Robotic grasping is a crucial topic in robotics and computer vision, with broad applications in industrial production and intelligent manufacturing. Although some methods have begun addressing instance-level grasping, most remain limited to predefined instances and categories, lacking flexibility for open-vocabulary grasp prediction based on user-specified instructions. To address this, we propose RoG-SAM, a language-driven, instance-level grasp detection framework built on Segment Anything Model (SAM). RoG-SAM utilizes open-vocabulary prompts for object localization and grasp pose prediction, adapting SAM through transfer learning with encoder adapters and multi-head decoders to extend its segmentation capabilities to grasp pose estimation. Experimental results show that RoG-SAM achieves competitive performance on single-object datasets (Cornell and Jacquard) and cluttered datasets (GraspNet-1Billion and OCID), with instance-level accuracies of 91.2% and 90.1%, respectively, while using only 28.3% of SAM's trainable parameters. The effectiveness of RoG-SAM was also validated in real-world environments. A demonstration video is available athttps://www.youtube.com/playlist?list=PL7et4nGJAImLGytsJbglGbXl1hacA2dy_.

The H ∞ control method is an effective approach for attenuating the effect of disturbances on practical systems, but it is difficult to obtain the H ∞ controller due to the nonlinear Hamilton—Jacobi—Isaacs equation, even for linear systems. This study deals with the design of an H ∞ controller for linear discrete-time systems. To solve the related game algebraic Riccati equation (GARE), a novel model-free minimax Q -learning method is developed, on the basis of an offline policy iteration algorithm, which is shown to be Newton’s method for solving the GARE. The proposed minimax Q -learning method, which employs off-policy reinforcement learning, learns the optimal control policies for the controller and the disturbance online, using only the state samples generated by the implemented behavior policies. Different from existing Q -learning methods, a novel gradient-based policy improvement scheme is proposed. We prove that the minimax Q -learning method converges to the saddle solution under initially admissible control policies and an appropriate positive learning rate, provided that certain persistence of excitation (PE) conditions are satisfied. In addition, the PE conditions can be easily met by choosing appropriate behavior policies containing certain excitation noises, without causing any excitation noise bias. In the simulation study, we apply the proposed minimax Q -learning method to design an H ∞ load-frequency controller for an electrical power system generator that suffers from load disturbance, and the simulation results indicate that the obtained H ∞ load-frequency controller has good disturbance rejection performance.

In this paper, a mathematical model for the joint allocation of two heterogeneous stochastic resources (namely, sensors and actuators) is presented, addressing the interdependencies between sensors and actuators, the resource constraints, the capability constraints as well as the strategy constraints. A heuristic initialized memetic algorithm (MA) is proposed to solve the joint allocation problem about stochastic resources (JASR). The integer-based dual-permutation encoding method is adopted and several permutation-based operators are involved in the process of crossover, mutation and local search. Besides, a hybrid initialization method is employed to maintain a balance between exploration and exploitation. For the performance evaluation, we build a general Monte Carlo simulation based JASR framework. Furthermore, we employ an extension of the state-of-the-art algorithm Swt_opt, MRBCH and BMA as competitors. Computational results show that the proposed MA performs very well in solving JASR instances of different scales, and it can generate better assignment schemes in most cases than its competitors in limited time.

Multi-Point Dynamic Aggregation (MPDA) is a novel task model for describing the process of multiple robots performing time-variant tasks. In the MPDA problem, several task points are located in different places and their states change over time. Multiple robots aggregate to these task points and execute the tasks cooperatively to make the states of all the task points change to zero. The task planning of MPDA is a typical NP-hard combinatorial optimization problem. Estimation of Distribution Algorithms (EDA) are evolutionary techniques based on probabilistic models. In this paper, a permutation-based EDA is proposed to solve the task planning problems in MPDA. The algorithm uses K-means clustering to update its probabilistic model which follows the multi-modal Gaussian distribution. Experimental results show that the proposed algorithm outperforms other compared methods in solving the task planning problems of MPDA.

This paper addresses a particular pursuit-evasion game, called as "fishing game" where a faster evader attempts to pass the gap between two pursuers. We are concerned with the conditions under which the evader or pursuers can win the game. This is a game of kind in which an essential aspect, barrier, separates the state space into disjoint parts associated with each player's winning region. We present a method of explicit policy to construct the barrier. This method divides the fishing game into two subgames related to the included angle and the relative distances between the evader and the pursuers, respectively, and then analyzes the possibility of capture or escape for each subgame to ascertain the analytical forms of the barrier. Furthermore, we fuse the games of kind and degree by solving the optimal control strategies in the minimum time for each player when the initial state lies in their winning regions. Along with the optimal strategies, the trajectories of the players are delineated and the upper bounds of their winning times are also derived.

The weapon target assignment (WTA) problem is a fundamental problem arising in defense-related applications of operations research. And the multi-stage weapon target assignment (MWTA) problem is the basis of dynamic weapon target assignment (DWTA) problems which commonly exist in practice. The MWTA problem considered in this paper is with uncertainties, namely the uncertain MWTA (UMWTA) problem, and is formulated into a multi-objective constrained combinatorial optimization problem with two competing objectives. Apart from maximizing damage to hostile targets, this paper follows the principle of minimizing ammunition consumption under the assumption that each element of the kill probability matrix follows four different probability distributions. In order to tackle the two challenges, i.e., multi-objective and the uncertainty, the multi-objective evolutionary algorithm based on decomposition with adaptive weight adjustment (MOEA/D-AWA) and the Max-Min robust operator are adopted to solve the problem efficiently. Then comparison studies between the MOEA/D-AWA and a single objective solver used for a relaxed formulation on solving both certain and uncertain instances of two different scaled MWTA problems which include four uncertain scenarios are conducted. Numerical results show that MOEA/D-AWA outperforms the single objective solver on solving both certain and uncertain multi-objective MWTA problems discussed in this paper. Comparisons between the results of the certain and uncertain formulation also indicate the necessity of the robust formulation of practical problems.

