ICRA 2022 论文列表

An oft-ignored challenge of real-world reinforcement learning is that the real world does not pause when agents make learning updates. As standard simulated environments do not address this real-time aspect of learning, most available implementations of RL algorithms process environment interactions and learning updates sequentially. As a consequence, when such implementations are deployed in the real world, they may make decisions based on significantly delayed observations and not act responsively. Asynchronous learning has been proposed to solve this issue, but no systematic comparison between sequential and asynchronous reinforcement learning was conducted using real-world environments. In this work, we set up two vision-based tasks with a robotic arm, implement an asynchronous learning system that extends a previous architecture, and compare sequential and asynchronous reinforcement learning across different action cycle times, sensory data dimensions, and mini-batch sizes. Our experiments show that when the time cost of learning updates increases, the action cycle time in sequential implementation could grow excessively long, while the asynchronous implementation can always maintain an appropriate action cycle time. Consequently, when learning updates are expensive, the performance of sequential learning diminishes and is outperformed by asynchronous learning by a substantial margin. Our system learns in real-time to reach and track visual targets from pixels within two hours of experience and does so directly using real robots, learning completely from scratch. Our code is available at: https://github.com/YufengYuan/ur5_async_r1.

Event cameras open up new possibilities for robotic perception due to their low latency and high dynamic range. On the other hand, developing effective event-based vision algorithms that fully exploit the beneficial properties of event cameras remains work in progress. In this paper, we focus on event-based visual odometry (VO). While existing event-driven VO pipelines have adopted continuous-time representations to asynchronously process event data, they either assume a known map, restrict the camera to planar trajectories, or integrate other sensors into the system. Towards map-free event-only monocular VO in SE(3), we propose an asynchronous structure-from-motion optimisation back-end. Our formulation is underpinned by a principled joint optimisation problem involving non-parametric Gaussian Process motion modelling and incremental maximum a posteriori inference. A high-performance incremental computation engine is employed to reason about the camera trajectory with every incoming event. We demonstrate the robustness of our asynchronous back-end in comparison to frame-based methods which depend on accurate temporal accumulation of measurements.

Collaborative localization is an essential capability for a team of robots such as connected vehicles to collaboratively estimate object locations from multiple perspectives with reliant cooperation. To enable collaborative localization, four key challenges must be addressed, including modeling complex relationships between observed objects, fusing observations from an arbitrary number of collaborating robots, quantifying localization uncertainty, and addressing latency of robot communications. In this paper, we introduce a novel approach that integrates uncertainty-aware spatiotemporal graph learning and model-based state estimation for a team of robots to collaboratively localize objects. Specifically, we introduce a new uncertainty-aware graph learning model that learns spatiotemporal graphs to represent historical motions of the objects observed by each robot over time and provides uncertainties in object localization. Moreover, we propose a novel method for integrated learning and model-based state estimation, which fuses asynchronous observations obtained from an arbitrary number of robots for collaborative localization. We evaluate our approach in two collaborative object localization scenarios in simulations and on real robots. Experimental results show that our approach outperforms previous methods and achieves state-of-the-art performance on asynchronous collaborative localization.

We present AstroLoc, an efficient and robust monocular visual-inertial graph-based localization system used by the Astrobee free-flying robots onboard the International Space Station (ISS). We provide a novel localization system that limits the traditionally higher computation times for graph-based localization systems and enables the resource constrained Astrobee robots to benefit from their increased accuracy. We also introduce methods for handling cheirality issues for visual odometry and localization factors that further increase localization robustness. We evaluate the performance of AstroLoc on a dataset of ISS activities and show that it greatly improves pose, velocity, and IMU bias estimation accuracy while efficiently running in a limited computation environment. AstroLoc has improved the localization accuracy for the Astrobee robots on the ISS and has led to more successful and longer duration activities. While the AstroLoc system is tuned for the Astrobee robots, it can be configured for any resource constrained platform. The source code for AstroLoc is released to the public.

Operating an articulated machine is a complex and hierarchical task, involving several levels of decision making. Motivated by the timber-harvesting applications of these machines, we are interested in developing a collaborative framework for operating an articulated machine/robot in order to increase its level of autonomy. In this paper, we consider two problems in the context of collaborative operation of a feller-buncher: first, the problem of planning a sequence of cut/grasp/bunch tasks for the trees in the vicinity of the machine. Here we propose a human-inspired planning algorithm based on our observations of the operators in the field. Then, a Markov Decision Process (MDP) framework is provided, which enables us to obtain an optimal sequence of tasks. We provide numerical illustrations of how our MDP framework works. Second is the problem of inferring the operator's goal from the motions of the machine. The goal inference algorithm presented here enables the robot equipped with the planning intelligence to perceive the human's intent in real-time. We evaluate the performance of our goal inference algorithm through a user-study with a feller-buncher simulator. The results show the benefits of our algorithm over a robot that assumes the human is moving to the closest target.

