ICRA 2023 论文列表

Advances in robotic skill acquisition have made it possible to build general-purpose libraries of learned skills for downstream manipulation tasks. However, naively executing these skills one after the other is unlikely to succeed without accounting for dependencies between actions prevalent in longhorizon plans. We present Sequencing Task-Agnostic Policies (STAP), a scalable framework for training manipulation skills and coordinating their geometric dependencies at planning time to solve long-horizon tasks never seen by any skill during training. Given that Q-functions encode a measure of skill feasibility, we formulate an optimization problem to maximize the joint success of all skills sequenced in a plan, which we estimate by the product of their Q-values. Our experiments indicate that this objective function approximates ground truth plan feasibility and, when used as a planning objective, reduces myopic behavior and thereby promotes long-horizon task success. We further demonstrate how STAP can be used for task and motion planning by estimating the geometric feasibility of skill sequences provided by a task planner. We evaluate our approach in simulation and on a real robot. Qualitative results and code are made available at sites.google.com/stanford.edu/stap.

We propose a novel scene-robot interaction graph (SRI-Graph) that exploits the known position of a mobile manipulator for robust and accurate scene understanding. Compared to the state-of-the-art scene graph approaches, the proposed SRI-Graph captures not only the relationships between the objects, but also the relationships between the robot manipulator and objects with which it interacts. To improve the detection accuracy of spatial relationships, we leverage the 3D position of the mobile manipulator in addition to RGB images. The manipulator's ego information is crucial for a successful scene understanding when the relationships are visually uncertain. The proposed model is validated for a real-world 3D robot-assisted feeding task. We release a new dataset named 3DRF-Pos for training and validation. We also develop a tool, named LabelImg-Rel, as an extension of the open-sourced image annotation tool LabelImg for a convenient annotation in robot-environment interaction scenarios*. Our experimental results using the Movo platform show that SRI-Graph outperforms the state-of-the-art approach and improves detection accuracy by up to 9.83%.

This paper addresses the robustness problem of visual-inertial state estimation for underwater operations. Underwater robots operating in a challenging environment are required to know their pose at all times. All vision-based localization schemes are prone to failure due to poor visibility conditions, color loss, and lack of features. The proposed approach utilizes a model of the robot's kinematics together with proprioceptive sensors to maintain the pose estimate during visual-inertial odometry (VIO) failures. Furthermore, the trajectories from successful VIO and the ones from the model-driven odometry are integrated in a coherent set that maintains a consistent pose at all times. Health-monitoring tracks the VIO process ensuring timely switches between the two estimators. Finally, loop closure is implemented on the overall trajectory. The resulting framework is a robust estimator switching between model-based and visual-inertial odometry (SM/VIO). Experimental results from numerous deployments of the Aqua2 vehicle demonstrate the robustness of our approach over coral reefs and a shipwreck.

Liquids and granular media are pervasive throughout human environments. Their free-flowing nature causes people to constrain them into containers. We do so with thousands of different types of containers made out of different materials with varying sizes, shapes, and colors. In this work, we present a state-of-the-art sensing technique for robots to perceive what liquid is inside of an unknown container. We do so by integrating Visible to Near Infrared (VNIR) reflectance spectroscopy into a robot's end effector. We introduce a hierarchical model for inferring the material classes of both containers and internal contents given spectral measurements from two integrated spectrometers. To train these inference models, we capture and open source a dataset of spectral measurements from over 180 different combinations of containers and liquids. Our technique demonstrates over 85% accuracy in identifying 13 different liquids and granular media contained within 13 different containers. The sensitivity of our spectral readings allow our model to also identify the material composition of the containers themselves with 96% accuracy. Overall, VNIR spectroscopy presents a promising method to give household robots a general-purpose ability to infer the liquids inside of containers, without needing to open or manipulate the containers.

Most current LiDAR simultaneous localization and mapping (SLAM) systems build maps in point clouds, which are sparse when zoomed in, even though they seem dense to human eyes. Dense maps are essential for robotic applications, such as map-based navigation. Due to the low memory cost, mesh has become an attractive dense model for mapping in recent years. However, existing methods usually produce mesh maps by using an offline post-processing step to generate mesh maps. This two-step pipeline does not allow these methods to use the built mesh maps online and to enable localization and meshing to benefit each other. To solve this problem, we propose the first CPU-only real-time LiDAR SLAM system that can simultaneously build a mesh map and perform localization against the mesh map. A novel and direct meshing strategy with Gaussian process reconstruction realizes the fast building, registration, and updating of mesh maps. We perform experiments on several public datasets. The results show that our SLAM system can run at around 40Hz. The localization and meshing accuracy also outperforms the state-of-the-art methods, including the TSDF map and Poisson reconstruction. Our code and video demos are available at: https://github.com/lab-sun/SLAMesh.

