This paper tackles the critical challenge of object navigation in autonomous navigation systems, particularly focusing on the problem of target approach and episode termination in environments with long optimal episode length in Deep Reinforcement Learning (DRL) based methods. While effective in environment exploration and object localization, conventional DRL methods often struggle with optimal path planning and termination recognition due to a lack of depth information. To overcome these limitations, we propose a novel approach, namely the Depth-Inference Termination Agent (DITA), which incorporates a supervised model called the Judge Model to implicitly infer object-wise depth and decide termination jointly with reinforcement learning. We train our judge model along with reinforcement learning in parallel and supervise the former efficiently by reward signal. Our evaluation shows the method is demonstrating superior performance, we achieve a 9.3% gain on success rate than our baseline method across all room types and gain 51.2% improvements on long episodes environment while maintaining slightly better Success Weighted by Path Length (SPL). Code and resources, visualization are available at: //github.com/HuskyKingdom/DITA_acml2023
相關內容
This work describes a Bayesian framework for reconstructing the boundaries that represent targeted features in an image, as well as the regularity (i.e., roughness vs. smoothness) of these boundaries.This regularity often carries crucial information in many inverse problem applications, e.g., for identifying malignant tissues in medical imaging. We represent the boundary as a radial function and characterize the regularity of this function by means of its fractional differentiability. We propose a hierarchical Bayesian formulation which, simultaneously, estimates the function and its regularity, and in addition we quantify the uncertainties in the estimates. Numerical results suggest that the proposed method is a reliable approach for estimating and characterizing object boundaries in imaging applications, as illustrated with examples from X-ray CT and image inpainting. We also show that our method is robust under various noise types, noise levels, and incomplete data.
The human ability to learn, generalize, and control complex manipulation tasks through multi-modality feedback suggests a unique capability, which we refer to as dexterity intelligence. Understanding and assessing this intelligence is a complex task. Amidst the swift progress and extensive proliferation of large language models (LLMs), their applications in the field of robotics have garnered increasing attention. LLMs possess the ability to process and generate natural language, facilitating efficient interaction and collaboration with robots. Researchers and engineers in the field of robotics have recognized the immense potential of LLMs in enhancing robot intelligence, human-robot interaction, and autonomy. Therefore, this comprehensive review aims to summarize the applications of LLMs in robotics, delving into their impact and contributions to key areas such as robot control, perception, decision-making, and path planning. We first provide an overview of the background and development of LLMs for robotics, followed by a description of the benefits of LLMs for robotics and recent advancements in robotics models based on LLMs. We then delve into the various techniques used in the model, including those employed in perception, decision-making, control, and interaction. Finally, we explore the applications of LLMs in robotics and some potential challenges they may face in the near future. Embodied intelligence is the future of intelligent science, and LLMs-based robotics is one of the promising but challenging paths to achieve this.
Live migration of an application or VM is a well-known technique for load balancing, performance optimization, and resource management. To minimize the total downtime during migration, two popular methods -- pre-copy or post-copy -- are used in practice. These methods scale to large VMs and applications since the downtime is independent of the memory footprint of an application. However, in a secure, trusted execution environment (TEE) like Intel's scalable SGX, the state-of-the-art still uses the decade-old stop-and-copy method, where the total downtime is proportional to the application's memory footprint. This is primarily due to the fact that TEEs like Intel SGX do not expose memory and page table accesses to the OS, quite unlike unsecure applications. However, with modern TEE solutions that efficiently support large applications, such as Intel's Scalable SGX and AMD's Epyc, it is high time that TEE migration methods also evolve to enable live migration of large TEE applications with minimal downtime (stop-and-copy cannot be used any more). We present OptMig, an end-to-end solution for live migrating large memory footprints in TEE-enabled applications. Our approach does not require a developer to modify the application; however, we need a short, separate compilation pass and specialized software library support. Our optimizations reduce the total downtime by 98% for a representative microbenchmark that uses 20GB of secure memory and by 90 -- 96% for a suite of Intel SGX applications that have multi-GB memory footprints.
This paper presents a vision guidance and control method for autonomous robotic capture and stabilization of orbital objects in a time-critical manner. The method takes into account various operational and physical constraints, including ensuring a smooth capture, handling line-of-sight (LOS) obstructions of the target, and staying within the acceleration, force, and torque limits of the robot. Our approach involves the development of an optimal control framework for an eye-to-hand visual servoing method, which integrates two sequential sub-maneuvers: a pre-capturing maneuver and a post-capturing maneuver, aimed at achieving the shortest possible capture time. Integrating both control strategies enables a seamless transition between them, allowing for real-time switching to the appropriate control system. Moreover, both controllers are adaptively tuned through vision feedback to account for the unknown dynamics of the target. The integrated estimation and control architecture also facilitates fault detection and recovery of the visual feedback in situations where the feedback is temporarily obstructed. The experimental results demonstrate the successful execution of pre- and post-capturing operations on a tumbling and drifting target, despite multiple operational constraints.
