000150246 001__ 150246 000150246 005__ 20250131203739.0 000150246 0247_ $$2doi$$a10.1109/LRA.2021.3060408 000150246 0248_ $$2sideral$$a124059 000150246 037__ $$aART-2021-124059 000150246 041__ $$aeng 000150246 100__ $$0(orcid)0000-0001-7251-0753$$aGarcia-Barcos, J.$$uUniversidad de Zaragoza 000150246 245__ $$aRobust Policy Search for Robot Navigation 000150246 260__ $$c2021 000150246 5203_ $$aComplex robot navigation and control problems can be framed as policy search problems. However, interactive learning in uncertain environments can be expensive, requiring the use of data-efficient methods. Bayesian optimization is an efficient nonlinear optimization method where queries are carefully selected to gather information about the optimum location. This is achieved by a surrogate model, which encodes past information, and the acquisition function for query selection. Bayesian optimization can be very sensitive to uncertainty in the input data or prior assumptions. In this letter, we incorporate both robust optimization and statistical robustness, showing that both types of robustness are synergistic. For robust optimization we use an improved version of unscented Bayesian optimization which provides safe and repeatable policies in the presence of policy uncertainty. We also provide new theoretical insights. For statistical robustness, we use an adaptive surrogate model and we introduce the Boltzmann selection as a stochastic acquisition method to have convergence guarantees and improved performance even with surrogate modeling errors. We present results in several optimization benchmarks and robot tasks. 000150246 536__ $$9info:eu-repo/grantAgreement/ES/MINECO-FEDER/RTI2018-096903-B-I00 000150246 540__ $$9info:eu-repo/semantics/closedAccess$$aAll rights reserved$$uhttp://www.europeana.eu/rights/rr-f/ 000150246 590__ $$a4.321$$b2021 000150246 591__ $$aROBOTICS$$b11 / 30 = 0.367$$c2021$$dQ2$$eT2 000150246 592__ $$a2.206$$b2021 000150246 593__ $$aArtificial Intelligence$$c2021$$dQ1 000150246 593__ $$aBiomedical Engineering$$c2021$$dQ1 000150246 593__ $$aMechanical Engineering$$c2021$$dQ1 000150246 593__ $$aControl and Optimization$$c2021$$dQ1 000150246 593__ $$aControl and Systems Engineering$$c2021$$dQ1 000150246 593__ $$aComputer Vision and Pattern Recognition$$c2021$$dQ1 000150246 594__ $$a8.0$$b2021 000150246 655_4 $$ainfo:eu-repo/semantics/article$$vinfo:eu-repo/semantics/publishedVersion 000150246 700__ $$0(orcid)0000-0002-6741-844X$$aMartinez-Cantin, R.$$uUniversidad de Zaragoza 000150246 7102_ $$15007$$2520$$aUniversidad de Zaragoza$$bDpto. Informát.Ingenie.Sistms.$$cÁrea Ingen.Sistemas y Automát. 000150246 7102_ $$15007$$2570$$aUniversidad de Zaragoza$$bDpto. Informát.Ingenie.Sistms.$$cÁrea Lenguajes y Sistemas Inf. 000150246 773__ $$g6, 2 (2021), 2389-2396$$pIEEE Robot. autom. let.$$tIEEE Robotics and Automation Letters$$x2377-3766 000150246 8564_ $$s2217369$$uhttps://zaguan.unizar.es/record/150246/files/texto_completo.pdf$$yVersión publicada 000150246 8564_ $$s3472630$$uhttps://zaguan.unizar.es/record/150246/files/texto_completo.jpg?subformat=icon$$xicon$$yVersión publicada 000150246 909CO $$ooai:zaguan.unizar.es:150246$$particulos$$pdriver 000150246 951__ $$a2025-01-31-20:05:32 000150246 980__ $$aARTICLE