A novel approach to solve optimal control problems dealing simultaneously with fractional differential equations and time delay is proposed in this work. More precisely, a set of global radial basis functions are firstly used to approximate the states and control variables in the problem. Then, a collocation method is applied to convert the time-delay fractional optimal control problem to a nonlinear programming one. By solving the resulting challenge, the unknown coefficients of the original one will be finally obtained. In this way, the proposed strategy introduces a very tunable framework for direct trajectory optimization, according to the discretization procedure and the range of arbitrary nodes. The algorithm’s performance has been analyzed for several non-trivial examples, and the obtained results have shown that this scheme is more accurate, robust, and efficient than most previous methods.
21 June 2022 in Mathematical Methods in the Applied Sciences
Guzheng tune progression involves intricately harmonizing melodic motif transitions. Effectively navigating this vast creative possibility space to expose musically consistent elaborations presents challenges. We develop a specialized large long short-term memory (LSTM) model for generating musically consistent Guzheng tune transitions. First, we propose novel firefly algorithm (FA) enhancements, e.g., adaptive diversity preservation and adaptive swim parameters, to boost exploration effectiveness for navigating the vast creative combinatorics when generating Guzheng tune transitions. Then, we develop a specialized stacked LSTM architecture incorporating residual connections and conditioned embedding vectors that can leverage long-range temporal dependencies in Guzheng music patterns, including unsupervised learning of concise Guzheng-specific melody embedding vectors via a variational autoencoder, encapsulating unique harmonic signatures from performance descriptors to provide style guidance. Finally, we use LSTM networks to develop adversarial generative large models that enable realistic synthesis and evaluation of Guzheng tunes switching. We gather an extensive 10+ hour corpus of solo Guzheng recordings spanning 230 musical pieces, 130 distinguished performing artists, and 600+ audio tracks. Simultaneously, we conduct thorough Guzheng data analysis. Comparative assessments against strong baselines over systematic musical metrics and professional listeners validate significant generation fidelity improvements. Our model achieves a 63 % reduction in reconstruction error compared to the standard FA optimization after 1000 iterations. It also outperforms baselines in capturing characteristic motifs, maintaining modality coherence with under 2 % dissonant pitch errors, and retaining desired rhythmic cadences. User studies further confirm the superior naturalness, novelty, and stylistic faithfulness of the generated tune transitions, with ratings close to real data.
Personalized path-finding allows a user to search for a travel path that can visit its several regions of interest (ROIs). Existing independent searching methods may bring many users to ROIs simultaneously, such as popular attractions in scenic, and thus induce traffic jams. Therefore, it is necessary to find a personalized travel path considering the load balance of traffic flow. However, it is challenging to ensure load balancing on the road and ROIs while allowing each user to visit their ROIs. In this paper, we propose a personalized path-finding method with a multi-agent path-finding (MAPF) framework. Based on the MAPF, it allows users to find the path independently, and we designed a reward to guide the agent simultaneously to navigate to the destination and ROIs for personalized travel. We also improve the architecture of MAPF to guide the agents in learning the load balance through a centralized value network. We evaluated the algorithm with up to 1024 agents on randomly generated road network graphs and compared it against state-of-the-art MAPF planners. We also trained the agent using imitation learning and validated our framework on real-world datasets. Experimental results show that our algorithm is efficient and better than the existing algorithms.
The development of 5G technology has enabled the cloud-internet of things (IoT) to impact all areas of our lives. Sensors in cloud-IoT generate large-scale data, and the demand for massive data processing is also increasing. The performance of a single machine can no longer meet the needs of existing users. In contrast, a data center (DC) integrates computing power and storage resources through a specific network topology and satisfies the need to process massive data. Regarding large-scale heterogeneous traffic in DCs, differentiated traffic scheduling on demand reduces transmission latency and improves throughput. Therefore, this paper presents a traffic scheduling method based on deep Q-networks (DQN). This method collects network parameters, delivers them to the environment module, and completes the environment construction of network information and reinforcement learning elements through the environment module. Thus, the final transmission path of the elephant flow is converted based on the action given by DQN. The experimental results show that the method proposed in this paper effectively reduces the transmission latency and improves the link utilization and throughput to a certain extent.