This paper presents a simultaneous localization and map-assisted environment recognition (SLAMER) method. Mobile robots usually have an environment map and environment information can be assigned to the map. Important information such as no entry zone can be predicted from the map if localization has succeeded. However, this prediction is failed when localization does not work. Uncertainty of pose estimate must be considered for robust-map-based environ-mental object prediction. Robots also have external sensors and can recognize environmental object; however, sensor-based recognition of course contain uncertainty. SLAMER fuses map-based prediction and sensor-based recognition while coping with these uncertainties and achieves accurate localization and environment recognition. In this paper, we demonstrate LiDAR-based implementation of SLAMER in two cases. In the first case, we use the SemanticKITTI dataset and show that SLAMER achieves accurate estimate more than traditional methods. In the second case, we use an indoor mobile robot and show that unmeasurable environmental objects such as open doors and no entry lines can be recognized.

We present SIERRA, a novel framework for accelerating development and improving reproducibility of results in robotics research. SIERRA accelerates research by automating the process of generating experiments from queries over independent variables, executing experiments, and processing the results to generate deliverables such as graphs and videos. It shifts the paradigm for testing hypotheses from procedural (“Do these steps to answer the query”) to declarative (“Here is the query to test—GO!”), reducing the burden on researchers. It employs a modular architecture enabling easy customization and extension for the needs of individual researchers, thereby eliminating manual configuration and processing via throw-away scripts. SIERRA improves reproducibility of research by providing automation independent of the execution environment (HPC hardware, real robots, etc.) and targeted platform (simulator, real robots, etc.). This enables exact experiment replication, up to the limit of the execution environment and platform, as well as making it easy for researchers to test hypotheses in different computational environments. Though SIERRA is targeted at robotics research, its design makes it extendable to other fields.

Accurate mapping of large-scale environments is an essential building block of most outdoor autonomous systems. Challenges of traditional mapping methods include the balance between memory consumption and mapping accuracy. This paper addresses the problem of achieving large-scale 3D reconstruction using implicit representations built from 3D LiDAR measurements. We learn and store implicit features through an octree-based, hierarchical structure, which is sparse and extensible. The implicit features can be turned into signed distance values through a shallow neural network. We leverage binary cross entropy loss to optimize the local features with the 3D measurements as supervision. Based on our implicit representation, we design an incremental mapping system with regularization to tackle the issue of forgetting in continual learning. Our experiments show that our 3D reconstructions are more accurate, complete, and memory-efficient than current state-of-the-art 3D mapping methods.

Designing a safe and human-like decision-making system for an autonomous vehicle is a challenging task. Generative imitation learning is one possible approach for automating policy-building by leveraging both real-world and simulated decisions. Previous work that applies generative imitation learning to autonomous driving policies focuses on learning a low-level controller for simple settings. However, to scale to complex settings, many autonomous driving systems combine fixed, safe, optimization-based low-level controllers with high-level decision-making logic that selects the appropriate task and associated controller. In this paper, we attempt to bridge this gap in complexity by employing Safety-Aware Hierarchical Adversarial Imitation Learning (SHAIL), a method for learning a high-level policy that selects from a set of low-level controller instances in a way that imitates low-level driving data on-policy. We introduce an urban roundabout simulator that controls non-ego vehicles using real data from the Interaction dataset. We then demonstrate empirically that even with simple controller options, our approach can produce better behavior than previous approaches in driver imitation that have difficulty scaling to complex environments. Our implementation is available at https://github.com/sisl/InteractionImitation.

Cables are commonplace in homes, hospitals, and industrial warehouses and are prone to tangling. This paper extends prior work on autonomously untangling long cables by introducing novel uncertainty quantification metrics and actions that interact with the cable to reduce perception uncertainty. We present Sliding and Grasping for Tangle Manipulation 2.0 (SGTM 2.0), a system that autonomously untangles cables approximately 3 meters in length with a bilateral robot using estimates of uncertainty at each step to inform actions. By interactively reducing uncertainty, SGTM 2.0 significantly reduces run-time. Physical experiments with 84 trials suggest that SGTM 2.02.0 can achieve 83% untangling success on cables with 1 or 2 overhand and figure-8 knots, and 70% termination detection success across these configurations, outperforming SGTM 1.0 by 43% in untangling accuracy and 200% in completion time. Supplementary material, visualizations, and videos can be found at sites.google.com/view/sgtm2.

