ICRA 2024 Papers

Searching Gas Source actively and efficiently in unknown hazard environments is an important but challenging issue. Using mobile robots to autonomously search and navigate to gas source location provides a promising way. Existing methods are mostly based on the modularization framework which investigates the gas-source search and robot navigation tasks independently, leading to a decoupled approach that results in higher collision risks and lower navigation efficiency. Moreover, existing robot navigation techniques grapple with the intricacies of navigating through unknown environments. To tackle these complexities, we introduce an integrated framework that merges gas source localization with robot navigation. This unified structure, underpinned by an end-to-end learning approach, resolves the inherent conflicts between gas exploration and collision avoidance. Our approach aggregates the local observations (raw 3D-LiDAR data) and the expert guidance information (gas distribution), and directly generates navigation actions by implementing the reinforcement learning with a novel reward function based on region dynamic guidances, thus effectively addressing the challenges of active gas source searching in unknown environments. Simulation results underscore the adaptability of our method to diverse unknown environments, along with its superior gas source searching capabilities compared to conventional approaches. Finally, we conduct real-world experiments to demonstrate our feasibility.

Gas Source Localization (GSL) is a challenging field of research within the robotics community, with high-stakes search-and-rescue applications. Existing methods vary widely and each has its strengths and weaknesses. Comparisons of different methods are limited due to the lack of a broadly adopted and standardized testing methodology. Existing GSL evaluations vary in environment size, wind conditions, and gas simulation fidelity. They also lack photo-realistic rendering for the integration of obstacle avoidance. In this paper, we propose GSL-Bench, a benchmarking tool that can evaluate the performance of existing GSL algorithms. GSL-Bench features high-fidelity graphics and gas simulation, featuring NVIDIA’s® Isaac Sim and OpenFOAM computational fluid dynamics software (CFD). Realism is further increased by simulating relevant gas and wind sensors. Scene generation is simplified with the introduction of AutoGDM+, capable of procedural environment generation, CFD and particle-based gas dispersion simulation. To illustrate GSL-Bench’s capabilities, three algorithms are compared in six warehouse settings of increasing complexity: E. Coli, dung beetle, and a random walker. Our results demonstrate GSL-Bench’s ability to provide valuable insights into algorithm performance.Site: https://sites.google.com/view/gslbench/

It is promising but challenging to design flocking control for a robot swarm to autonomously follow changing patterns or shapes in a optimal distributed manner. The optimal flocking control with dynamic pattern formation is, therefore, investigated in this paper. A predictive flocking control algorithm is proposed based on a Gibbs random field (GRF), where bio-inspired potential energies are used to charaterize "robot-robot" and "robot-environment" interactions. Specialized performance-related energies, e.g., motion smoothness, are introduced in the proposed design to improve the flocking behaviors. The optimal control is obtained by maximizing a posterior distribution of a GRF. A region-based shape control is accomplished for pattern formation in light of a mean shift technique. The proposed algorithm is evaluated via the comparison with two state-of-the-art flocking control methods in an environment with obstacles. Both numerical simulations and real-world experiments are conducted to demonstrate the efficiency of the proposed design.

This paper proposes a GPU-accelerated optimization framework for collision avoidance problems where the controlled objects and the obstacles can be modeled as the finite union of convex polyhedra. A novel collision avoidance constraint is proposed based on scale-based collision detection and the strong duality of convex optimization. Under this constraint, the high-dimensional non-convex optimization problems of collision avoidance can be decomposed into several low-dimensional quadratic programmings (QPs) following the paradigm of alternating direction method of multipliers (ADMM). Furthermore, these low-dimensional QPs can be solved parallel with GPUs, significantly reducing computational time. High-fidelity simulations are conducted to validate the proposed method’s effectiveness and practicality.

Accurate global localization is crucial for autonomous navigation and planning. To this end, various GPS-aided Visual-Inertial Odometry (GPS-VIO) fusion algorithms are proposed in the literature. This paper presents a novel GPS-VIO system that is able to significantly benefit from the online calibration of the rotational extrinsic parameter between the GPS reference frame and the VIO reference frame. The behind reason is this parameter is observable. This paper provides novel proof through nonlinear observability analysis. We also evaluate the proposed algorithm extensively on diverse platforms, including flying UAV and driving vehicle. The experimental results support the observability analysis and show increased localization accuracy in comparison to state-of-the-art (SOTA) tightly-coupled algorithms.

