Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysis

Building general-purpose robots that can operate seamlessly, in any environment, with any object, and utilizing various skills to complete diverse tasks has been a long-standing goal in Artificial Intelligence. Unfortunately, however, most existing robotic systems have been constrained - having been designed for specific tasks, trained on specific datasets, and deployed within specific environments. These systems usually require extensively-labeled data, rely on task-specific models, have numerous generalization issues when deployed in real-world scenarios, and struggle to remain robust to distribution shifts. Motivated by the impressive open-set performance and content generation capabilities of web-scale, large-capacity pre-trained models (i.e., foundation models) in research fields such as Natural Language Processing (NLP) and Computer Vision (CV), we devote this survey to exploring (i) how these existing foundation models from NLP and CV can be applied to the field of robotics, and also exploring (ii) what a robotics-specific foundation model would look like. We begin by providing an overview of what constitutes a conventional robotic system and the fundamental barriers to making it universally applicable. Next, we establish a taxonomy to discuss current work exploring ways to leverage existing foundation models for robotics and develop ones catered to robotics. Finally, we discuss key challenges and promising future directions in using foundation models for enabling general-purpose robotic systems. We encourage readers to view our living GitHub repository of resources, including papers reviewed in this survey as well as related projects and repositories for developing foundation models for robotics.

Multi-Robot Multi-Room Exploration with Geometric Cue Extraction and Spherical Decomposition

This work proposes an autonomous multi-robot exploration pipeline that coordinates the behaviors of robots in an indoor environment composed of multiple rooms. Contrary to simple frontier-based exploration approaches, we aim to enable robots to methodically explore and observe an unknown set of rooms in a structured building, keeping track of which rooms are already explored and sharing this information among robots to coordinate their behaviors in a distributed manner. To this end, we propose (1) a geometric cue extraction method that processes 3D map point cloud data and detects the locations of potential cues such as doors and rooms, (2) a spherical decomposition for open spaces used for target assignment. Using these two components, our pipeline effectively assigns tasks among robots, and enables a methodical exploration of rooms. We evaluate the performance of our pipeline using a team of up to 3 aerial robots, and show that our method outperforms the baseline by 36.6% in simulation and 26.4% in real-world experiments.

Robotic Interestingness via Human-Informed Few-Shot Object Detection

Interestingness recognition is crucial for decision making in autonomous exploration for mobile robots. Previous methods proposed an unsupervised online learning approach that can adapt to environments and detect interesting scenes quickly, but lack the ability to adapt to human-informed interesting objects. To solve this problem, we introduce a human-interactive framework, AirInteraction, that can detect human-informed objects via few-shot online learning. To reduce the communication bandwidth, we first apply an online unsupervised learning algorithm on the unmanned vehicle for interestingness recognition and then only send the potential interesting scenes to a base-station for human inspection. The human operator is able to draw and provide bounding box annotations for particular interesting objects, which are sent back to the robot to detect similar objects via few-shot learning. Only using few human-labeled examples, the robot can learn novel interesting object categories during the mission and detect interesting scenes that contain the objects. We evaluate our method on various interesting scene recognition datasets. To the best of our knowledge, it is the first human-informed few-shot object detection framework for autonomous exploration.

Calibrated Bagging Deep Learning for Image Semantic Segmentation: A Case Study on COVID-19 Chest X-ray Image

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causes coronavirus disease 2019 (COVID-19). Imaging tests such as chest X-ray (CXR) and computed tomography (CT) can provide useful information to clinical staff for facilitating a diagnosis of COVID-19 in a more efficient and comprehensive manner. As a breakthrough of artificial intelligence (AI), deep learning has been applied to perform COVID-19 infection region segmentation and disease classification by analyzing CXR and CT data. However, prediction uncertainty of deep learning models for these tasks, which is very important to safety-critical applications like medical image processing, has not been comprehensively investigated. In this work, we propose a novel ensemble deep learning model through integrating bagging deep learning and model calibration to not only enhance segmentation performance, but also reduce prediction uncertainty. The proposed method has been validated on a large dataset that is associated with CXR image segmentation. Experimental results demonstrate that the proposed method can improve the segmentation performance, as well as decrease prediction uncertainties.

