Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870369
Bo Sun, Wei Li, Y. Huang
Particle swarm optimization has been extensively noticed for its fast convergence speed with few parameters. However, it would be plagued by the premature convergence only affected by the global particles. In this study, Composite Proactive Particles in Swarm Optimization (Co-PPSO) is proposed. In Co-PPSO, the composite strategy framework is embedded into Proactive Particles in Swarm Optimization (PPSO), that is, three learning strategies are proposed to evaluate their differences and select the most suitable one for each particle. In addition, an elite group is constructed to make the particles jump out of the situation that they are only affected by the global best one in the particle swarm, and further improve the convergence accuracy. CEC2022 competition of single objective bound-constrained numerical optimization is employed to test the effect of 10-$D$ and 20-$D$ optimization, and four well-known PSO variants were used for comparison. The experimental results show that the Co-PPSO has certain competitiveness to improve premature convergence.
粒子群算法以其收敛速度快、参数少而受到广泛关注。但是，它会受到仅受全局粒子影响的过早收敛的困扰。本研究提出了一种复合主动粒子群优化算法(Co-PPSO)。在Co-PPSO中，将复合策略框架嵌入到主动粒子群优化(Proactive Particles In Swarm Optimization, PPSO)中，提出了三种学习策略来评估它们之间的差异，并为每个粒子选择最合适的学习策略。此外，构造了一个精英群，使粒子群跳出了粒子群中只受全局最优粒子影响的局面，进一步提高了收敛精度。采用CEC2022竞争单目标约束数值优化来测试10-$D$和20-$D$优化的效果，并使用四种知名的PSO变体进行比较。实验结果表明，Co-PPSO在改善过早收敛方面具有一定的竞争力。
{"title":"Performance of Composite PPSO on Single Objective Bound Constrained Numerical Optimization Problems of CEC 2022","authors":"Bo Sun, Wei Li, Y. Huang","doi":"10.1109/CEC55065.2022.9870369","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870369","url":null,"abstract":"Particle swarm optimization has been extensively noticed for its fast convergence speed with few parameters. However, it would be plagued by the premature convergence only affected by the global particles. In this study, Composite Proactive Particles in Swarm Optimization (Co-PPSO) is proposed. In Co-PPSO, the composite strategy framework is embedded into Proactive Particles in Swarm Optimization (PPSO), that is, three learning strategies are proposed to evaluate their differences and select the most suitable one for each particle. In addition, an elite group is constructed to make the particles jump out of the situation that they are only affected by the global best one in the particle swarm, and further improve the convergence accuracy. CEC2022 competition of single objective bound-constrained numerical optimization is employed to test the effect of 10-$D$ and 20-$D$ optimization, and four well-known PSO variants were used for comparison. The experimental results show that the Co-PPSO has certain competitiveness to improve premature convergence.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122642682","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870405
P. Krömer, Vojtěch Uher
The traveling salesman problem (TSP) is an iconic permutation problem with a number of applications in planning, scheduling, and logistics. It has also attracted much attention as a benchmarking problem frequently used to assess the properties of a variety of nature-inspired optimization methods. However, the standard libraries of TSP instances, such as the TSPLIB, are often decades old and might not reflect the requirements of modern real-world applications very well. In this work, we introduce several novel TSP instances representing real-world locations of pharmacies in several major cities of the Czech Republic. We look for the optimum routes between the pharmacies by selected nature-inspired algorithms and compare the results obtained on the real-world instances with their results on standard TSPLIB instances.