This paper presents a Bayesian framework for inferring the posterior of the extended state of a target, incorporating its underlying goal or intent, such as any intermediate waypoints and/or final destination. The methodology is thus for joint tracking and intent recognition. Several novel latent intent models are proposed here within a virtual leader formulation. They capture the influence of the target's hidden goal on its instantaneous behaviour. In this context, various motion models, including for highly maneuvering objects, are also considered. The a priori unknown target intent (e.g. destination) can dynamically change over time and take any value within the state space (e.g. a location or spatial region). A sequential Monte Carlo (particle filtering) approach is introduced for the simultaneous estimation of the target's (kinematic) state and its intent. Rao-Blackwellisation is employed to enhance the statistical performance of the inference routine. Simulated data and real radar measurements are used to demonstrate the efficacy of the proposed techniques.
This paper presents a novel method for the reconstruction of high-resolution temporal images in dynamic tomographic imaging, particularly for discrete objects with smooth boundaries that vary over time. Addressing the challenge of limited measurements per time point, we propose a technique that synergistically incorporates spatial and temporal information of the dynamic objects. This is achieved through the application of the level-set method for image segmentation and the representation of motion via a sinusoidal basis. The result is a computationally efficient and easily optimizable variational framework that enables the reconstruction of high-quality 2D or 3D image sequences with a single projection per frame. Compared to current methods, our proposed approach demonstrates superior performance on both synthetic and pseudo-dynamic real X-ray tomography datasets. The implications of this research extend to improved visualization and analysis of dynamic processes in tomographic imaging, finding potential applications in diverse scientific and industrial domains.
This paper explores the integration of deep learning techniques for joint sensing and communications, with an extension to semantic communications. The integrated system comprises a transmitter and receiver operating over a wireless channel, subject to noise and fading effects. The transmitter employs a deep neural network, namely an encoder, for joint operations of source coding, channel coding, and modulation, while the receiver utilizes another deep neural network, namely a decoder, for joint operations of demodulation, channel decoding, and source decoding to reconstruct the data samples. The transmitted signal serves a dual purpose, supporting communication with the receiver and enabling sensing. When a target is present, the reflected signal is received, and another deep neural network decoder is utilized for sensing. This decoder is responsible for detecting the target's presence and determining its range. All these deep neural networks, including one encoder and two decoders, undergo joint training through multi-task learning, considering data and channel characteristics. This paper extends to incorporate semantic communications by introducing an additional deep neural network, another decoder at the receiver, operating as a task classifier. This decoder evaluates the fidelity of label classification for received signals, enhancing the integration of semantics within the communication process. The study presents results based on using the CIFAR-10 as the input data and accounting for channel effects like Additive White Gaussian Noise (AWGN) and Rayleigh fading. The results underscore the effectiveness of multi-task deep learning in achieving high-fidelity joint sensing and semantic communications.
This paper surveys research works in the quickly advancing field of instruction tuning (IT), a crucial technique to enhance the capabilities and controllability of large language models (LLMs). Instruction tuning refers to the process of further training LLMs on a dataset consisting of \textsc{(instruction, output)} pairs in a supervised fashion, which bridges the gap between the next-word prediction objective of LLMs and the users' objective of having LLMs adhere to human instructions. In this work, we make a systematic review of the literature, including the general methodology of IT, the construction of IT datasets, the training of IT models, and applications to different modalities, domains and applications, along with an analysis on aspects that influence the outcome of IT (e.g., generation of instruction outputs, size of the instruction dataset, etc). We also review the potential pitfalls of IT along with criticism against it, along with efforts pointing out current deficiencies of existing strategies and suggest some avenues for fruitful research.
Due to the significance and value in human-computer interaction and natural language processing, task-oriented dialog systems are attracting more and more attention in both academic and industrial communities. In this paper, we survey recent advances and challenges in an issue-specific manner. We discuss three critical topics for task-oriented dialog systems: (1) improving data efficiency to facilitate dialog system modeling in low-resource settings, (2) modeling multi-turn dynamics for dialog policy learning to achieve better task-completion performance, and (3) integrating domain ontology knowledge into the dialog model in both pipeline and end-to-end models. We also review the recent progresses in dialog evaluation and some widely-used corpora. We believe that this survey can shed a light on future research in task-oriented dialog systems.
We propose a novel method for automatic reasoning on knowledge graphs based on debate dynamics. The main idea is to frame the task of triple classification as a debate game between two reinforcement learning agents which extract arguments -- paths in the knowledge graph -- with the goal to promote the fact being true (thesis) or the fact being false (antithesis), respectively. Based on these arguments, a binary classifier, called the judge, decides whether the fact is true or false. The two agents can be considered as sparse, adversarial feature generators that present interpretable evidence for either the thesis or the antithesis. In contrast to other black-box methods, the arguments allow users to get an understanding of the decision of the judge. Since the focus of this work is to create an explainable method that maintains a competitive predictive accuracy, we benchmark our method on the triple classification and link prediction task. Thereby, we find that our method outperforms several baselines on the benchmark datasets FB15k-237, WN18RR, and Hetionet. We also conduct a survey and find that the extracted arguments are informative for users.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191