Deploying Unmanned Aerial Vehicles (UAVs) for traffic monitoring has been a hotspot given their flexibility and broader view. However, a UAV is usually constrained by battery capacity due to limited payload. On the ot...
详细信息
Deploying Unmanned Aerial Vehicles (UAVs) for traffic monitoring has been a hotspot given their flexibility and broader view. However, a UAV is usually constrained by battery capacity due to limited payload. On the other hand, the development of wireless charging technology has allowed UAVs to replenish energy from charging *** this paper, we study a UAV routing problem in the presence of multiple charging stations (URPMCS) with the objective of minimizing the total distance traveled by the UAV during traffic monitoring. We present a deep reinforcement learning based method, where a multi-head heterogeneous attention mechanism is designed to facilitate learning a policy that automatically and sequentially constructs the route, while taking the energy consumption into account. In our method, two types of attentions are leveraged to learn the relations between monitoring targets and charging station nodes, adopting an encoder-decoder-like policy network. Moreover, we also employ a curriculum learning strategy to enhance generalization to different numbers of charging stations. Computational results show that our method outperforms conventional algorithms with higher solution quality (except for exact methods such as Gurobi) and shorter runtime in general, and also exhibits strong generalized performance on problem instances with different distributions and sizes.
The unrelated parallel machine scheduling problem with sequence dependent setup times (UPMSP-SDST) addressed in this study refers to allocating jobs among a given number of machines and determining their processing se...
详细信息
The unrelated parallel machine scheduling problem with sequence dependent setup times (UPMSP-SDST) addressed in this study refers to allocating jobs among a given number of machines and determining their processing sequence on each machine, to minimize the makespan (i.e., the maximum completion time). To deal with large-scale UPMSP-SDST with higher efficiency, this study presents an enhanced adaptive large neighborhood search (EALNS) algorithm with various destroy and repair operators and an efficiency-enhancement mechanism. The efficiency-enhancement mechanism is mainly composed of a simplified calculation method and a hierarchical comparison mechanism, which are applied to improve the implementation process of the greedy-based operators. The simplified calculation method obtains a new makespan by an incremental or decremental transformation, to avoid reluctant calculations. The hierarchical mechanism refines the comparison of different removal or insertion strategies of the operators, thereby arresting high-quality solutions with more metrics. The proposed algorithm is tested on 1640 instances, and numerical results demonstrate the superior performance of the EALNS over the existing methods, especially for large-scale problems. Notably, 834 instances' best-known solutions are updated from this study. In addition, deep analysis of the impact of the distribution of setup time on the performance of the algorithm is provided, which further verifies the potential wide applicability of the proposed EALNS.
This paper proposed a collaborative neurodynamic optimization (CNO) method to solve traveling salesman problem (TSP). First, we construct a Hopfield neural network (HNN) with n x n neurons for the n cities. Second, to...
详细信息
This paper proposed a collaborative neurodynamic optimization (CNO) method to solve traveling salesman problem (TSP). First, we construct a Hopfield neural network (HNN) with n x n neurons for the n cities. Second, to ensure the convergence of continuous HNN (CHNN), we reformulate TSP to satisfy the convergence condition of CHNN and solve TSP by CHNN. Finally, a population of CHNNs is used to search for local optimal solutions of TSP and the globally optimal solution is obtained using particle swarm optimization. Experimental results show the effectiveness of the CNO approach for solving TSP.
Quantum annealers search for optimal solutions to a combinatorialoptimization problem by solving a quadratic unconstrained binary optimization (QUBO) model. Due to integrated control errors (ICEs), input QUBO coeffic...
详细信息
ISBN:
(纸本)9798331541378
Quantum annealers search for optimal solutions to a combinatorialoptimization problem by solving a quadratic unconstrained binary optimization (QUBO) model. Due to integrated control errors (ICEs), input QUBO coefficients are temporarily erroneous while a quantum annealer is running. Previous studies indicate that ICEs prevent quantum annealer from accurately running to obtain an optimal solution. This paper proposes a new method for reducing the ICE-induced noise effect. After a QUBO is input to a quantum annealer, the range of the QUBO coefficients is scaled within the hardware limitation. At that time, if the range is too large, the lowest absolute value (LAV) and near-LAV coefficients of the QUBO become too much small after scaling and these small coefficients can be more sensitive to the noise. To reduce the noise effect, our proposed method shrinks the range of QUBO coefficients by splitting it into multiple QUBOs. In splitting, LAV and near-LAV coefficients are left only in one QUBO, and the other coefficients are equally split so that each QUBO has a shrunk range. Finally, comparing the (quasi-)optimal solutions obtained from all split QUBOs, we obtain a solution closer to the original optimal solution. Experimental evaluation results show that the proposed method obtains more near-optimal solutions than the QUBO input as-is for all benchmarks.
Attention-based models (AMs) with deep reinforcement learning (DRL) training schemes have emerged as a prominent research direction in the field of machine learning for combinatorialoptimization (ML4CO). These models...
详细信息
ISBN:
(纸本)9798350359329;9798350359312
Attention-based models (AMs) with deep reinforcement learning (DRL) training schemes have emerged as a prominent research direction in the field of machine learning for combinatorialoptimization (ML4CO). These models have demonstrated great potential for quickly constructing solutions, without depending on expert knowledge or relying on costly supervised labels. However, existing AMs often neglect the use of positional embeddings and fail to exploit the relative distance information between nodes. To address this limitation, this paper proposes a novel adaptive distance-aware (ADA) mechanism for AMs. Drawing inspiration from the positional encoding technology used in Transformers, the ADA mechanism encodes the distance between pairs of nodes to re-scale the raw self-attention weights. Additionally, the proposed mechanism leverages multiple distance-aware structures to provide the decoder with diverse horizons, enabling it to effectively solve routing problems. The effectiveness of the proposed adaptive DA mechanism is validated through numerous experiments. The results demonstrate that integrating the ADA mechanism into existing attention models significantly improves the quality of the constructed solutions and enhances the generalization capability across different problem sizes.
