hardware to be added to our platform. We thus contribute to “closing the loop” for human-robot interaction underwater by proposing and testing this system, suggesting a library of possible body language gestures for underwater robots, and offering insight on the design of nonverbal robot-to-human communication methods.

Recent advances in 3D printing and manufacturing of miniaturized robotic hardware and computing are paving the way to build inexpensive and disposable robots. This will have a large impact on several applications including scientific discovery (e.g., hurricane monitoring), search-and-rescue (e.g., operation in confined spaces), and entertainment (e.g., nano drones). The need for inexpensive and task-specific robots clashes with the current practice, where human experts are in charge of designing hardware and software aspects of the robotic platform. This makes the robot design process expensive and time consuming, and ultimately unsuitable for small-volumes low-cost applications. This paper considers the computational robot co-design problem, which aims to create an automatic algorithm that selects the best robotic modules (sensing, actuation, computing) in order to maximize the performance on a task, while satisfying given specifications (e.g., maximum cost of the resulting design). We propose a binary optimization formulation of the co-design problem and show that such formulation generalizes previous work based on strong modeling assumptions. We show that the proposed formulation can solve relatively large co-design problems in seconds and with minimal human intervention. We demonstrate the proposed approach in two applications: the co-design of an autonomous drone racing platform and the co-design of a multi-robot system.

Autonomous service robots require computational frameworks that allow them to generalize knowledge to new situations in a manner that models uncertainty while scaling to real-world problem sizes. The Robot Common Sense Embedding (RoboCSE) showcases a class of computational frameworks, multi-relational embeddings, that have not been leveraged in robotics to model semantic knowledge. We validate RoboCSE on a realistic home environment simulator (AI2Thor) to measure how well it generalizes learned knowledge about object affordances, locations, and materials. Our experiments show that RoboCSE can perform prediction better than a baseline that uses pre-trained embeddings, such as Word2Vec, achieving statistically significant improvements while using orders of magnitude less memory than our Bayesian Logic Network baseline. In addition, we show that predictions made by RoboCSE are robust to significant reductions in data available for training as well as domain transfer to MatterPort3D, achieving statistically significant improvements over a baseline that memorizes training data.

In this paper, we propose a road detection method with LiDAR-camera fusion in a novel conditional random field (CRF) framework to exploit both range and color information. In the LiDAR based part, a fast height-difference based scanning strategy is applied in the 2D LiDAR range-image domain and a dense road detection result in camera image domain can be obtained through geometric upsampling given the LiDAR-camera calibration parameters. In the camera based part, a fully convolutional network is applied in the camera image domain. Finally, we fuse the dense and binary road detection results from both LiDAR and camera in a single CRF framework. Experiments show that using a single thread of CPU, the proposed LiDAR based part can operate at a frequency of over 250Hz with sparse output in range image and 40Hz with dense result in camera image for the 64-beam Velodyne scanner. Our CRF fusion method achieves very promising road detection performance on the KITTI-Road dataset.

It is necessary to employ smart sensory systems in dynamic and mobile workspaces where industrial robots are mounted on mobile platforms. Such systems should be aware of flexible and non-stationary workspaces and able to react autonomously to changing situations. Building upon our previously presented RoPose-system [1], which employs a convolutional neural network architecture that has been trained on pure synthetic data to estimate the kinematic chain of an industrial robot arm system, we now present RoPose-Real. RoPose-Real extends the prior system with a comfortable and targetless extrinsic calibration tool, to allow for the production of automatically annotated datasets for real robot systems. Furthermore, we use the novel datasets to train the estimation network with real world data. The extracted pose information is used to automatically estimate the observing sensor pose relative to the robot system. Finally we evaluate the performance of the presented subsystems in a real world robotic scenario.

Deep reinforcement learning has recently made significant progress in solving computer games and robotic control tasks. A known problem, though, is that policies overfit to the training environment and may not avoid rare, catastrophic events such as automotive accidents. A classical technique for improving the robustness of reinforcement learning algorithms is to train on a set of randomized environments, but this approach only guards against common situations. Recently, robust adversarial reinforcement learning (RARL) was developed, which allows efficient applications of random and systematic perturbations by a trained adversary. A limitation of RARL is that only the expected control objective is optimized; there is no explicit modeling or optimization of risk. Thus the agents do not consider the probability of catastrophic events (i.e., those inducing abnormally large negative reward), except through their effect on the expected objective. In this paper we introduce risk-averse robust adversarial reinforcement learning (RARARL), using a risk-averse protagonist and a risk-seeking adversary. We test our approach on a self-driving vehicle controller. We use an ensemble of policy networks to model risk as the variance of value functions. We show through experiments that a risk-averse agent is better equipped to handle a risk-seeking adversary, and experiences substantially fewer crashes compared to agents trained without an adversary. Supplementary materials are available at https://sites.google.com/view/rararl.

