IROS 2023 论文列表

Automatic Robotic Assembly Sequence Planning (RASP) can significantly improve productivity and resilience in modern manufacturing along with the growing need for greater product customization. One of the main challenges in realizing such automation resides in efficiently finding solutions from a growing number of potential sequences for increasingly complex assemblies. Besides, costly feasibility checks are always required for the robotic system. To address this, we propose a holistic graphical approach including a graph representation called Assembly Graph for product assemblies and a policy architecture, Graph Assembly Processing Network, dubbed GRACE for assembly sequence generation. With GRACE, we are able to extract meaningful information from the graph input and predict assembly sequences in a step-by-step manner. In experiments, we show that our approach can predict feasible assembly sequences across product variants of aluminum profiles based on data collected in simulation of a dual-armed robotic system. We further demonstrate that our method is capable of detecting infeasible assemblies, substantially alleviating the undesirable impacts from false predictions, and hence facilitating real-world deployment soon. Code and training data are available at https://github.com/DLR-RM/GRACE.

For robot manipulation, a complete and accurate object shape is desirable. Here, we present a method that combines visual and haptic reconstruction in a closed-loop pipeline. From an initial viewpoint, the object shape is reconstructed using an implicit surface deep neural network. The location with highest uncertainty is selected for haptic exploration, the object is touched, the new information from touch and a new point cloud from the camera are added, object position is re-estimated and the cycle is repeated. We extend Rustler et al. (2022) by using a new theoretically grounded method to determine the points with highest uncertainty, and we increase the yield of every haptic exploration by adding not only the contact points to the point cloud but also incorporating the empty space established through the robot movement to the object. Additionally, the solution is compact in that the jaws of a closed two-finger gripper are directly used for exploration. The object position is re-estimated after every robot action and multiple objects can be present simultaneously on the table. We achieve a steady improvement with every touch using three different metrics and demonstrate the utility of the better shape reconstruction in grasping experiments on the real robot. On average, grasp success rate increases from 63.3 % to 70.4 % after a single exploratory touch and to 82.7% after five touches. The collected data and code are publicly available (https://osf.io/j6rkd/, https://github.com/ctu-vras/vishac).

Convolutional neural networks trained using supervised learning can improve visual perception for human-robot walking. These advances have been possible due to largescale datasets like ExoNet and StairNet - the largest open-source image datasets of real-world walking environments. However, these datasets require vast amounts of manually annotated data, the development of which is time consuming and labor intensive. Here we present a novel semi-supervised learning system (ExoNet-SSL) that uses over 1.2 million unlabelled images from ExoNet to improve training efficiency. We developed a deep learning model based on mobile vision transformers and trained the model using semi-supervised learning for image classification. Compared to standard supervised learning (98.4%), our ExoNet-SSL system was able to maintain high prediction accuracy (98.8%) when tested on previously unseen environments, while requiring 35% fewer labelled images during training. These results show that semi-supervised learning can improve training efficiency by leveraging large amounts of unlabelled data and minimize the size requirements for manually annotated images. Future research \text{will}\text{will} focus on model deployment for onboard real-time inference and control of human-robot walking.

This work introduces efficient symbolic algorithms for quantitative reactive synthesis. We consider resource-constrained robotic manipulators that need to interact with a human to achieve a complex task expressed in linear temporal logic. Our framework generates reactive strategies that not only guarantee task completion but also seek cooperation with the human when possible. We model the interaction as a two-player game and consider regret-minimizing strategies to encourage cooperation. We use symbolic representation of the game to enable scalability. For synthesis, we first introduce value iteration algorithms for such games with min-max objectives. Then, we extend our method to the regret-minimizing objectives. Our benchmarks reveal that our the symbolic framework not only significantly improves computation time (up to an order of magnitude) but also can scale up to much larger instances of manipulation problems with up to 2× number of objects and locations than the state of the art.

The robot exploration task has been widely studied with applications spanning from novel environment mapping to item delivery. For some time-critical tasks, such as rescue catastrophes, the agent is required to explore as efficiently as possible. Recently, Visit Frequency-based map representation achieved great success in such scenarios by discouraging repetitive visits with a frequency-based penalty. However, its relatively large size and single-agent settings hinder its further development. In this context, we propose Integrated Visit Frequency Map, which encodes identical information as Visit Frequency Map with a more compact size, and a visit frequency-based multi-agent information exchange and control scheme that is able to accommodate both representations. Through tests in diverse settings, the results indicate our proposed methods can achieve a comparable level of performance of VFM with lower bandwidth requirements and generalize well to different multi-agent setups including real-world environments.