HOI detection is essential for human-computer interaction, especially in behavior detection and robot manipulation. Existing mainstream transformer methods of HOI detection are focused on single-stream detection only, e.g., image \rightarrow HOI(\mathcal{P}_{1})image \rightarrow HOI(\mathcal{P}_{1}), or image \rightarrow HO\rightarrow I(\mathcal{P}_{2})image \rightarrow HO\rightarrow I(\mathcal{P}_{2}). Both paths have their own characteristics of concern, so we propose a novel method, using the Secondary path (\mathcal{P}_{2})(\mathcal{P}_{2}) Guides the Primary path (\mathcal{P}_{1})(\mathcal{P}_{1}) in Transformers (SGPT). SGPT contains two core modules: the Dual-Path Consistency (DPC) module and the Instance Interaction Attention (IIA) module. DPC keeps human, object and interaction consistent on the dual-path and lets \mathcal{P}_{2}\mathcal{P}_{2} guide \mathcal{P}_{1}\mathcal{P}_{1} to learn more meaningful features. IIA fuses human and object to enhance interaction in \mathcal{P}_{2}\mathcal{P}_{2}, which allows instance to constrain interaction. Our proposed dual-path are employed during training, and only the \mathcal{P}_{1}\mathcal{P}_{1} path is used for inference. Hence, SGPT improves generalization without increasing model capacity in HICO-DET and V-COCO datasets compared to the state-of-the-arts. The code of this work is available at https://github.com/visualVk/sgpt.git.

Vision-based object tracking has boosted extensive autonomous applications for unmanned aerial vehicles (UAVs). However, the dynamic changes in flight maneuver and viewpoint encountered in UAV tracking pose significant difficulties, e.g., aspect ratio change, and scale variation. The conventional cross-correlation operation, while commonly used, has limitations in effectively capturing perceptual similarity and incorporates extraneous background information. To mitigate these limitations, this work presents a novel saliency-guided dynamic vision Transformer (SGDViT) for UAV tracking. The proposed method designs a new task-specific object saliency mining network to refine the cross-correlation operation and effectively discriminate foreground and background information. Additionally, a saliency adaptation embedding operation dynamically generates tokens based on initial saliency, thereby reducing the computational complexity of the Transformer architecture. Finally, a lightweight saliency filtering Transformer further refines saliency information and increases the focus on appearance information. The efficacy and robustness of the proposed approach have been thoroughly assessed through experiments on three widely-used UAV tracking benchmarks and real-world scenarios, with results demonstrating its superiority. The source code and demo videos are available at https://github.com/vision4robotics/SGDViT.

In robotic manipulation, acquiring samples is extremely expensive because it often requires interacting with the real world. Traditional image-level data augmentation has shown the potential to improve sample efficiency in various machine learning tasks. However, image-level data augmentation is insufficient for an imitation learning agent to learn good manipulation policies in a reasonable amount of demonstrations. We propose Simulation-augmented Equivariant Imitation Learning (SEIL), a method that combines a novel data augmentation strategy of supplementing expert trajectories with simulated transitions and an equivariant model that exploits the O(2) symmetry in robotic manipulation. Experimental evaluations demonstrate that our method can learn non-trivial manipulation tasks within ten demonstrations and outperform the baselines by a significant margin.

We address the problem of efficient 3-D exploration in indoor environments for micro aerial vehicles with limited sensing capabilities and payload/power constraints. We develop an indoor exploration framework that uses learning to predict the occupancy of unseen areas, extracts semantic features, samples viewpoints to predict information gains for different exploration goals, and plans informative trajectories to enable safe and smart exploration. Extensive experimentation in simulated and real-world environments shows the proposed approach outperforms the state-of-the-art exploration framework by 24% in terms of the total path length in a structured indoor environment and with a higher success rate during exploration.