This paper introduces the open-source framework, GIRA, which implements fundamental robotics algorithms for reconstruction, pose estimation, and occupancy modeling using compact generative models. Compactness enables perception in the large by ensuring that the perceptual models can be communicated through low-bandwidth channels during large-scale mobile robot deployments. The generative property enables perception in the small by providing high-resolution reconstruction capability. These properties address perception needs for diverse robotic applications, including multi-robot exploration and dexterous manipulation. State-of-the-art perception systems construct perceptual models via multiple disparate pipelines that reuse the same underlying sensor data, which leads to increased computation, redundancy, and complexity. GIRA bridges this gap by providing a unified perceptual modeling framework using Gaussian mixture models (GMMs) as well as a novel systems contribution, which consists of GPUaccelerated functions to learn GMMs 10-100x faster compared to existing CPU implementations. Because few GMM-based frameworks are open-sourced, this work seeks to accelerate innovation and broaden adoption of these techniques.

A robot in a human-centric environment needs to account for the human’s intent and future motion in its task and motion planning to ensure safe and effective operation. This requires symbolic reasoning about probable future actions and the ability to tie these actions to specific locations in the physical environment. While one can train behavioral models capable of predicting human motion from past activities, this approach requires large amounts of data to achieve acceptable long-horizon predictions. More importantly, the resulting models are constrained to specific data formats and modalities. Moreover, connecting predictions from such models to the environment at hand to ensure the applicability of these predictions is an unsolved problem. We present a system that utilizes a Large Language Model (LLM) to infer a human’s next actions from a range of modalities without fine-tuning. A novel aspect of our system that is critical to robotics applications is that it links the predicted actions to specific locations in a semantic map of the environment. Our method leverages the fact that LLMs, trained on a vast corpus of text describing typical human behaviors, encode substantial world knowledge, including probable sequences of human actions and activities. We demonstrate how these localized activity predictions can be incorporated in a human-aware task planner for an assistive robot to reduce the occurrences of undesirable human-robot interactions by 29.2% on average.

Hand-eye calibration is the problem of solving the transformation from the end-effector of a robot to the sensor attached to it. Commonly employed techniques, such as AXXB or AXZB formulations, rely on regression methods that require collecting pose data from different robot configurations, which can produce low accuracy and repeatability. However, the derived transformation should solely depend on the geometry of the end-effector and the sensor attachment. We propose Geometry-Based End-Effector Calibration (GBEC) that enhances the repeatability and accuracy of the derived transformation compared to traditional hand-eye calibrations. To demonstrate improvements, we apply the approach to two different robot-assisted procedures: Transcranial Magnetic Stimulation (TMS) and femoroplasty. We also discuss the generalizability of GBEC for camera-in-hand and marker-in-hand sensor mounting methods. In the experiments, we perform GBEC between the robot end-effector and an optical tracker’s rigid body marker attached to the TMS coil or femoroplasty drill guide. Previous research documents low repeatability and accuracy of the conventional methods for robot-assisted TMS hand-eye calibration. Applying GBEC to repeated calibrations, we obtain transformations with standard deviations of 0.37mm, 0.65mm, and 0.40mm (translation) along x, y, and z axes of the end-effector, respectively. The tool alignment experiments after using GBEC achieve a mean accuracy around 0.2mm in Euclidean distance. When compared to some existing methods, the proposed method relies solely on the geometry of the flange and the pose of the rigid-body marker, making it independent of workspace constraints or robot accuracy, without sacrificing the orthogonality of the rotation matrix. Our results validate the accuracy and applicability of the approach, providing a new and generalizable methodology for obtaining the transformation from the end-effector to a sensor.

Mobile manipulation constitutes a fundamental task for robotic assistants and garners significant attention within the robotics community. A critical challenge inherent in mobile manipulation is the effective observation of the target while approaching it for grasping. In this work, we propose a graspability-aware mobile manipulation approach powered by an online grasping pose fusion framework that enables a temporally consistent grasping observation. Specifically, the predicted grasping poses are online organized to eliminate the redundant, outlier grasping poses, which can be encoded as a grasping pose observation state for reinforcement learning. Moreover, on-the-fly fusing the grasping poses enables a direct assessment of graspability, encompassing both the quantity and quality of grasping poses. This assessment can subsequently serve as an observe-to-grasp reward, motivating the agent to prioritize actions that yield detailed observations while approaching the target object for grasping. Through extensive experiments conducted on the Habitat and Isaac Gym simulators, we find that our method attains a good balance between observation and manipulation, yielding high performance under various grasping metrics. Furthermore, we discover that the incorporation of temporal information from grasping poses aids in mitigating the sim-to-real gap, leading to robust performance in challenging real-world experiments. Project page: https://pku-epic.github.io/GAMMA/