{"title":"Optimization of real-world supply routes by nature-inspired metaheuristics","authors":"P. Krömer, Vojtěch Uher","doi":"10.1109/CEC55065.2022.9870405","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870405","url":null,"abstract":"The traveling salesman problem (TSP) is an iconic permutation problem with a number of applications in planning, scheduling, and logistics. It has also attracted much attention as a benchmarking problem frequently used to assess the properties of a variety of nature-inspired optimization methods. However, the standard libraries of TSP instances, such as the TSPLIB, are often decades old and might not reflect the requirements of modern real-world applications very well. In this work, we introduce several novel TSP instances representing real-world locations of pharmacies in several major cities of the Czech Republic. We look for the optimum routes between the pharmacies by selected nature-inspired algorithms and compare the results obtained on the real-world instances with their results on standard TSPLIB instances.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129992182","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870436
Hao Hao, Shuai Wang, Bingdong Li, Aimin Zhou
The estimation of distribution algorithm (EDA) is an efficient heuristic method for handling black-box optimization problems since the ability for global population distribution modeling and gradient-free searching. However, the trial and error search mechanism relies on a large number of function evaluations, which is a considerable challenge under expensive black-box problems. Therefore, this article presents a surrogate assisted EDA with multi-acquisition functions. Firstly, a variable-width histogram is used as the global distribution model that focuses on promising areas. Next, the evaluated-free local search method improves the quality of new generation solutions. Fi-nally, model management with multiple acquisitions maintains global and local exploration preferences. Several commonly used benchmark functions with 20 and 50 dimensions are adopted to evaluate the proposed algorithm compared with several state-of-the-art surrogate assisted evaluation algorithms (SAEAs) and Bayesian optimization method. In addition, a rover trajectories optimizing problem is used to verify the ability to solve complex problems. The experimental results demonstrate the superiority of the proposed algorithm over these comparison algorithms.
{"title":"A Surrogate Model Assisted Estimation of Distribution Algorithm with Mutil-acquisition Functions for Expensive Optimization","authors":"Hao Hao, Shuai Wang, Bingdong Li, Aimin Zhou","doi":"10.1109/CEC55065.2022.9870436","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870436","url":null,"abstract":"The estimation of distribution algorithm (EDA) is an efficient heuristic method for handling black-box optimization problems since the ability for global population distribution modeling and gradient-free searching. However, the trial and error search mechanism relies on a large number of function evaluations, which is a considerable challenge under expensive black-box problems. Therefore, this article presents a surrogate assisted EDA with multi-acquisition functions. Firstly, a variable-width histogram is used as the global distribution model that focuses on promising areas. Next, the evaluated-free local search method improves the quality of new generation solutions. Fi-nally, model management with multiple acquisitions maintains global and local exploration preferences. Several commonly used benchmark functions with 20 and 50 dimensions are adopted to evaluate the proposed algorithm compared with several state-of-the-art surrogate assisted evaluation algorithms (SAEAs) and Bayesian optimization method. In addition, a rover trajectories optimizing problem is used to verify the ability to solve complex problems. The experimental results demonstrate the superiority of the proposed algorithm over these comparison algorithms.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128707315","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870308
S. Samarakoon, M. Muthugala, M. R. Elara
Area coverage is essential for robots used in cleaning, painting, and exploration applications. Reconfigurable robots have been introduced to solve the area coverage limitation of fixed-shape robots. The existing global coverage algorithms of reconfigurable robots are limited to consideration of a limited set of predefined shapes for the reconfiguration and do not consider the exact geometrical shape of obstacles. Therefore, degraded coverage performance could be observed from the existing methods. On the other hand, the coverage methods that consider reconfiguring beyond a limited set of predefined shapes are limited to local coverage. Furthermore, these methods only consider a single reconfiguration for the coverage. Therefore, this paper proposes a novel coverage method for a reconfigurable robot consisting of both global and local path planners. The global path planner uses boustrophedon motion combined with the A * algorithm. The optimum grid positioning that maximizes the global coverage is determined through a Genetic Algorithm (GA). The local coverage planner performs continuous reconfig-uration of the robot to adequately cover obstacle zones while navigating through narrow spaces without collisions. A GA is used to determine the reconfiguration parameters of the robot at each instance of the local coverage. Simulation results confirm that the proposed method is effective in performing both global and local coverage path planning for improving the area coverage performance.