Finding relative pose for cameras is of vital importance in computer vision and robotics. We investigate the problem of relative motion estimation between successive frames from a minimal number of correspondences. Existing approximated methods use a first-order approximation to relative pose in order to simplify the problem and produce an estimate quickly. Our solution uses Cayley parameterization to represent rotation and simplifies the high-degree polynomials only at the very end of the formulation, resulting in more accurate models. Furthermore, our method can be more effective if the camera rotates mainly around one coordinate axis. By treating the main rotation component as the hidden variable in the solution, we can retain more high-degree terms for the main rotation part, considerably widening the effective approximation range. Our experiments show that our method is more accurate than existing approximated solver, and that it is still effective for relatively large motions. Besides, our method produces far fewer solutions than essential matrix parameterized solvers.

Pavement cleaning is a labor-intensive, repetitive task and can be automated. Several autonomous pavement cleaning robots have been developed, pushing research towards their design and autonomous capabilities. Advances in design have been reported in earlier works on a self-reconfigurable robot with four independent steering drive (4ISD) capabilities, Panthera, for pavement cleaning and maintenance. Moreover, autonomous navigation requires sharp turns, heading angle adjustments, sideways movement, and locomotion without col-lision through constrained pavement conditions. The present work proposes an algorithm to ingeniously select the instan-taneous center of rotation (ICR) within the self-reconfigurable robot footprint and perform static rotation to adjust its heading angle during the waypoint navigation while avoiding collision with the constrained environment. Finally, the proposed algorithm is implemented, and experiments are conducted in real-world pavement scenarios. The experimental outcome success-fully demonstrates the self-reconfigurable robot's capability to navigate constrained pavement scenarios using the proposed algorithm during autonomous cleaning and maintenance tasks.

We focus on the problem of analyzing multiagent interactions in traffic domains. Understanding the space of behavior of real-world traffic may offer significant advantages for algorithmic design, data-driven methodologies, and bench-marking. However, the high dimensionality of the space and the stochasticity of human behavior may hinder the identification of important interaction patterns. Our key insight is that traffic environments feature significant geometric and temporal structure, leading to highly organized collective behaviors, often drawn from a small set of dominant modes. In this work, we propose a representation based on the formalism of topological braids that can summarize arbitrarily complex multiagent behavior into a compact object of dual geometric and symbolic nature, capturing critical events of interaction. This representation allows us to formally enumerate the space of outcomes in a traffic scene and characterize their complexity. We illustrate the value of the proposed representation in summarizing critical aspects of real-world traffic behavior through a case study on recent driving datasets. We show that despite the density of real-world traffic, observed behavior tends to follow highly organized patterns of low interaction. Our framework may be a valuable tool for evaluating the richness of driving datasets, but also for synthetically designing balanced training datasets or benchmarks.

Previous incremental estimation methods consider estimating a single line, requiring as many observers as the number of lines to be mapped. This leads to the need for having at least 4N state variables, with N being the number of lines. This paper presents the first approach for multi-line incremental estimation. Since lines are common in structured environments, we aim to exploit that structure to reduce the state space. The modeling of structured environments proposed in this paper reduces the state space to 3N + 3 and is also less susceptible to singular configurations. An assumption the previous methods make is that the camera velocity is available at all times. However, the velocity is usually retrieved from odometry, which is noisy. With this in mind, we propose coupling the camera with an Inertial Measurement Unit (IMU) and an observer cascade. A first observer retrieves the scale of the linear velocity and a second observer for the lines mapping. The stability of the entire system is analyzed. The cascade is shown to be asymptotically stable and shown to converge in experiments with simulated data.

