Abstract--Learning tool use from a single human demonstration video offers a highly intuitive and efficient approach to robot teaching. While humans can effortlessly generalize a demonstrated tool manipulation skill to diverse tools that support the same function (e.g., pouring with a mug versus a teapot), current one-shot imitation learning (OSIL) methods struggle to achieve this. A key challenge lies in establishing functional correspondences between demonstration and test tools, considering significant geometric variations among tools with the same function (i.e., intra-function variations). To address this challenge, we propose FUNCTO (Function-Centric OSIL for Tool Manipulation), an OSIL method that establishes function-centric correspondences with a 3D functional keypoint representation, enabling robots to generalize tool manipulation skills from a single human demonstration video to novel tools with the same function despite significant intra-function variations. We evaluate FUNCTO against exiting modular OSIL methods and end-to-end behavioral cloning methods through real-robot experiments on diverse tool manipulation tasks. The results demonstrate the superiority of FUNCTO when generalizing to novel tools with intra-function geometric variations. More details are available at https://sites.google.com/view/functo. The ability to use tools has long been recognized as a hallmark of human intelligence [1]. Endowing robots with the same capability holds the promise of unlocking a wide range of downstream tasks and applications [2, 3, 4]. As a step towards this goal, we tackle the problem of one-shot imitation learning (OSIL) for tool manipulation, which involves teaching robots a tool manipulation skill with a single human demonstration video. Previous OSIL methods [4, 5, 6, 7, 8, 9, 10] above, it remains a non-trivial challenge for robots due assume that tools supporting the same function share highly to significant geometric variations (e.g., shape, size, topology) similar shapes or appearances.

By injecting action components into the VLMs, Vision-Language-Action models (VLAs) can be naturally formed and also show promising performance. Existing work has demonstrated the effectiveness and generalization of VLAs in multiple scenarios and tasks. Nevertheless, the transfer from VLMs to VLAs is not trivial since existing VLAs differ in their backbones, action-prediction formulations, data distributions, and training recipes. This leads to a missing piece for a systematic understanding of the design choices of VLAs. In this work, we disclose the key factors that significantly influence the performance of VLA and focus on answering three essential design choices: which backbone to select, how to formulate the VLA architectures, and when to add cross-embodiment data. The obtained results convince us firmly to explain why we prefer VLA and develop a new family of VLAs, RoboVLMs, which require very few manual designs and achieve a new state-of-the-art performance in three simulation tasks and real-world experiments. Through our extensive experiments, which include over 8 VLM backbones, 4 policy architectures, and over 600 distinct designed experiments, we provide a detailed guidebook for the future design of VLAs. In addition to the study, the highly flexible RoboVLMs framework, which supports easy integrations of new VLMs and free combinations of various design choices, is made public to facilitate future research.

Grasping has been a crucial but challenging problem in robotics for many years. One of the most important challenges is how to make grasping generalizable and robust to novel objects as well as grippers in unstructured environments. We present \regnet, a robotic grasping system that can adapt to different parallel jaws to grasp diversified objects. To support different grippers, \regnet embeds the gripper parameters into point clouds, based on which it predicts suitable grasp configurations. It includes three components: Score Network (SN), Grasp Region Network (GRN), and Refine Network (RN). In the first stage, SN is used to filter suitable points for grasping by grasp confidence scores. In the second stage, based on the selected points, GRN generates a set of grasp proposals. Finally, RN refines the grasp proposals for more accurate and robust predictions. We devise an analytic policy to choose the optimal grasp to be executed from the predicted grasp set. To train \regnet, we construct a large-scale grasp dataset containing collision-free grasp configurations using different parallel-jaw grippers. The experimental results demonstrate that \regnet with the analytic policy achieves the highest success rate of $74.98\%$ in real-world clutter scenes with $20$ objects, significantly outperforming several state-of-the-art methods, including GPD, PointNetGPD, and S4G. The code and dataset are available at https://github.com/zhaobinglei/REGNet-V2.

GR-2 is first pre-trained on a vast number of Internet videos to capture the dynamics of the world. This large-scale pre-training, involving 38 million video clips and over 50 billion tokens, equips GR-2 with the ability to generalize across a wide range of robotic tasks and environments during subsequent policy learning. Following this, GR-2 is fine-tuned for both video generation and action prediction using robot trajectories. It exhibits impressive multi-task learning capabilities, achieving an average success rate of 97.7% across more than 100 tasks. Moreover, GR-2 demonstrates exceptional generalization to new, previously unseen scenarios, including novel backgrounds, environments, objects, and tasks.

Linguistic ambiguity is ubiquitous in our daily lives. Previous works adopted interaction between robots and humans for language disambiguation. Nevertheless, when interactive robots are deployed in daily environments, there are significant challenges for natural human-robot interaction, stemming from complex and unpredictable visual inputs, open-ended interaction, and diverse user demands. In this paper, we present SInViG, which is a self-evolving interactive visual agent for human-robot interaction based on natural languages, aiming to resolve language ambiguity, if any, through multi-turn visual-language dialogues. It continuously and automatically learns from unlabeled images and large language models, without human intervention, to be more robust against visual and linguistic complexity. Benefiting from self-evolving, it sets new state-of-the-art on several interactive visual grounding benchmarks. Moreover, our human-robot interaction experiments show that the evolved models consistently acquire more and more preferences from human users. Besides, we also deployed our model on a Franka robot for interactive manipulation tasks. Results demonstrate that our model can follow diverse user instructions and interact naturally with humans in natural language, despite the complexity and disturbance of the environment.