{"title":"Global and Local Area Coverage Path Planner for a Reconfigurable Robot","authors":"S. Samarakoon, M. Muthugala, M. R. Elara","doi":"10.1109/CEC55065.2022.9870308","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870308","url":null,"abstract":"Area coverage is essential for robots used in cleaning, painting, and exploration applications. Reconfigurable robots have been introduced to solve the area coverage limitation of fixed-shape robots. The existing global coverage algorithms of reconfigurable robots are limited to consideration of a limited set of predefined shapes for the reconfiguration and do not consider the exact geometrical shape of obstacles. Therefore, degraded coverage performance could be observed from the existing methods. On the other hand, the coverage methods that consider reconfiguring beyond a limited set of predefined shapes are limited to local coverage. Furthermore, these methods only consider a single reconfiguration for the coverage. Therefore, this paper proposes a novel coverage method for a reconfigurable robot consisting of both global and local path planners. The global path planner uses boustrophedon motion combined with the A * algorithm. The optimum grid positioning that maximizes the global coverage is determined through a Genetic Algorithm (GA). The local coverage planner performs continuous reconfig-uration of the robot to adequately cover obstacle zones while navigating through narrow spaces without collisions. A GA is used to determine the reconfiguration parameters of the robot at each instance of the local coverage. Simulation results confirm that the proposed method is effective in performing both global and local coverage path planning for improving the area coverage performance.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"292 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114398247","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870378
Patrik Kolenovsky, P. Bujok
In this paper, new strategy options are developed for the adaptive jSO algorithm. The proposed variant of jSO is based on the competition of a binomial and exponential crossover. Moreover, an Eigen transformation approach is employed in the selected crossover with a given probability. The proposed variant of jSO is applied to the CEC 2022 benchmark set, which contains 12 functions with dimensionality $D=10$, 20. The proposed algorithm found the optima values in seven problems out of 24. When comparing the new variant of jSO with the original jSO algorithm, nine functions were improved, where two of them significantly.
{"title":"An adaptive variant of jSO with multiple crossover strategies employing Eigen transformation","authors":"Patrik Kolenovsky, P. Bujok","doi":"10.1109/CEC55065.2022.9870378","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870378","url":null,"abstract":"In this paper, new strategy options are developed for the adaptive jSO algorithm. The proposed variant of jSO is based on the competition of a binomial and exponential crossover. Moreover, an Eigen transformation approach is employed in the selected crossover with a given probability. The proposed variant of jSO is applied to the CEC 2022 benchmark set, which contains 12 functions with dimensionality $D=10$, 20. The proposed algorithm found the optima values in seven problems out of 24. When comparing the new variant of jSO with the original jSO algorithm, nine functions were improved, where two of them significantly.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122623590","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870285
Brendan Rogers, N. Noman, S. Chalup, P. Moscato
Deep Neural Networks (DNN) require specifically tuned architectures and hyperparameters when being applied to any given task. Nature-inspired algorithms have been successfully applied for optimising various hyperparameters in different types of DNNs such as convolutional and recurrent for sentence classification. Hybrid networks, which contain multiple types of neural architectures have more recently been used for sentence classification in order to achieve better performance. However, the inclusion of hybrid architectures creates numerous possibilities of designing the network and those sub-networks also need fine-tuning. At present these hybrid networks are designed manually and various organisation attempts are noticed. In order to understand the benefit and the best design principle of such hybrid DNNs for sentence classification, in this work we used an Evolutionary Algorithm (EA) to optimise the topology and various hyperparameters in different types of layers within the network. In our experiments, the proposed EA designed the hybrid networks by using a single dataset and evaluated the evolved networks on multiple other datasets to validate their generalisation capability. We compared the EA-designed hybrid networks with human-designed hybrid networks in addition to other EA-optimised and expert-designed non-hybrid architectures.
