Distributed manufacturing is the mainstream model to accelerate production. However, the heterogeneous production environment makes engineer hard to find the optimal scheduling. This work investigates the energy-efficient distributed heterogeneous permutation flow scheduling problem with flexible machine speed (DHPFSP-FMS) with minimizing makespan and energy consumption simultaneously. In DHPFSP-FMS, the local search misleads the population falling into local optima which reduces the convergence and diversity. To solve this problem, a bi-roles co-evolutionary algorithm is proposed which contains the following improvements: First, the global search and local search is divided into two swarms producer and consumer to balance computation. Second, three heuristic rules are designed to get a high-quality initialization population. Next, five problem-based local search strategies are designed to accelerate converging. Then, an efficient energy-saving strategy is presented to save energy. Finally, to verify the performance of the proposed algorithm, 22 instances are generated based on the Taillard benchmark, and a number of numerical experiments are adopted. The experiment results state that our algorithm is superior to the state-of-arts and more efficient for DHPFSP-FMS.

In fleet air defense, the efficient coordination of multiple ships to complete weapon-target assignment has always been a critical challenge, primarily due to the varying combat capabilities and duties associated with each ship. Consequently, the traditional “weapon-target” assignment mode has turned into a “ship-weapon-target” assignment mode in the multi-ship dynamic weapon-target assignment (MS-DWTA) problem we proposed, with a larger solution space. In this problem, different ships possess distinct attributes, such as defense duties, weapon types, and loaded missile quantities. To solve this problem, we proposed an Attention enhanced multi-agent Distributional reinforcement learning method with Dynamic Reward (ADDR). Different from standard reinforcement learning method, ADDR learns to estimate the distribution, as opposed to only the expectation of future return, enabling better adaptation to air defense scenarios with significant randomness. The multi-head attention network integrates both the ship situation and the target situation to appropriately adjust the output of each agent, which explicitly considers the agent-level impact of ships to the whole fleet. Moreover, due to the missile fight time, ships may not immediately receive rewards after executing actions. To address this delayed phenomenon, we designed a dynamic reward mechanism to accurately adjust the delayed rewards. Through extensive simulation experiments, ADDR has demonstrated superior performance over multiple evaluation metrics.

We propose a novel online coregularization framework for multiview semisupervised learning based on the notion of duality in constrained optimization. Using the weak duality theorem, we reduce the online coregularization to the task of increasing the dual function. We demonstrate that the existing online coregularization algorithms in previous work can be viewed as an approximation of our dual ascending process using gradient ascent. New algorithms are derived based on the idea of ascending the dual function more aggressively. For practical purpose, we also propose two sparse approximation approaches for kernel representation to reduce the computational complexity. Experiments show that our derived online coregularization algorithms achieve risk and accuracy comparable to offline algorithms while consuming less time and memory. Specially, our online coregularization algorithms are able to deal with concept drift and maintain a much smaller error rate. This paper paves a way to the design and analysis of online coregularization algorithms.

Due to the advantages of low cost and high flexibility, drones have been applied to urban surveillance, vegetation monitoring, and other fields with the need for coverage of regions. To expand UAVs’ coverage, we designed the Capsule Airport (CA) to recharge and restore drones and provide take-off and landing services. Meanwhile, the combination of drones’ coverage path planning (CPP) and the deployment of CAs is a crucial problem with few relevant studies. We propose a solution approach to the CPP problem based on selecting scanning patterns and trapezoidal decomposition. In addition, we construct a 0–1 integer programming model to minimize the cost of the distance between CAs and the scanning missions. Specifically, a solution approach based on greedy and clustering heuristics is designed to solve this problem. Furthermore, we then develop a local-search-based algorithm with the operators of CA location exchange and drone scanning mission exchange to further optimize the solution. Random instances at different sizes are used to validate the performance of proposed algorithms, through which the sensitivity analysis is conducted with some factors. Finally, a case study based on the Maolichong forest park in Changsha, China, is presented to illustrate the application of the proposed method.