We present a principled method for motion prediction via dynamic simulation for rigid bodies in intermittent contact with each other where the contact is assumed to be a planar non-convex contact patch. The planar non-convex contact patch can either be a topologically connected set or disconnected set. Such algorithms are useful in planning and control for robotic manipulation. Most work in rigid body dynamic simulation assume that the contact between objects is a point contact, which may not be valid in many applications. In this paper, by using the convex hull of the contact patch, we build on our recent work on simulating rigid bodies with convex contact patches, for simulating motion of objects with planar non-convex contact patches. We formulate a discrete-time mixed complementarity problem where we solve the contact detection and integration of the equations of motion simultaneously. Thus, our method is a geometrically-implicit method and we prove that in our formulation, there is no artificial penetration between the contacting rigid bodies. We solve for the equivalent contact point (ECP) and contact impulse of each contact patch simultaneously along with the state, i.e., configuration and velocity of the objects. We provide empirical evidence to show that our method can seamlessly capture transition between different contact modes like patch contact to multiple or single point contact during simulation.

Untethered magnetic microrobots hold the potential to penetrate hard-to-reach areas of the human body and to perform therapy in a controlled way. In the past decade, impressive advancements have been made in this field but the clinical adoption of magnetoresponsive microrobots is still hampered by safety issues. A tool appointed for magnetic microrobots retrieval within body fluids could enable a real paradigm change, fostering their clinical translation.By starting from the general problem to retrieve magnetic microrobots injected into the bloodstream, the authors introduce a magnetic capture model that allows to design retrieval tools for magnetic cores of different diameters (down to 10 nm) and in different environmental conditions (fluid speed up to 7 cms-1). The model robustness is demonstrated by the design and testing of a retrieval catheter. In its optimal configuration, the catheter includes 27 magnets and fits a 12 F catheter. The model provides a good prediction of capture efficiency for 250 nm magnetic particles (experimental data: 77.6%, model prediction: 65%) and a very good prediction for 500 nm particles (experimental data: 93.6%, model prediction: 94%). The results support the proposed model-based design approach, which can be extended to retrieve other magnetoresponsive agents from body compartments.

This paper describes a stable knee-stretched walking of a humanoid by the resolved viscoelasticity control (RVC). The RVC method resolves multiple viscoelasticities in task-space, including the center of mass viscoelasticity for balancing, into joint-space viscoelasticity. Although a robust and compliant motion was achieved by the RVC method in previous studies, the conventional knee-bent posture to avoid the kinematic singularity suffered large knee joint torque. In this study, we propose an extension of the RVC capable of the kinematic singularity. We demonstrate through simulations and experiments that the RVC method considering the singularity achieves a stable and human-like walking, reducing the knee joint torque and improving the energy efficiency.

Soft robots utilize compliant materials to perform motions and behaviors not typically achievable by rigid bodied systems. These materials and soft actuator fabrication methods have been leveraged to create multigait walking soft robots. However, soft materials are prone to failure, restricting the ability of soft robots to accomplish tasks. In this work we address the problem of generating reactive controllers for multigait walking soft robots that are resilient to actuator failure by applying methods of formal synthesis. We present a sensing-based abstraction for actuator performance, provide a framework for encoding multigait behavior and actuator failure in Linear Temporal Logic (LTL), and demonstrate synthesized controllers on a physical soft robot.

Conventional feedback control methods can solve various types of robot control problems very efficiently by capturing the structure with explicit models, such as rigid body equations of motion. However, many control problems in modern manufacturing deal with contacts and friction, which are difficult to capture with first-order physical modeling. Hence, applying control design methodologies to these kinds of problems often results in brittle and inaccurate controllers, which have to be manually tuned for deployment. Reinforcement learning (RL) methods have been demonstrated to be capable of learning continuous robot controllers from interactions with the environment, even for problems that include friction and contacts. In this paper, we study how we can solve difficult control problems in the real world by decomposing them into a part that is solved efficiently by conventional feedback control methods, and the residual which is solved with RL. The final control policy is a superposition of both control signals. We demonstrate our approach by training an agent to successfully perform a real-world block assembly task involving contacts and unstable objects.

Consider many particles actuated by a uniform global external field (e.g. gravitational or magnetic fields). This paper presents analytical results using workspace obstacles and global inputs to reshape such a group of particles. Shape control of many particles is necessary for conveying information, construction, and navigation. First we show how the particles' characteristic angle of repose can be used to reshape the particles by controlling angle of attack and the magnitude of the driving force. These can then be used to control the force and torque applied to a rectangular rigid body. Next, we examine the full set of stable, achievable mean and variance configurations for the shape of a particle group in two canonical environments: a square and a circular workspace. Finally, we show how workspaces with linear boundary layers can be used to achieve a more rich set of mean and variance configurations.

Hamilton-Jacobi (HJ) reachability provides a flexible framework for the verification of safety in robotic systems: it accounts for nonlinear system dynamics and provides safety-preserving controllers. However, computational scalability limits its direct application to systems of less than five continuous state dimensions. To alleviate this computational burden, system decomposition methods have been proposed; however, safety guarantees are lost in situations involving “leaking corners which arise when there are conflicting controls between subsystems. In this paper, a coupled HJ formulation is presented, which addresses leaking corners and guarantees safety, while incorporating dimensionality reduction. We demonstrate our method in two examples, one of which is a vehicle obstacle avoidance problem with a 5D car model, whose HJ computation was previously considered to be intractable.