Articulated objects like cabinets and doors are widespread in daily life. However, directly manipulating 3D articulated objects is challenging because they have diverse geometrical shapes, semantic categories, and kinetic constraints. Prior works mostly focused on recognizing and manipulating articulated objects with specific joint types. They can either estimate the joint parameters or distinguish suitable grasp poses to facilitate trajectory planning. Although these approaches have succeeded in certain types of articulated objects, they lack generalizability to unseen objects, which significantly impedes their application in broader scenarios. In this paper, we propose a novel framework of Generalizable Articulation Modeling and Manipulating for Articulated Objects (GAMMA), which learns both articulation modeling and grasp pose affordance from diverse articulated objects with different categories. In addition, GAMMA adopts adaptive manipulation to iteratively reduce the modeling errors and enhance manipulation performance. We train GAMMA with the PartNet-Mobility dataset and evaluate with comprehensive experiments in SAPIEN simulation and real-world Franka robot. Results show that GAMMA significantly outperforms SOTA articulation modeling and manipulation algorithms in unseen and cross-category articulated objects. Images, videos and codes are published on the project website at: sites.google.com/view/gamma-articulation.

Self-supervised depth estimation has evolved into an image reconstruction task that minimizes a photometric loss. While recent methods have made strides in indoor depth estimation, they often produce inconsistent depth estimation in textureless areas and unsatisfactory depth discrepancies at object boundaries. To address these issues, in this work, we propose GAM-Depth, developed upon two novel components: gradient-aware mask and semantic constraints. The gradient-aware mask enables adaptive and robust supervision for both key areas and textureless regions by allocating weights based on gradient magnitudes. The incorporation of semantic constraints for indoor self-supervised depth estimation improves depth discrepancies at object boundaries, leveraging a co-optimization network and proxy semantic labels derived from a pretrained segmentation model. Experimental studies on three indoor datasets, including NYUv2, ScanNet, and InteriorNet, show that GAM-Depth outperforms existing methods and achieves state-of-the-art performance, signifying a meaningful step forward in indoor depth estimation. Our code will be available at https://github.com/AnqiCheng1234/GAM-Depth.

We present FuncGrasp, a framework that can infer dense yet reliable grasp configurations for unseen objects using one annotated object and single-view RGB-D observation via categorical priors. Unlike previous works that only transfer a set of grasp poses, FuncGrasp aims to transfer infinite configurations parameterized by an object-centric continuous grasp function across varying instances. To ease the transfer process, we propose Neural Surface Grasping Fields (NSGF), an effective neural representation defined on the surface to densely encode grasp configurations. Further, we exploit function-to-function transfer using sphere primitives to establish semantically meaningful categorical correspondences, which are learned in an unsupervised fashion without any expert knowledge. We showcase the effectiveness through extensive experiments in both simulators and the real world. Remarkably, our framework significantly outperforms several strong baseline methods in terms of density and reliability for generated grasps.

Nano-sized unmanned aerial vehicles (UAVs) are well-fit for indoor applications and for close proximity to humans. To enable autonomy, the nano-UAV must be able to self-localize in its operating environment. This is a particularly-challenging task due to the limited sensing and compute resources on board. This work presents an online and onboard approach for localization in floor plans annotated with semantic information. Unlike sensor-based maps, floor plans are readily-available, and do not increase the cost and time of deployment. To overcome the difficulty of localizing in sparse maps, the proposed approach fuses geometric information from miniaturized time-of-flight sensors and semantic cues. The semantic information is extracted from images by deploying a state-of-the-art object detection model on a high-performance multi-core microcontroller onboard the drone, consuming only 2.5mJ per frame and executing in 38ms. In our evaluation, we globally localize in a real-world office environment, achieving 90% success rate. We also release an open-source implementation of our work 1.

In minimally invasive procedures such as biopsies and prostate cancer brachytherapy, accurate needle placement remains challenging due to limitations in current tracking methods related to interference, reliability, resolution or image contrast. This often leads to frequent needle adjustments and reinsertions. To address these shortcomings, we introduce an optimized needle shape-sensing method using a fully distributed grating-based sensor. The proposed method uses simple trigonometric and geometric modeling of the fiber using optical frequency domain reflectometry (OFDR), without requiring prior knowledge of tissue properties or needle deflection shape and amplitude. Our optimization process includes a reproducible calibration process and a novel tip curvature compensation method. We validate our approach through experiments in artificial isotropic and inhomogeneous animal tissues, establishing ground truth using 3D stereo vision and cone beam computed tomography (CBCT) acquisitions, respectively. Our results yield an average RMSE ranging from 0.58 ± 0.21 mm to 0.66 ± 0.20 mm depending on the chosen spatial resolution, achieving the submillimeter accuracy required for interventional procedures.