Visual navigation by mobile robots is classically tackled through SLAM plus optimal planning, and more recently through end-to-end training of policies implemented as deep networks. While the former are often limited to waypoint planning, but have proven their efficiency even on real physical environments, the latter solutions are most frequently employed in simulation, but have been shown to be able learn more complex visual reasoning, involving complex semantic regularities. Navigation by real robots in physical environments is still an open problem. End-to-end training approaches have been thoroughly tested in simulation only, with experiments involving real robots being restricted to rare performance evaluations in simplified laboratory conditions. In this work we present an in-depth study of the performance and reasoning capacities of real physical agents, trained in simulation and deployed to two different physical environments. Beyond benchmarking, we provide insights into the generalization capabilities of different agents training in different conditions. We visualize sensor usage and the importance of the different types of signals. We show, that for the PointGoal task, an agent pre-trained on wide variety of tasks and fine-tuned on a simulated version of the target environment can reach competitive performance without modelling any sim2real transfer, i.e. by deploying the trained agent directly from simulation to a real physical robot.

A pressing challenge in the design of actuated knee prostheses is the ability to address the high variation of speed and torque requirements for the different types and phases of locomotion. This manuscript presents a novel over-actuated knee prosthesis which makes use of a dual motor actuation architecture to address this issue. It utilizes a high speed/low torque motor to enable natural and highly dynamical motion, as required for swing phases of walking, which is permanently engaged. In addition to this motor, a clutchable uni-directional low dynamics high torque motor is present to assist during the execution of tasks which demand active torque. Preliminary experimental validations have been performed on a healthy subject provided with an able-bodied adapter to demonstrate natural walk patterns and power-assisted sit-to-stand activities.

Strategies for safe human-robot interaction (HRI), such as the well-established Safe Motion Unit, provide a velocity scaling for biomechanically safe robot motion. In addition, psychologically-based safety approaches are required for trustworthy HRI. Such schemes can be very conservative and robot motion complying with such safety approaches should be time efficient within the robot motion planning. In this study, we improve the efficiency of a previously introduced approach for psychologically-based safety in HRI via a Model Predictive Control robot motion planner that simultaneously adjusts Cartesian path and speed to minimise the distance to the target pose as fast as possible. A subordinate real-time motion generator ensures human physical safety by integrating the Safe Motion Unit. Our motion planner is validated by two experiments. The simultaneous adjustment of path and velocity accomplishes highly time efficient robot motion, while considering the human physical and psychological safety. Compared to direct path velocity scaling approaches our planner enables 28 % faster motion execution.

Traditional robotic manipulator design methods require extensive, time-consuming, and manual trial and error to produce a viable design. During this process, engineers often spend their time redesigning or reshaping components as they discover better topologies for the robotic manipula-tor. Tactile sensors, while useful, often complicate the design due to their bulky form factor. We propose an integrated design pipeline to streamline the design and manufacturing of robotic manipulators with knitted, glove-like tactile sensors. The proposed pipeline allows a designer to assemble a collection of modular, open-source components by applying predefined graph grammar rules. The end result is an intuitive design paradigm that allows the creation of new virtual designs of manipulators in a matter of minutes. Our framework allows the designer to fine-tune the manipulator's shape through cage-based geometry deformation. Finally, the designer can select surfaces for adding tactile sensing. Once the manipulator design is finished, the program will automatically generate 3D printing and knitting files for manufacturing. We demonstrate the utility of this pipeline by creating four custom manipulators tested on real-world tasks: screwing in a wing nut, pouring water from a bottle, picking up an egg, and cutting paper with scissors.

To date, untethered micro/nanorobots have attracted considerable attention in various aspects due to their unique potential for in-vivo applications such as the targeted therapy. One of the most promising types of micro/nanorobots is the class of ferromagnetic microrobots which can be efficiently actuated via gradient/rotational magnetic field generated by less costly electromagnetic coil systems. For performing successful operations, locomotion control of the magnetic microrobots is non-trivial. Modern controllers commonly require motion status-based feedback. To fully utilize those advanced approaches, motion state of one microrobot should be supplied, however it is still challenging in cases. It is noted that, during locomotion, one ferromagnetic microrobot can combine with others to form an unstructured larger one, namely growing magnetic microrobot (GMM), whose dynamic behavior keeps changing, and thus the model-based observers are never applicable. Besides, tracking and estimating states of those unstructured time-varying GMMs in complex surroundings are always challenging, especially for an uneven sampling scenario. In order to accurately estimate the GMM motion status in a complex environment via micro-vision, this study develops an indeterministic observer leveraging on the approach of discriminative correlation filter with channel/spatial reliability (CSR-DCF) and the variable-step finite-time sliding mode (FSM-V) state estimation theory. Experimental study verifies that the proposed observation scheme can effectively estimate motion states of one GMM moving in obstacle surroundings throughout.

