Search

Deep reinforcement learning has achieved great success in laser-based collision avoidance works because the laser can sense accurate depth information without too much redundant data, which can maintain the robustness of the algorithm when it is migrated from the simulation environment to the real world. However, high-cost laser devices are not only difficult to deploy for a large scale of robots but also demonstrate unsatisfactory robustness towards the complex obstacles, including irregular obstacles, e.g., tables, chairs, and shelves, as well as complex ground and special materials. In this paper, we propose a novel monocular camera-based complex obstacle avoidance framework. Particularly, we innovatively transform the captured RGB images to pseudo-laser measurements for efficient deep reinforcement learning. Compared to the traditional laser measurement captured at a certain height that only contains one-dimensional distance information away from the neighboring obstacles, our proposed pseudo-laser measurement fuses the depth and semantic information of the captured RGB image, which makes our method effective for complex obstacles. We also design a feature extraction guidance module to weight the input pseudo-laser measurement, and the agent has more reasonable attention for the current state, which is conducive to improving the accuracy and efficiency of the obstacle avoidance policy., Comment: arXiv admin note: substantial text overlap with arXiv:2108.06887

Deep Neural Networks (DNNs) are widely used for computer vision tasks. However, it has been shown that deep models are vulnerable to adversarial attacks, i.e. , their performances drop when imperceptible perturbations are made to the original inputs, which may further degrade the following visual tasks or introduce new problems such as data and privacy security. Hence, metrics for evaluating the robustness of deep models against adversarial attacks are desired. However, previous metrics are mainly proposed for evaluating the adversarial robustness of shallow networks on the small-scale datasets. Although the Cross Lipschitz Extreme Value for nEtwork Robustness (CLEVER) metric has been proposed for large-scale datasets ( e.g. , the ImageNet dataset), it is computationally expensive and its performance relies on a tractable number of samples. In this paper, we propose the Adversarial Converging Time Score (ACTS), an attack-dependent metric that quantifies the adversarial robustness of a DNN on a specific input. Our key observation is that local neighborhoods on a DNN’s output surface would have different shapes given different inputs. Hence, given different inputs, it requires different time for converging to an adversarial sample. Based on this geometry meaning, ACTS measures the converging time as an adversarial robustness metric. We validate the effectiveness and generalization of the proposed ACTS metric against different adversarial attacks on the large-scale ImageNet dataset using state-of-the-art deep networks. Extensive experiments show that our ACTS metric is an efficient and effective adversarial metric over the previous CLEVER metric.

Recently, deep convolutional neural networks (CNNs) have been widely explored in single image super-resolution (SISR) and contribute remarkable progress. However, most of the existing CNNs-based SISR methods do not adequately explore contextual information in the feature extraction stage and pay little attention to the final high-resolution (HR) image reconstruction step, hence hindering the desired SR performance. To address the above two issues, in this paper, we propose a two-stage attentive network (TSAN) for accurate SISR in a coarse-to-fine manner. Specifically, we design a novel multi-context attentive block (MCAB) to make the network focus on more informative contextual features. Moreover, we present an essential refined attention block (RAB) which could explore useful cues in HR space for reconstructing fine-detailed HR image. Extensive evaluations on four benchmark datasets demonstrate the efficacy of our proposed TSAN in terms of quantitative metrics and visual effects. Code is available at https://github.com/Jee-King/TSAN.

Background: The open and accessible regions of the chromosome are more likely to be bound by transcription factors which are important for nuclear processes and biological functions. Studying the change of chromosome flexibility can help to discover and analyze disease markers and improve the efficiency of clinical diagnosis. Current methods for predicting chromosome flexibility based on Hi-C data include the Flexibility-Rigidity Index (FRI) and the Gaussian Network Model (GNM), which have been proposed to characterize chromosome flexibility. However, these methods require the chromosome structure data based on 3D biological experiments, which is time-consuming and expensive. Objective: Generally, the folding and curling of the double helix sequence of DNA have a great impact on chromosome flexibility and function. Motivated by the success of genomic sequence analysis in biomolecular function analysis, we hope to propose a method to predict chromosome flexibility only based on genomic sequence data. Methods: We propose a new method (named "DeepCFP") using deep learning models to predict chromosome flexibility based on only genomic sequence features. The model has been tested in the GM12878 cell line. Results: The maximum accuracy of our model has reached 91%. The performance of DeepCFP is close to FRI and GNM. Conclusion: The DeepCFP can achieve high performance only based on genomic sequence.

