TRC: Trust Region Conditional Value at Risk for Safe Reinforcement Learning

Author(s):  
Dohyeong Kim ◽  
Songhwai Oh
2014 ◽  
Vol 16 (6) ◽  
pp. 3-29 ◽  
Author(s):  
Samuel Drapeau ◽  
Michael Kupper ◽  
Antonis Papapantoleon

2014 ◽  
Vol 59 (2) ◽  
pp. 116-135 ◽  
Author(s):  
Stephan A. Trusevych ◽  
Roy H. Kwon ◽  
Andrew K. S. Jardine

2021 ◽  
Author(s):  
Pedram Eshaghieh Firoozabadi ◽  
sara nazif ◽  
Seyed Abbas Hosseini ◽  
Jafar Yazdi

Abstract Flooding in urban area affects the lives of people and could cause huge damages. In this study, a model is proposed for urban flood management with the aim of reducing the total costs. For this purpose, a hybrid model has been developed using SWMM and a quasi-two-dimensional model based on the cellular automata (CA) capable of considering surface flow infiltration. Based on the hybrid model outputs, the best management practices (BMPs) scenarios are proposed. In the next step, a damage estimation model has been developed using depth-damage curves. The amount of damage has been estimated for the scenarios in different rainfall return periods to obtain the damage and cost- probability functions. The conditional value at risk (CVaR) are estimated based on these functions which is the basis of decision making about the scenarios. The proposed model is examined in an urban catchment located in Tehran, Iran. In this study, five scenarios have been designed on the basis of different BMPs. It has been found that the scenario of permeable pavements has the lowest risk. The proposed model enables the decision makers to choose the best scenario with the minimum cost taking into account the risk associated with each scenario.


Sign in / Sign up

Export Citation Format

Share Document