This paper discusses a low-cost, open-source and open-hardware design and performance evaluation of a low-speed, multi-fan wind system dedicated to micro air vehicle (MAV) testing. In addition, a set of experiments with a flapping wing MAV and rotorcraft is presented, demonstrating the capabilities of the system and the properties of these different types of drones in response to various types of wind. We performed two sets of experiments where a MAV is flying into the wake of the fan system, gathering data about states, battery voltage and current. Firstly, we focus on steady wind conditions with wind speeds ranging from 0.5 m S-1 to 3.4 m S-1. During the second set of experiments, we introduce wind gusts, by periodically modulating the wind speed from 1.3 m S−1 to 3.4 m S−1 with wind gust oscillations of 0.5 Hz, 0.25 Hz and 0.125 Hz. The “Flapper” flapping wing MAV requires much larger pitch angles to counter wind than the “CrazyFlie” quadrotor. This is due to the Flapper's larger wing surface. In forward flight, its wings do provide extra lift, considerably reducing the power consumption. In contrast, the CrazyFlie's power consumption stays more constant for different wind speeds. The experiments with the varying wind show a quicker gust response by the CrazyFlie compared with the Flapper drone, but both their responses could be further improved. We expect that the proposed wind gust system will provide a useful tool to the community to achieve such improvements.

We present the first realization of an assembly sequence planning framework for large-scale and complex 3D real-world CAD scenarios. Other than in academic benchmark data sets, in our scenario each assembled part is allowed to contain flexible fastening elements and the number of assembled parts is quite high. With our framework we are able to derive a meaningful assembly priority graph for the parts. Our framework divides the disassembly motion of each part into a NEAR- and a subsequent FAR planning phase and uses existing specialized motion planners for each phase. To reduce the number of unsuccessful motion planning requests we use a general Voronoi diagram graph and a novel collision perceiving method which significantly speed up our framework. At the end, we create an assembly priority graph to indicate which parts must be disassembled before others. In our experiments, we show that our framework is the first one which is able to generate a priority graph for a representative data set from the automotive industry. Moreover, the reported disassembly motions for the individual parts are shorter and can be computed faster than with other state-of-the-art frameworks.

This paper presents a simple but compact design of a bicycle-like robot for inspecting complex-shaped ferromagnetic structures. The design concept for versatile locomotion relies on two independently steered magnetic wheels formed in a bicycle-like configuration, allowing the robot to possess multi-directional mobility. The key feature of a reciprocating mechanism enables the robot to change its shape when passing obstacles. A dynamic joint of the robot configuration makes it naturally adapt to uneven and complex surfaces of steel structures. We demonstrate the usability and practical deployment of the robot for steel thickness measurement using an ultrasonic sensor.

This paper develops an adaptive PID autotuner for multicopters, and presents simulation and experimental results. The autotuner consists of adaptive digital control laws based on retrospective cost adaptive control implemented in the PX4 flight stack. A learning trajectory is used to optimize the autopilot during a single flight. The autotuned autopilot is then compared with the default PX4 autopilot by flying a test trajectory constructed using the second-order Hilbert curve. In order to investigate the sensitivity of the autotuner to the quadcopter dynamics, the mass of the quadcopter is varied, and the performance of the autotuned and default autopilot is compared. It is observed that the autotuned autopilot outperforms the default autopilot.

Learning controllers that reproduce legged locomotion in nature has been a longtime goal in robotics and computer graphics. While yielding promising results, recent approaches are not yet flexible enough to be applicable to legged systems of different morphologies. This is partly because they often rely on precise motion capture references or elaborate learning environments that ensure the naturality of the emergent locomotion gaits but prevent generalization. This work proposes a generic approach for ensuring realism in locomotion by guiding the learning process with the spring-loaded inverted pendulum model as a reference. Leveraging on the exploration capacities of Reinforcement Learning (RL), we learn a control policy that fills in the information gap between the template model and full-body dynamics required to maintain stable and periodic locomotion. The proposed approach can be applied to robots of different sizes and morphologies and adapted to any RL technique and control architecture. We present experimental results showing that even in a model-free setup and with a simple reactive control architecture, the learned policies can generate realistic and energy-efficient locomotion gaits for a bipedal and a quadrupedal robot. And most importantly, this is achieved without using motion capture, strong constraints in the dynamics or kinematics of the robot, nor prescribing limb coordination. We provide supplemental videos for qualitative analysis of the naturality of the learned gaits4.