Recent progress in vision language foundation models has shown their ability to understand multimodal data and resolve complicated vision language tasks, including robotics manipulation. We seek a straightforward way of making use of existing vision-language models (VLMs) with simple fine-tuning on robotics data. To this end, we derive a simple and novel vision-language manipulation framework, dubbed RoboFlamingo, built upon the open-source VLMs, OpenFlamingo. Unlike prior works, RoboFlamingo utilizes pre-trained VLMs for single-step vision-language comprehension, models sequential history information with an explicit policy head, and is slightly fine-tuned by imitation learning only on language-conditioned manipulation datasets. Such a decomposition provides RoboFlamingo the flexibility for open-loop control and deployment on low-performance platforms. By exceeding the state-of-the-art performance with a large margin on the tested benchmark, we show RoboFlamingo can be an effective and competitive alternative to adapt VLMs to robot control. Our extensive experimental results also reveal several interesting conclusions regarding the behavior of different pre-trained VLMs on manipulation tasks. We believe RoboFlamingo has the potential to be a cost-effective and easy-to-use solution for robotics manipulation, empowering everyone with the ability to fine-tune their own robotics policy.

Interactive visual grounding in Human-Robot Interaction (HRI) is challenging yet practical due to the inevitable ambiguity in natural languages. It requires robots to disambiguate the user input by active information gathering. Previous approaches often rely on predefined templates to ask disambiguation questions, resulting in performance reduction in realistic interactive scenarios. In this paper, we propose TiO, an end-to-end system for interactive visual grounding in human-robot interaction. Benefiting from a unified formulation of visual dialogue and grounding, our method can be trained on a joint of extensive public data, and show superior generality to diversified and challenging open-world scenarios. In the experiments, we validate TiO on GuessWhat?! and InViG benchmarks, setting new state-of-the-art performance by a clear margin. Moreover, we conduct HRI experiments on the carefully selected 150 challenging scenes as well as real-robot platforms. Results show that our method demonstrates superior generality to diversified visual and language inputs with a high success rate. Codes and demos are available at https://github.com/jxu124/TiO.

This paper presents INVIGORATE, a robot system that interacts with human through natural language and grasps a specified object in clutter. The objects may occlude, obstruct, or even stack on top of one another. INVIGORATE embodies several challenges: (i) infer the target object among other occluding objects, from input language expressions and RGB images, (ii) infer object blocking relationships (OBRs) from the images, and (iii) synthesize a multi-step plan to ask questions that disambiguate the target object and to grasp it successfully. We train separate neural networks for object detection, for visual grounding, for question generation, and for OBR detection and grasping. They allow for unrestricted object categories and language expressions, subject to the training datasets. However, errors in visual perception and ambiguity in human languages are inevitable and negatively impact the robot's performance. To overcome these uncertainties, we build a partially observable Markov decision process (POMDP) that integrates the learned neural network modules. Through approximate POMDP planning, the robot tracks the history of observations and asks disambiguation questions in order to achieve a near-optimal sequence of actions that identify and grasp the target object. INVIGORATE combines the benefits of model-based POMDP planning and data-driven deep learning. Preliminary experiments with INVIGORATE on a Fetch robot show significant benefits of this integrated approach to object grasping in clutter with natural language interactions. A demonstration video is available at https://youtu.be/zYakh80SGcU.

Ambiguity is ubiquitous in human communication. Previous approaches in Human-Robot Interaction (HRI) have often relied on predefined interaction templates, leading to reduced performance in realistic and open-ended scenarios. To address these issues, we present a large-scale dataset, \invig, for interactive visual grounding under language ambiguity. Our dataset comprises over 520K images accompanied by open-ended goal-oriented disambiguation dialogues, encompassing millions of object instances and corresponding question-answer pairs. Leveraging the \invig dataset, we conduct extensive studies and propose a set of baseline solutions for end-to-end interactive visual disambiguation and grounding, achieving a 45.6\% success rate during validation. To the best of our knowledge, the \invig dataset is the first large-scale dataset for resolving open-ended interactive visual grounding, presenting a practical yet highly challenging benchmark for ambiguity-aware HRI. Codes and datasets are available at: \href{https://openivg.github.io}{https://openivg.github.io}.

Recent advancements in Large Language Models (LLMs) such as GPT4 have displayed exceptional multi-modal capabilities in following open-ended instructions given images. However, the performance of these models heavily relies on design choices such as network structures, training data, and training strategies, and these choices have not been extensively discussed in the literature, making it difficult to quantify progress in this field. To address this issue, this paper presents a systematic and comprehensive study, quantitatively and qualitatively, on training such models. We implement over 20 variants with controlled settings. Concretely, for network structures, we compare different LLM backbones and model designs. For training data, we investigate the impact of data and sampling strategies. For instructions, we explore the influence of diversified prompts on the instruction-following ability of the trained models. For benchmarks, we contribute the first, to our best knowledge, comprehensive evaluation set including both image and video tasks through crowd-sourcing. Based on our findings, we present Lynx, which performs the most accurate multi-modal understanding while keeping the best multi-modal generation ability compared to existing open-sourced GPT4-style models.