The path planning problems arising in manipulation planning and in task and motion planning settings are typically repetitive: the same manipulator moves in a space that only changes slightly. Despite this potential for reuse of information, few planners fully exploit the available information. To better enable this reuse, we decompose the collision checking into reusable, and non-reusable parts. We then treat the sequences of path planning problems in manipulation planning as a multiquery path planning problem. This allows the usage of planners that actively minimize planning effort over multiple queries, and by doing so, actively reuse previous knowledge. We implement this approach in EIRM* and effort ordered LazyPRM*, and benchmark it on multiple simulated robotic examples. Further, we show that the approach of decomposing collision checks additionally enables the reuse of the gained knowledge over multiple different instances of the same problem, i.e., in a multiquery manipulation planning scenario. The planners using the decomposed collision checking out-perform the other planners in initial solution time by up to a factor of two while providing a similar solution quality.

This paper presents an efficient approach to object manipulation planning using Monte Carlo Tree Search (MCTS) to find contact sequences and an efficient ADMM-based trajectory optimization algorithm to evaluate the dynamic feasibility of candidate contact sequences. To accelerate MCTS, we propose a methodology to learn a goal-conditioned policy-value network and a feasibility classifier to direct the search towards promising nodes. Further, manipulation-specific heuristics enable to drastically reduce the search space. Systematic object manipulation experiments in a physics simulator and on real hardware demonstrate the efficiency of our approach. In particular, our approach scales favorably for long manipulation sequences thanks to the learned policy-value network, significantly improving planning success rate. All source code including the baseline can be found at https://hzhu.io/contact-mcts.

Optimal Multi-Robot Path Planning (MRPP) has garnered significant attention due to its many applications in domains including warehouse automation, transportation, and swarm robotics. Current MRPP solvers can be divided into reduction-based, search-based, and rule-based categories, each with their strengths and limitations. Regardless of the methodology, however, the issue of handling dense MRPP instances remains a significant challenge, where existing approaches generally demonstrate a dichotomy regarding solution optimality and efficiency. This study seeks to bridge the gap in optimal MRPP resolution for dense, highly-entangled scenarios, with potential applications to high-density storage systems and traffic congestion control. Toward that goal, we analyze the behaviors of SOTA MRPP algorithms in dense settings and develop two hybrid algorithms leveraging the strengths of existing SOTA algorithms: DCBS (database-accelerated enhanced conflict-based search) and SCBS (sparsified enhanced conflict-based search). Experimental validations demonstrate that DCBS and SCBS deliver a significant reduction in computational time compared to existing bounded-suboptimal methods and improve solution quality compared to existing rule-based methods, achieving a desirable balance between computational efficiency and solution optimality. As a result, DCBS and SCBS are particularly suitable for quickly computing good-quality solutions for multi-robot routing in dense settings. Simulation video https://youtu.be/dZxMPUr7Bqg Upon the publication of the manuscript source code and data will be released at https://github.com/arc-l/dcbs

Reinforcement learning (RL) has achieved promising results on most robotic control tasks. Safety of learning-based controllers is an essential notion of ensuring the effectiveness of the controllers. Current methods adopt whole consistency constraints during the training, thus resulting in inefficient exploration in the early stage. In this paper, we propose an algorithm named Constrained Policy Optimization with Extra Safety Budget (ESB-CPO) to strike a balance between the exploration efficiency and the constraints satis-faction. In the early stage, our method loosens the practical constraints of unsafe transitions (adding extra safety bud-get) with the aid of a new metric we propose. With the training process, the constraints in our optimization problem become tighter. Meanwhile, theoretical analysis and practical experiments demonstrate that our method gradually meets the cost limit's demand in the final training stage. When evaluated on Safety-Gym and Bullet-Safety-Gym benchmarks, our method has shown its advantages over baseline algorithms in terms of safety and optimality. Remarkably, our method gains remarkable performance improvement under the same cost limit compared with baselines.

Collaborative autonomous multi-agent systems covering a specified area have many potential applications. Traditional approaches for such problems involve designing model-based control policies; however, state-of-the-art classical control policy still exhibits a large degree of sub-optimality. We present a combined reinforcement learning (RL) and control approach for the multi-agent coverage problem involving agents with second-order dynamics, with the RL component being based on the Multi-Agent Proximal Policy Optimization Algorithm (MAPPO). Our proposed network architecture includes the incorporation of LSTM and self-attention, which allows the trained policy to adapt to a variable number of agents. Our trained policy significantly outperforms the state-of-the-art classical control policy. We demonstrate our proposed method in a variety of simulated experiments.