The major problem facing users of Hopfield neural networks is the automatic choice of hyperparameters depending on the optimisation problem. This work introduces an automatic method to overcome this problem based on a...
详细信息
The major problem facing users of Hopfield neural networks is the automatic choice of hyperparameters depending on the optimisation problem. This work introduces an automatic method to overcome this problem based on an original mathematical model minimizing the energy function. This methods ensures the feasibility of optimal solution obtained by decomposing the set of the feasible solutions. We illustrate the proposed model in the context of six well-known NP-hard problems: meeting scheduling problem, Kohonen network problem, portfolio selection problem, traveling salesman problem, task assignment problem, and max-stable problem. To show the effectiveness of our model, we use particle swarm and genetic algorithms to solve several instances of the last three problems. Numerical results show the good performance of the proposed approach compared to random tuning hyperparameters methods. Indeed, our approach permits an improvement of 49.75% for traveling salesman problem, 5.92% for task assignment problem, and 29.41% for max-stable problem.
For combinatorialoptimization problem (CSP) solving of spiking neural networks (SNNs), both excitatory and inhibitory synaptic connections are necessary for mapping of constraints, along with adaptively-stochastic ne...
详细信息
ISBN:
(纸本)9798350332520
For combinatorialoptimization problem (CSP) solving of spiking neural networks (SNNs), both excitatory and inhibitory synaptic connections are necessary for mapping of constraints, along with adaptively-stochastic neuron. In this work, for the first time a novel ferroelectric FET (FeFET) based signed synapse with only two transistors is proposed and experimentally demonstrated to achieve excitatory and inhibitory connections, enabling cascade circuit with our previous proposed FeFETbased adaptively-stochastic neuron for all ferroelectric SNN optimizer. Based on the proposed design, a stochastic SNN is implemented for fast solving CSPs with accuracy improvement by 200%, providing a promising ultralow-hardware-cost and energy-efficient solution for optimization.
A real-life problem is the rostering of nurses at *** is a famous nondeterministic,polynomial time(NP)-hard combinatorialoptimization *** the real-world nurse rostering problem(NRP)constraints in distributing workloa...
详细信息
A real-life problem is the rostering of nurses at *** is a famous nondeterministic,polynomial time(NP)-hard combinatorialoptimization *** the real-world nurse rostering problem(NRP)constraints in distributing workload equally between available nurses is still a difficult task to *** international shortage of nurses,in addition to the spread of COVID-19,has made it more difficult to provide convenient rosters for *** on the literature,heuristic-based methods are the most commonly used methods to solve the NRP due to its computational complexity,especially for large ***-based algorithms in general have problems striking the balance between diversification and ***,this paper aims to introduce a novel metaheuristic hybridization that combines the enhanced harmony search algorithm(EHSA)with the simulated annealing(SA)algorithm called the annealing harmony search algorithm(AHSA).The AHSA is used to solve NRP from a Malaysian *** AHSA performance is compared to the EHSA,climbing harmony search algorithm(CHSA),deluge harmony search algorithm(DHSA),and harmony annealing search algorithm(HAS).The results show that the AHSA performs better than the other compared algorithms for all the tested instances where the best ever results reported for the UKMMC dataset.
Vehicle routing problems have attracted increasing attention because of the rapid development of transportation. Companies want to reduce the cost by lowering the number of vehicles and the total distances, which can ...
详细信息
ISBN:
(纸本)9798400701207
Vehicle routing problems have attracted increasing attention because of the rapid development of transportation. Companies want to reduce the cost by lowering the number of vehicles and the total distances, which can be considered as a combinatorialoptimization problem. The ant colony algorithm shows great potential in solving vehicle routing problems. However, it suffers from a low convergence speed due to the randomly initialized pheromone, which may cause a waste of computational resources in the early search process. To address this problem, a graph neural network is pre-trained to provide prior knowledge to initialize the pheromone in the ant colony algorithm, which can boost the convergence process. In addition, some classic local research methods are applied to balance the exploration and exploitation of the evolutionary process.
Transiently chaotic neural network (TCNN) and its improved versions have been proven to have search abilities for combinatorialoptimization problem (COP). However, the TCNN may be able to maintain its solving ability...
详细信息
ISBN:
(纸本)9798350321050
Transiently chaotic neural network (TCNN) and its improved versions have been proven to have search abilities for combinatorialoptimization problem (COP). However, the TCNN may be able to maintain its solving ability while the chaotic dynamics are cut. In this paper, the mechanism of the continuous-time Hopfield neural network (CHNN) and the TCNN for COP are analyzed qualitatively from the view of the energy function. It is believed that the "annealing" progress, i.e., the dynamic relaxation of the energy function, helps the improvement of the TCNN over the CHNN. Another Hopfield network with a linear activation function and annealing strategy for the COP is proposed in this paper. Simulations on the TSP show that the improved network performs as well as TCNN but is much more efficient. The performance of the TCNN is improved when linear activation functions are used. Compared with the traditional sigmoid activation function, the improved network is more suitable for hardware implementation.
暂无评论