Multi-objective optimization problems are ubiquitous in robotics, e.g., the optimization of a robot manipulation task requires a joint consideration of grasp pose configurations, collisions and joint limits. While some demands can be easily hand-designed, e.g., the smoothness of a trajectory, several task-specific objectives need to be learned from data. This work introduces a method for learning data-driven SE(3) cost functions as diffusion models. Diffusion models can represent highly-expressive multimodal distributions and exhibit proper gradients over the entire space due to their score-matching training objective. Learning costs as diffusion models allows their seamless integration with other costs into a single differentiable objective function, enabling joint gradient-based motion optimization. In this work, we focus on learning SE(3) diffusion models for 6DoF grasping, giving rise to a novel framework for joint grasp and motion optimization without needing to decouple grasp selection from trajectory generation. We evaluate the representation power of our SE(3) diffusion models w.r.t. classical generative models, and we showcase the superior performance of our proposed optimization framework in a series of simulated and real-world robotic manipulation tasks against representative baselines. Videos, code and additional details are available at: https://sites.google.com/view/se3dif

In this paper, we propose a signed distance field (SDF)-based deep Q-learning framework for multi-object re-arrangement. Our method learns to rearrange objects with non-prehensile manipulation, e.g., pushing, in unstructured environments. To reliably estimate Q-values in various scenes, we train the Q-network using an SDF-based scene graph as the state-goal representation. To this end, we introduce SDFGCN, a scalable Q-network structure which can estimate Q-values from a set of SDF images satisfying permutation invariance by using graph convolutional networks. In contrast to grasping-based rearrangement methods that rely on the performance of grasp predictive models for perception and movement, our approach enables rearrangements on unseen objects, including hard-to-grasp objects. Moreover, our method does not require any expert demonstrations. We observe that SDFGCN is capable of unseen objects in challenging configurations, both in the simulation and the real world.

We present a novel initialization technique for the range-aided simultaneous localization and mapping (RA-SLAM) problem. In RA-SLAM we consider measurements of point-to-point distances in addition to measurements of rigid transformations to landmark or pose variables. Standard formulations of RA-SLAM approach the problem as non-convex optimization, which requires a good initialization to obtain quality results. The initialization technique proposed here relaxes the RA-SLAM problem to a convex problem which is then solved to determine an initialization for the original, non-convex problem. The relaxation is a second-order cone program (SOCP), which is derived from a quadratically constrained quadratic program (QCQP) formulation of the RA-SLAM problem. As a SOCP, the method is highly scalable. We name this relaxation Second-order COnic RElaxation for RA-SLAM (SCORE). To our knowledge, this work represents the first convex relaxation for RA-SLAM. We present real-world and simulated experiments which show SCORE initialization permits the efficient recovery of quality solutions for a variety of challenging single- and multi-robot RA-SLAM problems with thousands of poses and range measurements.

Recovering full 3D shapes from partial observations is a challenging task that has been extensively addressed in the computer vision community. Many deep learning methods tackle this problem by training 3D shape generation networks to learn a prior over the full 3D shapes. In this training regime, the methods expect the inputs to be in a fixed canonical form, without which they fail to learn a valid prior over the 3D shapes. We propose SCARP, a model that performs Shape C ompletion in ARbitrary Poses. Given a partial pointcloud of an object, SCARP learns a disentangled feature representation of pose and shape by relying on rotationally equivariant pose features and geometric shape features trained using a multi-tasking objective. Unlike existing methods that depend on an external canonicalization method, SCARP performs canonicalization, pose estimation, and shape completion in a single network, improving the performance by 45% over the existing baselines. In this work, we use SCARP for improving grasp proposals on tabletop objects. By completing partial tabletop objects directly in their observed poses, SCARP enables a SOTA grasp proposal network improve their proposals by 71.2% on partial shapes. Project page: https://bipashasen.github.io/scarp

Designing a socially-aware navigation method for crowded environments has become a critical issue in robotics. In order to perform navigation in a crowded environment without causing discomfort to nearby pedestrians, it is necessary to design a global planner that is able to consider both human-robot interaction (HRI) and prediction of future states. In this paper, we propose a socially-aware global planner called SCAN, which is a global planner that generates appropriate local goals considering HRI and prediction of future states. Our method simulates future states considering the effects of the robot's actions on the future intentions of pedestrians using Monte Carlo tree search (MCTS), which estimates the quality of local goals. For fast simulation, we execute pedestrian motion prediction using Y-net and future state simulation using MCTS in parallel. Neural networks are only used in Y-net and not in MCTS, which enables fast simulation and prediction of a long horizon of future states. We evaluate the proposed method based on the proposed socially-aware navigation metric using realistic pedestrian simulation and real-world experiments. The results show that the proposed method outperforms existing methods significantly, indicating the importance of considering human-robot interaction for socially-aware navigation.