{"title":"Joint Optimization of Topology and Hyperparameters of Hybrid DNNs for Sentence Classification","authors":"Brendan Rogers, N. Noman, S. Chalup, P. Moscato","doi":"10.1109/CEC55065.2022.9870285","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870285","url":null,"abstract":"Deep Neural Networks (DNN) require specifically tuned architectures and hyperparameters when being applied to any given task. Nature-inspired algorithms have been successfully applied for optimising various hyperparameters in different types of DNNs such as convolutional and recurrent for sentence classification. Hybrid networks, which contain multiple types of neural architectures have more recently been used for sentence classification in order to achieve better performance. However, the inclusion of hybrid architectures creates numerous possibilities of designing the network and those sub-networks also need fine-tuning. At present these hybrid networks are designed manually and various organisation attempts are noticed. In order to understand the benefit and the best design principle of such hybrid DNNs for sentence classification, in this work we used an Evolutionary Algorithm (EA) to optimise the topology and various hyperparameters in different types of layers within the network. In our experiments, the proposed EA designed the hybrid networks by using a single dataset and evaluated the evolved networks on multiple other datasets to validate their generalisation capability. We compared the EA-designed hybrid networks with human-designed hybrid networks in addition to other EA-optimised and expert-designed non-hybrid architectures.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"150 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122913652","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870306
G. Filippi, M. Vasile, E. Patelli, M. Fossati
This paper presents a novel approach to the gener-ative design optimisation of a resilient Drone Logistic Network (DLN) for the delivery of medical equipment in Scotland. A DLN is a complex system composed of a high number of different classes of drones and ground infrastructures. The corresponding DLN model is composed of a number of interconnected digital twins of each one of these infrastructures and vehicles, forming a single digital twin of the whole logistic network. The paper proposes a multi-agent bio-inspired optimisation approach based on the analogy with the Physarum Policefalum slime mould that incrementally generates and optimise the DLN. A graph theory methodology is also employed to evaluate the network resilience where random failures, and their cascade effect, are simulated. The different conflicting objectives are aggregated into a single global performance index by using Pascoletti-Serafini scalarisation.
{"title":"Generative Optimisation of Resilient Drone Logistic Networks","authors":"G. Filippi, M. Vasile, E. Patelli, M. Fossati","doi":"10.1109/CEC55065.2022.9870306","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870306","url":null,"abstract":"This paper presents a novel approach to the gener-ative design optimisation of a resilient Drone Logistic Network (DLN) for the delivery of medical equipment in Scotland. A DLN is a complex system composed of a high number of different classes of drones and ground infrastructures. The corresponding DLN model is composed of a number of interconnected digital twins of each one of these infrastructures and vehicles, forming a single digital twin of the whole logistic network. The paper proposes a multi-agent bio-inspired optimisation approach based on the analogy with the Physarum Policefalum slime mould that incrementally generates and optimise the DLN. A graph theory methodology is also employed to evaluate the network resilience where random failures, and their cascade effect, are simulated. The different conflicting objectives are aggregated into a single global performance index by using Pascoletti-Serafini scalarisation.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132545347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870241
Igor L. S. Russo, H. Barbosa
Bi-level programming (BLP) is a hierarchical decision-making problem in which part of the constraints is determined by solving other optimization problems. Classic op-timization techniques cannot be applied directly, while standard metaheuristics often demand high computational costs. The transfer optimization paradigm uses the experience acquired when solving one optimization problem to speed up a distinct but related task. In particular, the multitasking technique ad-dresses two or more optimization tasks simultaneously to explore similarities and improve convergence. BLPs can benefit from multitasking as many (potentially similar) lower-level problems must be solved. Recently, several studies used surrogate methods to save expensive upper-level function evaluations in BLPs. This work proposes an algorithm based on Differential Evolution supported by transfer optimization and surrogate models to solve BLPs more efficiently. Experiments show a reduction of up to 86% regarding the number of function evaluations of the upper-level problem while achieving similar or superior accuracy when compared to state-of-the-art solvers.
