While the exploration for embodied AI has spanned multiple decades, it remains a persistent challenge to endow agents with human-level intelligence, including perception, learning, reasoning, decision-making, control, and generalization capabilities, so that they can perform general-purpose tasks in open, unstructured, and dynamic environments. Recent advances in computer vision, natural language processing, and multi-modality learning have shown that the foundation models have superhuman capabilities for specific tasks. They not only provide a solid cornerstone for integrating basic modules into embodied AI systems but also shed light on how to scale up robot learning from a methodological perspective. This survey aims to provide a comprehensive and up-to-date overview of foundation models in robotics, focusing on autonomous manipulation and encompassing high-level planning and low-level control. Moreover, we showcase their commonly used datasets, simulators, and benchmarks. Importantly, we emphasize the critical challenges intrinsic to this field and delineate potential avenues for future research, contributing to advancing the frontier of academic and industrial discourse.
The language-conditioned robotic manipulation aims to transfer natural language instructions into executable actions, from simple pick-and-place to tasks requiring intent recognition and visual reasoning. Inspired by the dual process theory in cognitive science, which suggests two parallel systems of fast and slow thinking in human decision-making, we introduce Robotics with Fast and Slow Thinking (RFST), a framework that mimics human cognitive architecture to classify tasks and makes decisions on two systems based on instruction types. Our RFST consists of two key components: 1) an instruction discriminator to determine which system should be activated based on the current user instruction, and 2) a slow-thinking system that is comprised of a fine-tuned vision language model aligned with the policy networks, which allows the robot to recognize user intention or perform reasoning tasks. To assess our methodology, we built a dataset featuring real-world trajectories, capturing actions ranging from spontaneous impulses to tasks requiring deliberate contemplation. Our results, both in simulation and real-world scenarios, confirm that our approach adeptly manages intricate tasks that demand intent recognition and reasoning. The project is available at https://jlm-z.github.io/RSFT/
Recent work on visual representation learning has shown to be efficient for robotic manipulation tasks. However, most existing works pretrained the visual backbone solely on 2D images or egocentric videos, ignoring the fact that robots learn to act in 3D space, which is hard to learn from 2D observation. In this paper, we examine the effectiveness of pretraining for vision backbone with public-available large-scale 3D data to improve manipulation policy learning. Our method, namely Depth-aware Pretraining for Robotics (DPR), enables an RGB-only backbone to learn 3D scene representations from self-supervised contrastive learning, where depth information serves as auxiliary knowledge. No 3D information is necessary during manipulation policy learning and inference, making our model enjoy both efficiency and effectiveness in 3D space manipulation. Furthermore, we introduce a new way to inject robots' proprioception into the policy networks that makes the manipulation model robust and generalizable. We demonstrate in experiments that our proposed framework improves performance on unseen objects and visual environments for various robotics tasks on both simulated and real robots.
Imitation learning (IL), aiming to learn optimal control policies from expert demonstrations, has been an effective method for robot manipulation tasks. However, previous IL methods either only use expensive expert demonstrations and omit imperfect demonstrations or rely on interacting with the environment and learning from online experiences. In the context of robotic manipulation, we aim to conquer the above two challenges and propose a novel framework named Similarity Weighted Behavior Transformer (SWBT). SWBT effectively learn from both expert and imperfect demonstrations without interaction with environments. We reveal that the easy-to-get imperfect demonstrations, such as forward and inverse dynamics, significantly enhance the network by learning fruitful information. To the best of our knowledge, we are the first to attempt to integrate imperfect demonstrations into the offline imitation learning setting for robot manipulation tasks. Extensive experiments on the ManiSkill2 benchmark built on the high-fidelity Sapien simulator and real-world robotic manipulation tasks demonstrated that the proposed method can extract better features and improve the success rates for all tasks. Our code will be released upon acceptance of the paper.