Generally, the sample data of Multiple Attributes Decision Making (MADM) problems is incomplete because of variety of factors such as noise in data, compactness of representation, prediction capability and randomness of experiment. Rough set theory is a useful mathematical tool for this incomplete decision systems, while the fuzziness of relation-based classification and uncertainty of attribute reduction always exist in traditional extended rough sets model. In order to classify the incomplete decision systems effectively, a new refined classification method with tolerance relation-based rough sets was presented in this paper. Considering the randomness of missing value, this method used attribute importance to replace attribute reduction to establish refined classification rules directly. Not only it can reduce the computational complexity, but also can increase classification accuracy. From the analysis and comparison of examples about classification problems of air weapon targets, the effectiveness and stability of this method for incomplete decision systems were verified.

Differential evolution (DE) and particle swarm optimization (PSO) are two formidable population-based optimizers (POs) that follow different philosophies and paradigms, which are successfully and widely applied in scientific and engineering research. The hybridization between DE and PSO represents a promising way to create more powerful optimizers, especially for specific problem solving. In the past decade, numerous hybrids of DE and PSO have emerged with diverse design ideas from many researchers. This paper attempts to comprehensively review the existing hybrids based on DE and PSO with the goal of collection of different ideas to build a systematic taxonomy of hybridization strategies. Taking into account five hybridization factors, i.e., the relationship between parent optimizers, hybridization level, operating order (OO), type of information transfer (TIT), and type of transferred information (TTI), we propose several classification mechanisms and a versatile taxonomy to differentiate and analyze various hybridization strategies. A large number of hybrids, which include the hybrids of DE and PSO and several other representative hybrids, are categorized according to the taxonomy. The taxonomy can be utilized not only as a tool to identify different hybridization strategies, but also as a reference to design hybrid optimizers. The tradeoff between exploration and exploitation regarding hybridization design is discussed and highlighted. Based on the taxonomy proposed, this paper also indicates several promising lines of research that are worthy of devotion in future.

In this paper, we propose an efficient rule-based heuristic to solve asset-based dynamic weapon-target assignment (DWTA) problems. The main idea of the proposed heuristic is to utilize the domain knowledge of DWTA problems to directly achieve weapon assignment, without large number of function evaluations. We update the saturation states of constraints in the assignment process to guarantee the feasibility of generated solutions. For the purpose of testing the performance of the proposed heuristic, we build a general Monte Carlo simulation-based DWTA framework. For comparison, we also employ a Monte Carlo method (MCM) to make DWTA decisions in different defense scenarios. From simulations with DWTA instances under different scales, the heuristic has obvious advantages over the MCM with regard to solution quality and computation time. The proposed method can solve large-scale DWTA problems (e.g., those including 100 weapons, 100 targets, and four defense stages) within only a few seconds.

The dynamic weapon-target assignment (DWTA) problem is a typical constrained combinatorial optimization problem with the objective of maximizing the total value of surviving assets threatened by hostile targets through all defense stages. A generic asset-based DWTA model is established, especially for the warfare scenario of force coordination, to formulate this problem. Four categories of constraints, involving capability constraints, strategy constraints, resource constraints (i.e., ammunition constraints), and engagement feasibility constraints, are taken into account in the DWTA model. The concept of virtual permutation (VP) is proposed to facilitate the generation of feasible decisions. A construction procedure (CP) converts VPs into feasible DWTA decisions. With constraint satisfaction guaranteed by the synergy of VPs and the CP, an elaborate local search (LS) operator, namely move-to-head operator, is constructed to avoid repeatedly generating the same decisions. The operator is integrated into two tabu search (TS) algorithms to solve DWTA problems. Comparative experiments involving a random sampling method, an LS method, a hybrid genetic algorithm, a hybrid ant-colony optimization algorithm, and our TS algorithms show that the proposed TS heuristics for DWTA outperform their competitors in most test cases and they are competent for high-quality real-time DWTA decision makings.

Global optimization process can often be divided into two subprocesses: exploration and exploitation. The tradeoff between exploration and exploitation (T:Er&Ei) is crucial in search and optimization, having a great effect on global optimization performance, e.g., accuracy and convergence speed of optimization algorithms. In this paper, definitions of exploration and exploitation are first given based on information correlation among samplings. Then, some general indicators of optimization hardness are presented to characterize problem difficulties. By analyzing a typical contraction-based three-stage optimization process,Optimal Contraction Theoremis presented to show thatT:Er&Eidepends on the optimization hardness of problems to be optimized.T:Er&Eiwill gradually lean toward exploration as optimization hardness increases. In the case of great optimization hardness, exploration-dominated optimizers outperform exploitation-dominated optimizers. In particular, random sampling will become an outstanding optimizer when optimization hardness reaches a certain degree. Besides, the optimal number of contraction stages increases with optimization hardness. In an optimal contraction way, the whole sampling cost is evenly distributed in all contraction stages, and each contraction takes the same contracting ratio. Furthermore, the characterization of optimization hardness is discussed in detail. The experiments with several typical global optimization algorithms used to optimize three groups of test problems validate the correctness of the conclusions made byT:Er&Eianalysis.