Precise robotic manipulation skills are desirable in many industrial settings, reinforcement learning (RL) methods hold the promise of acquiring these skills autonomously. In this paper, we explicitly consider incorporating operational space force/torque information into reinforcement learning; this is motivated by humans heuristically mapping perceived forces to control actions, which results in completing high-precision tasks in a fairly easy manner. Our approach combines RL with force/torque information by incorporating a proper operational space force controller; where we also exploit different ablations on processing this information. Moreover, we propose a neural network architecture that generalizes to reasonable variations of the environment. We evaluate our method on the open-source Siemens Robot Learning Challenge, which requires precise and delicate force-controlled behavior to assemble a tight-fit gear wheel set.

Moving a human body or a large and bulky object may require the strength of whole arm manipulation (WAM). This type of manipulation places the load on the robot's arms and relies on global properties of the interaction to succeed- rather than local contacts such as grasping or non-prehensile pushing. In this paper, we learn to generate motions that enable WAM for holding and transporting of humans in certain rescue or patient care scenarios. We model the task as a reinforcement learning problem in order to provide a robot behavior that can directly respond to external perturbation and human motion. For this, we represent global properties of the robot-human interaction with topology-based coordinates that are computed from arm and torso positions. These coordinates also allow transferring the learned policy to other body shapes and sizes. For training and evaluation, we simulate a dynamic sea rescue scenario and show in quantitative experiments that the policy can solve unseen scenarios with differently-shaped humans, floating humans, or with perception noise. Our qualitative experiments show the subsequent transporting after holding is achieved and we demonstrate that the policy can be directly transferred to a real world setting.

The design of feedback controllers for bipedal robots is challenging due to the hybrid nature of its dynamics and the complexity imposed by high-dimensional bipedal models. In this paper, we present a novel approach for the design of feedback controllers using Reinforcement Learning (RL) and Hybrid Zero Dynamics (HZD). Existing RL approaches for bipedal walking are inefficient as they do not consider the underlying physics, often requires substantial training, and the resulting controller may not be applicable to real robots. HZD is a powerful tool for bipedal control with local stability guarantees of the walking limit cycles. In this paper, we propose a non traditional RL structure that embeds the HZD framework into the policy learning. More specifically, we propose to use RL to find a control policy that maps from the robot’s reduced order states to a set of parameters that define the desired trajectories for the robot’s joints through the virtual constraints. Then, these trajectories are tracked using an adaptive PD controller. The method results in a stable and robust control policy that is able to track variable speed within a continuous interval. Robustness of the policy is evaluated by applying external forces to the torso of the robot. The proposed RL framework is implemented and demonstrated in OpenAI Gym with the MuJoCo physics engine based on the well-known RABBIT robot model.

In autonomous racing, vehicles operate close to the limits of handling and a sensor failure can have critical consequences. To limit the impact of such failures, this paper presents the redundant perception and state estimation approaches developed for an autonomous race car. Redundancy in perception is achieved by estimating the color and position of the track delimiting objects using two sensor modalities independently. Specifically, learning-based approaches are used to generate color and pose estimates, from LiDAR and camera data respectively. The redundant perception inputs are fused by a particle filter based SLAM algorithm that operates in real-time. Velocity is estimated using slip dynamics, with reliability being ensured through a probabilistic failure detection algorithm. The sub-modules are extensively evaluated in real-world racing conditions using the autonomous race car gotthard driverless, achieving lateral accelerations up to 1. 7G and a top speed of 90km/h.

This paper presents a new methodology to quantify robot localization safety by evaluating integrity risk, a performance metric widely used in open-sky aviation applications that has been recently extended to mobile ground robots. Here, a robot is localized by feeding relative measurements to mapped landmarks into an Extended Kalman Filter while a sequence of innovations is evaluated for fault detection. The main contribution is the derivation of a sequential chi-squared integrity monitoring methodology that maintains constant computation requirements by employing a preceding time window and, at the same time, is robust against faults occurring prior to the window. Additionally, no assumptions are made on either the nature or shape of the faults because safety is evaluated under the worst possible combination of sensor faults.

This paper proposes a probabilistic framework for classification of evolving targets, leveraging the principles of recursive Bayesian estimation in a perception-oriented context. By implementing a Gaussian toroid prediction model of the perception target's evolution, the proposed recursive Bayesian classification (RBC) scheme provides probabilistically robust classification. Appropriate features are extracted from the target, which is then probabilistically represented in a belief space. This approach is capable of handling high-dimensional belief spaces, while simultaneously allowing for multi-Gaussian representation of belief without computational complexity that hinders real-time analysis. The proposed technique is validated over several parameter values by thousands of simulated experiments, where it is shown to outperform naıve classification when high observational uncertainty is present.