AirDet: Few-Shot Detection without Fine-tuning for Autonomous Exploration

Few-shot object detection has rapidly progressed owing to the success of meta-learning strategies. However, the requirement of a fine-tuning stage in existing methods is timeconsuming and significantly hinders their usage in real-time applications such as autonomous exploration of low-power robots. To solve this problem, we present a brand new architecture, AirDet, which is free of fine-tuning by learning class agnostic relation with support images. Specifically, we propose a support-guided cross-scale (SCS) feature fusion network to generate object proposals, a global-local relation network (GLR) for shots aggregation, and a relation-based prototype embedding network (R-PEN) for precise localization. Exhaustive experiments are conducted on COCO and PASCAL VOC datasets, where surprisingly, AirDet achieves comparable or even better results than the exhaustively finetuned methods, reaching up to 40-60% improvements on the baseline. To our excitement, AirDet obtains favorable performance on multi-scale objects, especially the small ones. Furthermore, we present evaluation results on real-world exploration tests from the DARPA Subterranean Challenge, which strongly validate the feasibility of AirDet in robotics. The source code, pre-trained models, along with the real world data for exploration, will be made public.

Unsupervised Online Learning for Robotic Interestingness with Visual Memory

Autonomous robots frequently need to detect "interesting" scenes to decide on further exploration, or to decide which data to share for cooperation. These scenarios often require fast deployment with little or no training data. Prior work considers "interestingness" based on data from the same distribution. Instead, we propose to develop a method that automatically adapts online to the environment to report interesting scenes quickly. To address this problem, we develop a novel translation-invariant visual memory and design a three-stage architecture for long-term, short-term, and online learning, which enables the system to learn human-like experience, environmental knowledge, and online adaption, respectively. With this system, we achieve an average of 20% higher accuracy than the state-of-the-art unsupervised methods in a subterranean tunnel environment. We show comparable performance to supervised methods for robot exploration scenarios showing the efficacy of our approach. We expect that the presented method will play an important role in the robotic interestingness recognition exploration tasks.

Semi-supervised Learning for COVID-19 Image Classification via ResNet

Coronavirus disease 2019 (COVID-19) is an ongoing global pandemic in over 200 countries and territories, which has resulted in a great public health concern across the international community. Analysis of X-ray imaging data can play a critical role in timely and accurate screening and fighting against COVID-19. Supervised deep learning has been successfully applied to recognize COVID-19 pathology from X-ray imaging datasets. However, it requires a substantial amount of annotated X-ray images to train models, which is often not applicable to data analysis for emerging events such as COVID-19 outbreak, especially in the early stage of the outbreak. To address this challenge, this paper proposes a two-path semi-supervised deep learning model, ssResNet, based on Residual Neural Network (ResNet) for COVID-19 image classification, where two paths refer to a supervised path and an unsupervised path, respectively. Moreover, we design a weighted supervised loss that assigns higher weight for the minority classes in the training process to resolve the data imbalance. Experimental results on a large-scale of X-ray image dataset COVIDx demonstrate that the proposed model can achieve promising performance even when trained on very few labeled training images.

Combating the Compounding-Error Problem with a Multi-step Model

Model-based reinforcement learning is an appealing framework for creating agents that learn, plan, and act in sequential environments. Model-based algorithms typically involve learning a transition model that takes a state and an action and outputs the next state---a one-step model. This model can be composed with itself to enable predicting multiple steps into the future, but one-step prediction errors can get magnified, leading to unacceptable inaccuracy. This compounding-error problem plagues planning and undermines model-based reinforcement learning. In this paper, we address the compounding-error problem by introducing a multi-step model that directly outputs the outcome of executing a sequence of actions. Novel theoretical and empirical results indicate that the multi-step model is more conducive to efficient value-function estimation, and it yields better action selection compared to the one-step model. These results make a strong case for using multi-step models in the context of model-based reinforcement learning.