The deployment of agile autonomous systems in challenging, unstructured environments requires adaptation capabilities and robustness to uncertainties. Existing robust and adaptive controllers, such as those based on model predictive control (MPC), can achieve impressive performance at the cost of heavy online onboard computations. Strategies that efficiently learn robust and onboard-deployable policies from MPC have emerged, but they still lack fundamental adaptation capabilities. In this work, we extend an existing efficient Imitation Learning (IL) algorithm for robust policy learning from MPC with the ability to learn policies that adapt to challenging model/environment uncertainties. The key idea of our approach consists in modifying the IL procedure by conditioning the policy on a learned lower-dimensional model/environment representation that can be efficiently estimated online. We tailor our approach to the task of learning an adaptive position and attitude control policy to track trajectories under challenging disturbances on a multirotor. Evaluations in simulation show that a high-quality adaptive policy can be obtained in about 1.3 hours. We additionally empirically demonstrate rapid adaptation to in- and out-of-training-distribution uncertainties, achieving a 6.1 cm average position error under wind disturbances that correspond to about 50% of the weight of the robot, and that are 36% larger than the maximum wind seen during training.

Although dynamic games provide a rich paradigm for modeling agents' interactions, solving these games for real-world applications is often challenging. Many real-world interactive settings involve general nonlinear state and input constraints that couple agents' decisions with one another. In this work, we develop an efficient and fast planner for interactive trajectory optimization in constrained setups using a constrained game-theoretical framework. Our key insight is to leverage the special structure of agents' objective and constraint functions that are common in multi-agent interactions for fast and reliable planning. More precisely, we identify the structure of agents' cost and constraint functions under which the resulting dynamic game is an instance of a constrained dynamic potential game. Constrained dynamic potential games are a class of games for which instead of solving a set of coupled constrained optimal control problems, a constrained Nash equilibrium, i.e. a Generalized Nash equilibrium, can be found by solving a single constrained optimal control problem. This simplifies constrained interactive trajectory optimization significantly. We compare the performance of our method in a navigation setup involving four planar agents and show that our method is on average 20 times faster than the state-of-the-art. We further provide experimental validation of our proposed method in a navigation setup involving two quadrotors carrying a rigid object while avoiding collisions with two humans.

This paper outlines a novel versatile geometric method for the estimation of the efficiency of multistage compound planetary gearboxes. The approach is based on a virtual pitch point modeling that allows for accurate representation of the gear interaction with forces applied at the pitch point. When this modeling is applied to all gears in a multistage compound planetary gearbox, the gearbox efficiency may be estimated directly from the free body diagram of the compound planets in a simple and easy to implement fashion. The gear design parameters that determine the gearing efficiency are identified and a consistent strategy for maximizing the efficiency of such gearboxes is outlined. Finally, the introduced methods are applied to the design of a three-stage compound planetary gearbox used in a novel snake-like robot.

Emerging partial-assistance exoskeletons can enhance able-bodied performance and aid people with patho-logical gait or age-related immobility. However, every person walks differently, which makes it difficult to directly compute assistance torques from joint kinematics. Gait-state estimation-based controllers use phase (normalized stride time) and task variables (e.g., stride length and ground inclination) to parameterize the joint torques. Using kinematic models that depend on the gait-state, prior work has used an Extended Kalman filter (EKF) to estimate the gait-state online. However, this EKF suffered from kinematic errors since it used a subject-independent measurement model, and it is still unknown how personalization of this measurement model would reduce gait-state tracking error. This paper quantifies how much gait-state tracking improvement a personalized measurement model can have over a subject-independent measurement model when using an EKF-based gait-state estimator. Since the EKF performance depends on the measurement model covariance matrix, we tested on multiple different tuning parameters. Across reasonable values of tuning parameters that resulted in good performance, personalization improved estimation error on average by 8.5 ± 13.8% for phase (mean ± standard deviation), 27.2 \pm \ 8.1{\%}\pm \ 8.1{\%} for stride length, and 10.5 \pm \ 13.5{\%}\pm \ 13.5{\%} for ground inclination. These findings support the hypothesis that personalization of the measurement model significantly improves gait-state estimation performance in EKF based gait-state tracking (P \ll 0.05)(P \ll 0.05), which could ultimately enable reliable responses to faster human gait changes.

