Visual reasoning between visual images and natural language is a long-standing challenge in computer vision. Most of the methods aim to look for answers to questions only on the basis of the analysis of the offered qu...
详细信息
Visual reasoning between visual images and natural language is a long-standing challenge in computer vision. Most of the methods aim to look for answers to questions only on the basis of the analysis of the offered questions and images. Other approaches treat knowledge graphs as flattened tables to search for the answer. However, there are two major problems with these works: 1) the model disregards the fact that the world we surrounding us interlinks our hearing and speaking of natural language and 2) the model largely ignores the structure of the KG. To overcome these challenging deficiencies, a model should jointly consider two modalities of vision and language, as well as the rich structural and logical information embedded in knowledge graphs. To this end, we propose a general joint representation learning framework for visual reasoning, namely, knowledge-embedded mutual guidance. It realizes mutual guidance not only between visual data and natural language descriptions but also between knowledge graphs and reasoning models. In addition, it exploits the knowledge derived from the reasoning model to boost knowledge graphs when applying the visual relation detection task. The experimental results demonstrate that the proposed approach performs dramatically better than state-of-the-art methods on two benchmarks for visual reasoning.
In this paper, we discuss a novel graph matching problem, namely the parameterized Koopmans- Beckmann's graph matching (KBGMw). KBGMw is defined by a weighted linear combination of a series of Koopmans-Beckmann...
详细信息
In this paper, we discuss a novel graph matching problem, namely the parameterized Koopmans- Beckmann's graph matching (KBGMw). KBGMw is defined by a weighted linear combination of a series of Koopmans-Beckmann's graph matching. First, we show that KBGMw can be taken as a special case of the parameterized Lawler's graph matching, subject to certain conditions. Second, based on structured SVM, we propose a supervised learning method for automatically estimating the parameters of KBGMw. Experimental results on both synthetic and real image matching data sets show that the proposed method achieves relatively better performances, even superior to some deep learning methods. (c) 2020 Elsevier B.V. All rights reserved.
Map inpainting is an important technology in the production of maps for autonomous driving vehicles. In recent years, scholars have often used methods such as point cloud inpainting, RGB image inpainting, and depth in...
详细信息
Map inpainting is an important technology in the production of maps for autonomous driving vehicles. In recent years, scholars have often used methods such as point cloud inpainting, RGB image inpainting, and depth inpainting to repair maps. However, these methods require high computational power and result in longer algorithmic processing times. To address this issue, we propose SynerFill, a synergistic RGB-D images inpainting method that can simultaneously inpaint RGB and depth images. We design its network architecture and loss functions, which include a generator, an RGB image discriminator, a depth image discriminator, and an edge image discriminator. Second, we collect real-world data and build a large-scale, multi-scene, multi-weather dataset called the Synthetic City RGB-D (SCRGB-D) Dataset based on 3ds Max, CARLA, and Unreal Engine 4. To verify SynerFill, we conduct experiments on the SCRGB-D dataset, DynaFill dataset, and SceneNet dataset. The experimental results show that SynerFill achieves state-of-the-art (SOTA) performance.
Our editorial focuses on intelligent vehicle carries-enabled future transportation and logistics, aiming at facilitating the realization of smart urban environments. Various promising applications can be expected rang...
Our editorial focuses on intelligent vehicle carries-enabled future transportation and logistics, aiming at facilitating the realization of smart urban environments. Various promising applications can be expected ranging from low-altitude air parcel delivery to offshore aquaculture. Our goal is to attract more good ideas in this direction.
The path tracking of the robotic fish is a hotspot with its high maneuverability and environmental friendliness. However, the periodic oscillation generated by bionic fish-like propulsion mode may lead to unstable con...
详细信息
The path tracking of the robotic fish is a hotspot with its high maneuverability and environmental friendliness. However, the periodic oscillation generated by bionic fish-like propulsion mode may lead to unstable control. To this end, this article proposes a novel framework involving a newly designed platform and multiagent reinforcement learning (MARL) method. First, a bionic robotic fish equipped with a reaction wheel is developed to enhance the stability. Second, an MARL-based control framework is proposed for the cooperative control of tail-beating and reaction wheel. Correspondingly, a hierarchical training method including initial training and iterative training is designed to deal with the control coupling and frequency difference between two agents. Finally, extensive simulations and experiments indicate that the developed robotic fish and the proposed MARL-based control framework can effectively improve the accuracy and stability of path tracking. Remarkably, headshaking is reduced about 40%. It provides a promising reference for the stability optimization and cooperative control of bionic swimming robots featuring oscillatory motions.
