Policy-Based Reinforcement Learning for Training Autonomous Driving Agents in Urban Areas With Affordance Learning

Ahmed, Mohammed; Abobakr, Ahmed; Lim, Chee Peng; Nahavandi, Saeid

doi:10.1109/tits.2021.3115235

Cited by 10 publications

(17 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 5 contains the results for the testing conditions, and here, Ahmed et al [37] achieved the best results, surpassing Chen et al [36] in two tasks. Agarwal et al [14] also achieved excellent scores in these conditions, achieving the same score as Ahmed et al [37] in the Regular and Dense tasks. In the Empty and Regular tasks, all approaches have achieved good results, always above 80%.…”

Section: Discussionmentioning

confidence: 85%

“…More recently, some authors have also used the aforementioned affordances to tackle the high-dimensional data issue of RL. Ahmed et al, proposed an end-to-end AD system comprised of two major components: supervised network and a Deep RL agent (DDPG) [37]. The supervised network encodes RGB images into a set of affordances.…”

Section: A Architecturesmentioning

confidence: 99%

“…Simplified version of the system proposed by Ahmed et al [37]. The system receives as input a stack of RGB images, vehicle measurements and navigational commands.…”

Section: Figurementioning

confidence: 99%

“…Multiple authors have also used high-level measurements about the state of the vehicle, such as current velocity or acceleration [37], [39], [44]. When the model only considers one frame to make a decision, the usage of the current velocity can be highly useful [24].…”

Section: Figurementioning

confidence: 99%

“…Over the years, the interest in the end-to-end approach for AD was scarce, especially compared with the amount of research done in modular approaches. However, due to the rise of Artificial Intelligence [28]- [30] in the past years, and due to the recent developments of Deep Reinforcement Learning (Deep RL) [31]- [33] by Deep Mind [34], [35], end-to-end approaches have begun to show promising results [36], [37].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Review of End-to-End Autonomous Driving in Urban Environments

Coelho

Oliveira

2022

IEEE Access

View full text Add to dashboard Cite

Autonomous driving in urban environments requires intelligent systems that are able to deal with complex and unpredictable scenarios. Traditional modular approaches focus on dividing the driving task into standard modules, and then use rule-based methods to connect those different modules. As such, these approaches require a significant effort to design architectures that combine all system components, and are often prone to error propagation throughout the pipeline. Recently, end-to-end autonomous driving systems have formulated the autonomous driving problem as an end-to-end learning process, with the goal of developing a policy that transforms sensory data into vehicle control commands. Despite promising results, the majority of end-to-end works in autonomous driving focus on simple driving tasks, such as lane-following, which do not fully capture the intricacies of driving in urban environments. The main contribution of this paper is to provide a detailed comparison between end-to-end autonomous driving systems that tackle urban environments. This analysis comprises two stages: a) a description of the main characteristics of the successful end-to-end approaches in urban environments; b) a quantitative comparison based on two CARLA simulator benchmarks (CoRL2017 and NoCrash). Beyond providing a detailed overview of the existent approaches, we conclude this work with the most promising aspects of end-to-end autonomous driving approaches suitable for urban environments.INDEX TERMS autonomous driving, end-to-end, imitation learning, reinforcement learning, urban environments.

show abstract

Section: Discussionmentioning

confidence: 85%

Section: A Architecturesmentioning

confidence: 99%

“…Simplified version of the system proposed by Ahmed et al [37]. The system receives as input a stack of RGB images, vehicle measurements and navigational commands.…”

Section: Figurementioning

confidence: 99%