We present a k -d tree construction algorithm designed to accelerate rendering of scenes with motion blur, in application scenarios where a k -d tree is either required or desired. Our associated data structure focuses on capturing incoherent motion within the nodes of a k -d tree and improves both data structure quality and efficiency over previous methods. At building time, we track primitives with motion that is significantly distinct from other primitives within the node, guarantee valid node references and the correctness of the data structure via primitive duplication heuristic and propagation rules. Then, we incorporate filtering heuristics to limit the propagation of primitives, and eliminate some incorrect duplicated references that may invalidate node references. We also demonstrate a strategy for updating existing MBKD for animated scenes. Our experiments with this hierarchy show artifact-free motion-blur rendering using a k -d tree, and demonstrate improvements against a traditional BVH with interpolation and a MSBVH structure designed to handle moving primitives, particularly in render time.

Task assignment is a critical technology for heterogeneous unmanned aerial vehicle (UAV) applications. Target precedence has typically been ignored in previous studies, such that it is possible to obtain a task assignment solution with an unreasonable target execution order. For this reason, a cooperative multiple task assignment problem with target precedence constraints (CMTAPTPC) model is proposed in this paper, which considers not only kinematic, resource, and task precedence constraints of the UAV, but also target precedence to achieve more realistic scenarios. In addition, a graph method is improved to detect and eliminate deadlocks in solutions that include target precedence constraints. We also introduced a waitable path coordination (WPC) algorithm to generate conflict-free flight paths. Unlike the traditional path elongation method, this method can reduce the number of path elongation operations and save the energy of UAVs. Based on the characteristics of the CMTAPTPC model, this study proposes a modified genetic algorithm that integrates the graph-based method and WPC algorithm to solve the task assignment problem. In the simulation, three problem-scale scenarios were designed, and the superior performance of the modified genetic algorithm was demonstrated by comparing it with a traditional genetic algorithm. Finally, a time series diagram shows the task assignment solution that meets all time constraints and illustrates the rationality of the WPC algorithm.

Dogfight is often a continuous and multi-round process with missile attacks. If the fighter only considers the security when evading the incoming missile, it will easily lose the superiority in subsequent air combat. Therefore, it is necessary to maintain as much tactical superiority as possible while ensuring a successful evasion. The amalgamative tactical requirements of achieving multiple evasive objectives in a dogfight are taken into account in this paper. A method of generating a nondominated maneuver strategy set for evading missiles with tactical requirements is proposed. The tactical requirements include higher miss distance, less energy consumption, and higher terminal superiority. Then the evasion problem is defined and reformulated into a multi-objective optimization problem, which is solved by a redesigned multi-objective evolutionary algorithm based on decomposition (MOEA/D). Simulations are used to demonstrate the feasibility and effectiveness of the approach. A set of approximate Pareto-optimal solutions satisfying the tactical requirements are obtained. These solutions can not only guide the fighter to avoid being hit but also achieve the goal of relatively reducing energy consumption and improving terminal superiority.

This study deals with the autonomous evasive maneuver strategy of unmanned combat air vehicle (UCAV), which is threatened by a high-performance beyond-visual-range (BVR) air-to-air missile (AAM). Considering tactical demands of achieving self-conflicting evasive objectives in actual air combat, including higher miss distance, less energy consumption and longer guidance support time, the evasive maneuver problem in BVR air combat is defined and reformulated into a multi-objective optimization problem. Effective maneuvers of UCAV used in different evasion phases are modeled in three-dimensional space. Then the three-level decision space structure is established according to qualitative evasive tactical planning. A hierarchical multi-objective evolutionary algorithm (HMOEA) is designed to find the approximate Pareto-optimal solutions of the problem. The approach combines qualitative tactical experience and quantitative maneuver decision optimization method effectively. Simulations are used to demonstrate the feasibility and effectiveness of the approach. The results show that the obtained set of decision variables constitutes nondominated solutions, which can meet different evasive tactical requirements of UCAV while ensuring successful evasion.

