Affiliation:
1. Tijuana Institute of Technology, Tecnólogico Nacional de México, Tijuana 22414, Mexico
2. Department of Computer Engineering, Robotics and Automation, University of Granada, 18071 Granada, Spain
Abstract
Designing a controller is typically an iterative process during which engineers must assess the performance of a design through time-consuming simulations; this becomes even more burdensome when using a population-based metaheuristic that evaluates every member of the population. Distributed algorithms can mitigate this issue, but these come with their own challenges. This is why, in this work, we propose a distributed and asynchronous bio-inspired algorithm to execute the simulations in parallel, using a multi-population multi-algorithmic approach. Following a cloud-native pattern, isolated populations interact asynchronously using a distributed message queue, which avoids idle cycles when waiting for other nodes to synchronize. The proposed algorithm can mix different metaheuristics, one for each population, first because it is possible and second because it can help keep total diversity high. To validate the speedup benefit of our proposal, we optimize the membership functions of a fuzzy controller for the trajectory tracking of a mobile autonomous robot using distributed versions of genetic algorithms, particle swarm optimization, and a mixed-metaheuristic configuration. We compare sequential versus distributed implementations and demonstrate the benefits of mixing the populations with distinct metaheuristics. We also propose a simple migration strategy that delivers satisfactory results. Moreover, we compare homogeneous and heterogenous configurations for the populations’ parameters. The results show that even when we use random heterogeneous parameter configuration in the distributed populations, we obtain an error similar to that in other work while significantly reducing the execution time.
Subject
Physics and Astronomy (miscellaneous),General Mathematics,Chemistry (miscellaneous),Computer Science (miscellaneous)
Reference52 articles.
1. A New Approach to Linear Filtering and Prediction Problems;Kalman;J. Basic Eng.,1960
2. Learning control systems and intelligent control systems: An intersection of artifical intelligence and automatic control;Fu;IEEE Trans. Autom. Control,1971
3. Application of fuzzy algorithms for control of simple dynamic plant;Mamdani;Proc. Inst. Electr. Eng.,1974
4. Driankov, D., and Saffiotti, A. (2013). Fuzzy Logic Techniques for Autonomous Vehicle Navigation, Physica.
5. Sutton, R.S., and Barto, A.G. (2018). Reinforcement Learning: An Introduction, MIT Press.
Cited by
6 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献