Pairwise similarity has been widely used for image classification by propagating the class information from labeled images to unlabeled images and predicting the classes of unlabeled images accordingly. Although widel...
详细信息
For bionic underwater robots, it is a great challenge for depth control due to model uncertainty and strong nonlinearity. To this end, we propose a data-driven model predictive control (MPC) approach using reinforceme...
详细信息
For bionic underwater robots, it is a great challenge for depth control due to model uncertainty and strong nonlinearity. To this end, we propose a data-driven model predictive control (MPC) approach using reinforcement learning (RL) for robotic penguin depth control. First, by imitating the underwater mode of the biological penguin, a robotic prototype with a tendon-driven head, two-degrees-of-freedom wings, and a tendon-driven tail was designed. Then, a data-driven MPC framework is proposed considering the structure and motion properties of the robotic penguin. Especially, a data-based learning environment is constructed using a motion capture system, computational fluid dynamics, and a backpropagation neural network. Meanwhile, to maximize the benefits of the controller while ensuring safety and stability, a data-driven MPC using the RL scheme is applied to approximate the optimal policy. Combined with an appropriate reward design and periodic training, the closed-loop controller performance is significantly improved, and the validity of the proposed framework is finally tested by extensive simulations and experiments. Notably, this work will provide valuable insights into the learning-based motion control of bionic underwater robots.
To tackle the complexity of human and social factors in manufacturing systems, parallel manufacturing for industrial metaverses is proposed as a new paradigm in smart manufacturing for effective and efficient operatio...
详细信息
To tackle the complexity of human and social factors in manufacturing systems, parallel manufacturing for industrial metaverses is proposed as a new paradigm in smart manufacturing for effective and efficient operations of those systems, where Cyber-Physical-Social systems(CPSSs) and the Internet of Minds(Io M) are regarded as its infrastructures and the "Artificial systems", "Computational experiments"and "Parallel execution"(ACP) method is its methodological foundation for parallel evolution, closed-loop feedback, and collaborative optimization. In parallel manufacturing, social demands are analyzed and extracted from social intelligence for product R&D and production planning, and digital workers and robotic workers perform the majority of the physical and mental work instead of human workers, contributing to the realization of low-cost, high-efficiency and zero-inventory manufacturing. A variety of advanced technologies such as Knowledge Automation(KA), blockchain, crowdsourcing and Decentralized Autonomous Organizations(DAOs) provide powerful support for the construction of parallel manufacturing, which holds the promise of breaking the constraints of resource and capacity, and the limitations of time and space. Finally, the effectiveness of parallel manufacturing is verified by taking the workflow of customized shoes as a case,especially the unmanned production line named Flex Vega.
This paper utilizes parallel control to investigate the problem of event-triggered deep reinforcement learning and develops an event-triggered deep Q-network (ETDQN) for decision-making of autonomous driving, without ...
详细信息
This paper utilizes parallel control to investigate the problem of event-triggered deep reinforcement learning and develops an event-triggered deep Q-network (ETDQN) for decision-making of autonomous driving, without training an explicit triggering condition. Based on the framework of parallel control, the developed ETDQN incorporates information of actions into the feedback and constructs a dynamic control policy. First, in the realization of the dynamic control policy, we integrate the current state and the previous action to construct the augmented state as well as the augmented Markov decision process. Meanwhile, it is shown theoretically that the goal of the developed dynamic control policy is to learn the variation rate of the action. The augmented state contains information of the current state and the previous action, which enables the developed ETDQN to directly design the immediate reward considering communication loss. Then, based on dueling double deep Q-network (dueling DDQN), we establish the augmented action-value, value, and advantage functions to directly learn the optimal event-triggered decision-making policy of autonomous driving without an explicit triggering condition. It is worth noticing that the developed ETDQN applies to various deep Q-networks (DQNs). Empirical results demonstrate that, in event-triggered control, the developed ETDQN outperforms dueling DDQN and reduces communication loss effectively.
暂无评论