Humans interpret scenes by recognizing both the identities and positions of objects in their observations. For a robot to perform tasks such as \enquote{pick and place}, understanding both what the objects are and where they are located is crucial. While the former has been extensively discussed in the literature that uses the large language model to enrich the text descriptions, the latter remains underexplored. In this work, we introduce the \textit{Object-Centric Instruction Augmentation (OCI)} framework to augment highly semantic and information-dense language instruction with position cues. We utilize a Multi-modal Large Language Model (MLLM) to weave knowledge of object locations into natural language instruction, thus aiding the policy network in mastering actions for versatile manipulation. Additionally, we present a feature reuse mechanism to integrate the vision-language features from off-the-shelf pre-trained MLLM into policy networks. Through a series of simulated and real-world robotic tasks, we demonstrate that robotic manipulator imitation policies trained with our enhanced instructions outperform those relying solely on traditional language instructions.
Recent advancements in Natural Language Processing (NLP) have led to the development of NLP-based recommender systems that have shown superior performance. However, current models commonly treat items as mere IDs and adopt discriminative modeling, resulting in limitations of (1) fully leveraging the content information of items and the language modeling capabilities of NLP models; (2) interpreting user interests to improve relevance and diversity; and (3) adapting practical circumstances such as growing item inventories. To address these limitations, we present GPT4Rec, a novel and flexible generative framework inspired by search engines. It first generates hypothetical "search queries" given item titles in a user's history, and then retrieves items for recommendation by searching these queries. The framework overcomes previous limitations by learning both user and item embeddings in the language space. To well-capture user interests with different aspects and granularity for improving relevance and diversity, we propose a multi-query generation technique with beam search. The generated queries naturally serve as interpretable representations of user interests and can be searched to recommend cold-start items. With GPT-2 language model and BM25 search engine, our framework outperforms state-of-the-art methods by $75.7\%$ and $22.2\%$ in Recall@K on two public datasets. Experiments further revealed that multi-query generation with beam search improves both the diversity of retrieved items and the coverage of a user's multi-interests. The adaptiveness and interpretability of generated queries are discussed with qualitative case studies.
Acoustic signal acts as an essential input to many systems. However, the pure acoustic signal is very difficult to extract, especially in noisy environments. Existing beamforming systems are able to extract the signal transmitted from certain directions. However, since microphones are centrally deployed, these systems have limited coverage and low spatial resolution. We overcome the above limitations and present ChordMics, a distributed beamforming system. By leveraging the spatial diversity of the distributed microphones, ChordMics is able to extract the acoustic signal from arbitrary points. To realize such a system, we further address the fundamental challenge in distributed beamforming: aligning the signals captured by distributed and unsynchronized microphones. We implement ChordMics and evaluate its performance under both LOS and NLOS scenarios. The evaluation results tell that ChordMics can deliver higher SINR than the centralized microphone array. The average performance gain is up to 15dB.
Sound recognition is an important and popular function of smart devices. The location of sound is basic information associated with the acoustic source. Apart from sound recognition, whether the acoustic sources can be localized largely affects the capability and quality of the smart device's interactive functions. In this work, we study the problem of concurrently localizing multiple acoustic sources with a smart device (e.g., a smart speaker like Amazon Alexa). The existing approaches either can only localize a single source, or require deploying a distributed network of microphone arrays to function. Our proposal called Symphony is the first approach to tackle the above problem with a single microphone array. The insight behind Symphony is that the geometric layout of microphones on the array determines the unique relationship among signals from the same source along the same arriving path, while the source's location determines the DoAs (direction-of-arrival) of signals along different arriving paths. Symphony therefore includes a geometry-based filtering module to distinguish signals from different sources along different paths and a coherence-based module to identify signals from the same source. We implement Symphony with different types of commercial off-the-shelf microphone arrays and evaluate its performance under different settings. The results show that Symphony has a median localization error of 0.694m, which is 68% less than that of the state-of-the-art approach.