Soft tactile sensors can enable robots to grasp objects easily and stably by simultaneously providing tactile data and mechanical compliance to robotic hands. If there are low-cost and easy-to-build robotic hands equipped with soft tactile sensors, they would be highly accessible and facilitate many robotics projects. To this end, we propose an accessible robot hand capable of tactile sensing, which can be produced through digital fabrication. We made the robot hand using commercial servo motors as well as components 3D printed from PETG, TPU, and conductive TPU. These materials allow the robot hand to have a soft, durable, and even functional structure. Specifically, the soft fingertip was crafted from TPU and conductive TPU, and their mechanical and electrical properties enable easy implementation of tactile sensing capabilities, such as force and capacitive touch, simply by adding off-the-shelf sensors (air-pressure and capacitance). The proposed robot hand could effectively sense interaction forces and proximity to conductive objects, and its utilization in various tasks was also demonstrated successfully.

In the past decade, there has been significant advancement in designing wearable neural interfaces for controlling neurorobotic systems, particularly bionic limbs. These interfaces function by decoding signals captured noninvasively from the skin’s surface. Portable high-density surface electromyography (HD-sEMG) modules combined with deep learning decoding have attracted interest by achieving excellent gesture prediction and myoelectric control of prosthetic systems and neurorobots. However, factors like small electrode size and unstable electrode-skin contacts make HD-sEMG susceptible to pixel electrode drops. The sparse electrode-skin disconnections rooted in issues such as low adhesion, sweating, hair blockage, and skin stretch challenge the reliability and scalability of these modules as the perception unit for neurorobotic systems. This paper proposes a novel deep-learning model providing resiliency for HD-sEMG modules, which can be used in the wearable interfaces of neurorobots. The proposed 3D Dilated Efficient CapsNet model trains on an augmented input space to computationally ‘force’ the network to learn channel dropout variations and thus learn robustness to channel dropout. The proposed framework maintained high performance under a sensor dropout reliability study conducted. Results show conventional models’ performance significantly degrades with dropout and is recovered using the proposed architecture and the training paradigm.

One of the key challenges of visual Simultaneous Localization and Mapping (SLAM) in large-scale environments is how to effectively use global localization to correct the cumulative errors from long-term tracking. This challenge presents itself in two main aspects: first, the difficulty for robots in revisiting previous locations to perform loop closure, and second, the considerable memory resources required to maintain point-cloud-based global maps. Recent solutions have resorted into neural networks, using satellite images as the references for ground-level localization. However, most of these methods merely provide cross-view patch-matching results, which leads to unfeasible in integration with the SLAM system. To address these issues, we present a semantic-based cross-view localization method. This approach combines semantic information with a reward and penalty mechanism, enabling us to obtain a global probability map and achieve precise 3-degree-of-freedom (3-DoF) localization. Based on that, we develop a SLAM system that capitalizes on satellite imagery for global localization. This strategy effectively bridges the gap between SLAM and real-world coordinates while also substantially reducing accumulated errors. Our experimental results demonstrate that our global localization method significantly outperforms existing satellite-based systems. Moreover, in scenarios where the robot struggles to find loop closures, employing our localization method improves the SLAM accuracy.

Task planning for robotic cooking involves generating a sequence of actions for a robot to prepare a meal successfully. This paper introduces a novel task tree generation pipeline producing correct planning and efficient execution for cooking tasks. Our method first uses a large language model (LLM) to retrieve recipe instructions and then utilizes a fine-tuned GPT-3 to convert them into a task tree, capturing sequential and parallel dependencies among subtasks. The pipeline then mitigates the uncertainty and unreliable features of LLM outputs using task tree retrieval. We combine multiple LLM task tree outputs into a graph and perform a task tree retrieval to avoid questionable nodes and high-cost nodes to improve planning correctness and execution efficiency. Our evaluation results show its superior performance in task planning accuracy and efficiency compared to previous works.

We explore Bird’s-Eye View (BEV) generation, converting a BEV map into its corresponding multi-view street images. Valued for its unified spatial representation aiding multi-sensor fusion, BEV is pivotal for various autonomous driving applications. Creating accurate street-view images from BEV maps is essential for portraying complex traffic scenarios and enhancing driving algorithms. Concurrently, diffusion-based conditional image generation models have demonstrated remarkable outcomes, adept at producing diverse, high-quality, and condition-aligned results. Nonetheless, the training of these models demands substantial data and computational resources. Hence, exploring methods to fine-tune these advanced models, like Stable Diffusion, for specific conditional generation tasks emerges as a promising avenue. In this paper, we introduce a practical framework for generating images from a BEV layout. Our approach comprises two main components: the Neural View Transformation and the Street Image Generation. The Neural View Transformation phase converts the BEV map into aligned multi-view semantic segmentation maps by learning the shape correspondence between the BEV and perspective views. Subsequently, the Street Image Generation phase utilizes these segmentations as a condition to guide a fine-tuned latent diffusion model. This finetuning process ensures both view and style consistency. Our model leverages the generative capacity of large pretrained diffusion models within traffic contexts, effectively yielding diverse and condition-coherent street view images.