{"title":"A multitasking surrogate-assisted differential evolution method for solving bi-level optimization problems","authors":"Igor L. S. Russo, H. Barbosa","doi":"10.1109/CEC55065.2022.9870241","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870241","url":null,"abstract":"Bi-level programming (BLP) is a hierarchical decision-making problem in which part of the constraints is determined by solving other optimization problems. Classic op-timization techniques cannot be applied directly, while standard metaheuristics often demand high computational costs. The transfer optimization paradigm uses the experience acquired when solving one optimization problem to speed up a distinct but related task. In particular, the multitasking technique ad-dresses two or more optimization tasks simultaneously to explore similarities and improve convergence. BLPs can benefit from multitasking as many (potentially similar) lower-level problems must be solved. Recently, several studies used surrogate methods to save expensive upper-level function evaluations in BLPs. This work proposes an algorithm based on Differential Evolution supported by transfer optimization and surrogate models to solve BLPs more efficiently. Experiments show a reduction of up to 86% regarding the number of function evaluations of the upper-level problem while achieving similar or superior accuracy when compared to state-of-the-art solvers.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"99 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124088359","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870359
Ye Tian, Haowen Chen, Xiaoshu Xiang, Hao Jiang, Xing-yi Zhang
Evolutionary algorithms and mathematical programming methods are currently the most popular optimizers for solving continuous optimization problems. Owing to the population based search strategies, evolutionary algorithms can find a set of promising solutions without using any problem-specific information. By contrast, with the assistance of gradient and other information of the functions, mathematical programming methods can quickly converge to a single optimum. While these two types of optimizers have their own advantages and disadvantages, the performance comparison between them is rarely touched. It is known that gradient descent methods generally converge faster than evolutionary algorithms, but when can evolutionary algorithms outperform gradient descent methods? How is the scalability of them? To answer these questions, this paper first gives a review of popular evolutionary algorithms and mathematical programming methods, then conducts several experiments to compare their performance from various aspects, and finally draws some conclusions.
{"title":"A Comparative Study on Evolutionary Algorithms and Mathematical Programming Methods for Continuous Optimization","authors":"Ye Tian, Haowen Chen, Xiaoshu Xiang, Hao Jiang, Xing-yi Zhang","doi":"10.1109/CEC55065.2022.9870359","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870359","url":null,"abstract":"Evolutionary algorithms and mathematical programming methods are currently the most popular optimizers for solving continuous optimization problems. Owing to the population based search strategies, evolutionary algorithms can find a set of promising solutions without using any problem-specific information. By contrast, with the assistance of gradient and other information of the functions, mathematical programming methods can quickly converge to a single optimum. While these two types of optimizers have their own advantages and disadvantages, the performance comparison between them is rarely touched. It is known that gradient descent methods generally converge faster than evolutionary algorithms, but when can evolutionary algorithms outperform gradient descent methods? How is the scalability of them? To answer these questions, this paper first gives a review of popular evolutionary algorithms and mathematical programming methods, then conducts several experiments to compare their performance from various aspects, and finally draws some conclusions.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127949079","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-18DOI: 10.1109/CEC55065.2022.9870257
V. Parque
The quest for the efficient adaptation of multilegged robotic systems to changing conditions is expected to render new insights into robotic control and locomotion. In this paper, we study the performance frontiers of the enumerative (factorial) encoding of hexapod gaits for fast recovery to conditions of leg failures. Our computational studies using five nature-inspired gradient-free optimization heuristics have shown that it is possible to render feasible recovery gait strategies that achieve minimal deviation to desired locomotion directives with a few evaluations (trials). For instance, it is possible to generate viable recovery gait strategies reaching 2.5 cm, (10 cm.) deviation on average with respect to a commanded direction with 40 – 60 (20) evaluations/trials. Our results are the potential to enable efficient adaptation to new conditions and to explore further the canonical representations for adaptation in robotic locomotion problems.
{"title":"Towards Hexapod Gait Adaptation using Enumerative Encoding of Gaits: Gradient-Free Heuristics","authors":"V. Parque","doi":"10.1109/CEC55065.2022.9870257","DOIUrl":"https://doi.org/10.1109/CEC55065.2022.9870257","url":null,"abstract":"The quest for the efficient adaptation of multilegged robotic systems to changing conditions is expected to render new insights into robotic control and locomotion. In this paper, we study the performance frontiers of the enumerative (factorial) encoding of hexapod gaits for fast recovery to conditions of leg failures. Our computational studies using five nature-inspired gradient-free optimization heuristics have shown that it is possible to render feasible recovery gait strategies that achieve minimal deviation to desired locomotion directives with a few evaluations (trials). For instance, it is possible to generate viable recovery gait strategies reaching 2.5 cm, (10 cm.) deviation on average with respect to a commanded direction with 40 – 60 (20) evaluations/trials. Our results are the potential to enable efficient adaptation to new conditions and to explore further the canonical representations for adaptation in robotic locomotion problems.","PeriodicalId":153241,"journal":{"name":"2022 IEEE Congress on Evolutionary Computation (CEC)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128974589","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：