Effectively rearranging heterogeneous objects constitutes a high-utility skill that an intelligent robot should master. Whereas significant work has been devoted to the grasp synthesis of heterogeneous objects, little attention has been given to the planning for sequentially manipulating such objects. In this work, we examine the long-horizon sequential rearrangement of heterogeneous objects in a tabletop setting, addressing not just generating feasible plans but near-optimal ones. Toward that end, and building on previous methods, including combinatorial algorithms and Monte Carlo tree search-based solutions, we develop state-of-the-art solvers for optimizing two practical objective functions considering key object properties such as size and weight. Thorough simulation studies show that our methods provide significant advantages in handling challenging heterogeneous object rearrangement problems, especially in cluttered settings. Real robot experiments further demonstrate and confirm these advantages. Source code and evaluation data associated with this research will be available at https//github.com/arc-l/TRLB upon the publication of this manuscript

Reinforcement learning (RL) has emerged as a promising approach for optimizing traffic signal control (TSC) to ensure the efficient operation of transportation networks. However, the traditional trial-and-error technique in RL is usually impractical in real-world applications. Offline RL, which trains models using pre-collected datasets, is a more practical approach. However, this presents challenges such as suboptimal datasets and limited generalization of pre-trained models. To address this, we propose an offline-to-online RL framework for TSC that pre-trains a generalized model and quickly adapts to new traffic scenarios through online refinement. In the offline stage, we augment the pre-collected datasets to cover a diverse set of possible scenarios and use an offline RL method to pretrain a control model. To ensure generalization, we use FRAP-like network as our base model, which is designed to learn the basic logic for signal control. In the online stage, we introduce a discrepancy measure to tackle inconsistencies between offline pre-trained models and online scenarios and prioritize samples based on it. In the experiments, the proposed approach achieves competitive performance and reduces the training time needed for learning in new scenarios, compared to several baselines.

Eye gaze can convey rich information of human intentions, which enables the social robots to comprehend the cognition and behavior of human targets. However, the existing 3D gaze estimation methods generally have high requirements either on the dedicated hardware or the quantity and quality of training databases, which largely limits their practical application values. This paper proposes EasyGaze3D, an effective 3D gaze estimation framework using a single RGB camera. First, the framework detects the 2D facial landmarks and recovers the 3D facial shape from the input image, and derives the required camera parameters with these features. Then, without loss of generality, the gaze direction can be regarded as the vector pointing from the eyeball center to the pupil center, which are derived respectively from the detected facial landmarks and the spherical fitting performed on the recovered 3D facial shape. Besides, we propose a flexible yet efficient calibration module, namely Easy-Cali, for deriving the subject-specific 3D facial shape and eyeball centers. The features calibrated by Easy-Cali can further boost the performance of EasyGaze3D. Experimental results show that our proposed method, being plug-and-play and without the need of training on large-scale dataset, can achieve superior performance against the existing methods based on deep models.

The Vision Transformer (ViT) architecture has established its place in computer vision literature, however, training ViTs for RGB-D object recognition remains an understudied topic, viewed in recent literature only through the lens of multi-task pretraining in multiple vision modalities. Such approaches are often computationally intensive, relying on the scale of multiple pretraining datasets to align RGB with 3D information. In this work, we propose a simple yet strong recipe for transferring pretrained ViTs in RGB-D domains for 3D object recognition, focusing on fusing RGB and depth representations encoded jointly by the ViT. Compared to previous works in multimodal Transformers, the key challenge here is to use the attested flexibility of ViTs to capture cross-modal interactions at the downstream and not the pretraining stage. We explore which depth representation is better in terms of resulting accuracy and compare early and late fusion techniques for aligning the RGB and depth modalities within the ViT architecture. Experimental results in the Washington RGB-D Objects dataset (ROD) demonstrate that in such RGB → RGB-D scenarios, late fusion techniques work better than most popularly employed early fusion. With our transfer baseline, fusion ViTs score up to 95.4% top-1 accuracy in ROD, achieving new state-of-the-art results in this benchmark. We further show the benefits of using our multimodal fusion baseline over unimodal feature extractors in a synthetic-to-real visual adaptation as well as in an open-ended lifelong learning scenario in the ROD benchmark, where our model outperforms previous works by a margin of >8%. Finally, we integrate our method with a robot framework and demonstrate how it can serve as a perception utility in an interactive robot learning scenario, both in simulation and with a real robot.

Lines are interesting geometrical features commonly seen in indoor and urban environments. There is missing a complete benchmark where one can evaluate lines from a sequential stream of images in all its stages: Line detection, Line Association and Pose error. To do so, we present a complete and exhaustive benchmark for visual lines in a SLAM front-end, both for RGB and RGBD, by providing a plethora of complementary metrics. We have also labeled data from well-known SLAM datasets in order to have all in one poses and accurately annotated lines. In particular, we have evaluated 17 line detection algorithms, 5 line associations methods and the resultant pose error for aligning a pair of frames with several combinations of detector-association. We have packaged all methods and evaluations metrics and made them publicly available on web-page33https://prime-slam.github.io/evolin/.