Background Protein–protein interactions (PPIs) are of great importance in cellular systems of organisms, since they are the basis of cellular structure and function and many essential cellular processes are related to that. Most proteins perform their functions by interacting with other proteins, so predicting PPIs accurately is crucial for understanding cell physiology. Results Recently, graph convolutional networks (GCNs) have been proposed to capture the graph structure information and generate representations for nodes in the graph. In our paper, we use GCNs to learn the position information of proteins in the PPIs networks graph, which can reflect the properties of proteins to some extent. Combining amino acid sequence information and position information makes a stronger representation for protein, which improves the accuracy of PPIs prediction. Conclusion In previous research methods, most of them only used protein amino acid sequence as input information to make predictions, without considering the structural information of PPIs networks graph. We first time combine amino acid sequence information and position information to make representations for proteins. The experimental results indicate that our method has strong competitiveness compared with several sequence-based methods.

For quite a long time, effective Beyond-Visual-Range (BVR) air combat tactics can only be discovered by human pilots in the actual combat process. However, due to the lack of actual combat opportunities, making new air combat tactics innovation was generally considered quite difficult. To address this challenge, we first introduced a solely end-to-end Reinforcement Learning (RL) approach for training competitive air combat agents with adversarial self-play from scratch in a high fidelity air combat simulation environment during training. Furthermore, a Key Air Combat Event Reward Shaping (KAERS) mechanism was proposed to provide sparse but objective shaped rewards beyond episodic win/lose signal to accelerate the initial machine learning process. Experimental results showed that multiple valuable air combat tactical behaviors emerged progressively. We hope this study could be extended to the future of air combat machine intelligence research.

Trajectory-ranked reward extrapolation (T-REX) provides a general framework to infer users’ intentions from sub-optimal demonstrations. However, it becomes inflexible when encountering multi-agent scenarios, due to its high complexity caused by rational behaviors, e.g., cooperation and communication. In this paper, we propose a novel Multi-Agent Trajectory-ranked Reward EXtrapolation framework (MA-TREX), which adopts inverse reinforcement learning to infer demonstrators’ cooperative intention in the environment with high-dimensional state-action space. Specifically, to reduce the dependence on demonstrators, the MA-TREX uses self-generated demonstrations to iteratively extrapolate the reward function. Moreover, a knowledge transfer method is adopted in the iteration process, by which the self-generated data required subsequently is only one third of the initial demonstrations. Experimental results on several multi-agent collaborative tasks demonstrate that the MA-TREX can effectively surpass the demonstrators and obtain the same level reward as the ground truth quickly and stably.

Air-to-air confrontation has attracted wide attention from artificial intelligence scholars. However, in the complex air combat process, operational strategy selection depends heavily on aviation expert knowledge, which is usually expensive and difficult to obtain. Moreover, it is challenging to select optimal action sequences efficiently and accurately with existing methods, due to the high complexity of action selection when involving hybrid actions, e.g., discrete/continuous actions. In view of this, we propose a novel Multi-Agent Hierarchical Policy Gradient algorithm (MAHPG), which is capable of learning various strategies and transcending expert cognition by adversarial self-play learning. Besides, a hierarchical decision network is adopted to deal with the complicated and hybrid actions. It has a hierarchical decision-making ability similar to humankind, and thus, reduces the action ambiguity efficiently. Extensive experimental results demonstrate that the MAHPG outperforms the state-of-the-art air combat methods in terms of both defense and offense ability. Notably, it is discovered that the MAHPG has the ability of Air Combat Tactics Interplay Adaptation, and new operational strategies emerged that surpass the level of experts.