Simulation-Based Reinforcement Learning for Real-World Autonomous Driving

Osiński, Błażej; Jakubowski, Adam; Zięcina, Paweł; Miłoś, Piotr; Galias, Christopher; Homoceanu, Silviu; Michalewski, Henryk

doi:10.1109/icra40945.2020.9196730

Cited by 82 publications

(32 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We selected the two cases based on the simple fact that they are -to the best of our knowledge -some of the only instances where RL has been successfully deployed in practice. We intentionally did not consider cases of self-driving cars and user-recommendation systems, as they are either not yet production-ready (Osiński et al, 2020), or use simpler forms of RL, such as contextual bandits (Amat, Chandrashekar, Jebara, & Basilico, 2018).…”

Section: Methodsmentioning

confidence: 99%

Port Community System Business Models

Jović¹,

Aksentijević²,

Plentaj³

et al. 2021

34th Bled eConference Digital Support From Crisis to Progressive Change: Conference Proceedings

View full text Add to dashboard Cite

Port Community Systems have become a staple technological platform used to exchange information between the public and private agents and entities involved in ship and cargo services within seaports. In this paper, the theoretical background of the Port Community System is provided, emphasizing the importance of its implementation, and stakeholder collaboration. Different models of introducing an integrated Port Community System in seaports are analysed using literature review and actual cases in some of the most prominent seaports.

show abstract

Section: Methodsmentioning

confidence: 99%

Port Community System Business Models

Jović¹,

Aksentijević²,

Plentaj³

et al. 2021

34th Bled eConference Digital Support From Crisis to Progressive Change: Conference Proceedings

View full text Add to dashboard Cite

show abstract

Section: Methodsmentioning

confidence: 99%

Designing Call to Action: Users’ Perception of Different Characteristics

Mejtoft

Hedlund²,

Cripps

et al. 2021

34th Bled eConference Digital Support From Crisis to Progressive Change: Conference Proceedings

View full text Add to dashboard Cite

This paper aims to provide guidance when designing a call to action in a digital system with the purpose to create an intended feeling and user engagement. The paper is based on a test of four different simple call to action constructions. The users clearly prefer constructions that have a high explainability and feels intuitive. Hence, the design should have a high level of transparency and show the user straight away what it demands from the user and what the result of the action is. Furthermore, the design should have a high usability to make it clear how to use the call to action.

show abstract

“…This allows for a better exploration of the training process. Overall, model-free methods are popular because of their scalability, which makes it a promising tool for many high-dimensional tasks ( Osiński et al, 2020 ; Schoettler et al, 2020 ; Ye et al, 2020 ). However, as we will discuss in Section 7.1 , data inefficiency is a common problem for model-free methods.…”

Section: Related Workmentioning

confidence: 99%

Risk-Aware Model-Based Control

Chen

Rosendo

2021

Front. Robot. AI

View full text Add to dashboard Cite

Model-Based Reinforcement Learning (MBRL) algorithms have been shown to have an advantage on data-efficiency, but often overshadowed by state-of-the-art model-free methods in performance, especially when facing high-dimensional and complex problems. In this work, a novel MBRL method is proposed, called Risk-Aware Model-Based Control (RAMCO). It combines uncertainty-aware deep dynamics models and the risk assessment technique Conditional Value at Risk (CVaR). This mechanism is appropriate for real-world application since it takes epistemic risk into consideration. In addition, we use a model-free solver to produce warm-up training data, and this setting improves the performance in low-dimensional environments and covers the shortage of MBRL’s nature in the high-dimensional scenarios. In comparison with other state-of-the-art reinforcement learning algorithms, we show that it produces superior results on a walking robot model. We also evaluate the method with an Eidos environment, which is a novel experimental method with multi-dimensional randomly initialized deep neural networks to measure the performance of any reinforcement learning algorithm, and the advantages of RAMCO are highlighted.

show abstract

Simulation-Based Reinforcement Learning for Real-World Autonomous Driving

Cited by 82 publications

References 22 publications

Port Community System Business Models

Port Community System Business Models

Designing Call to Action: Users’ Perception of Different Characteristics

Risk-Aware Model-Based Control

Contact Info

Product

Resources

About