Poster: Unraveling Reward Functions for Head-to-Head Autonomous Racing in AWS DeepRacer

Tian, Allen; John, Eddy Guerra; Yang, Kecheng

doi:10.1145/3565287.3617987

Cited by 1 publication

(2 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There was also research oriented on multi-agent autonomous navigation on the Deep-Racer platform [55,56]. This article focusses on vehicle platoon state control strategy and scheduling, using Gazebo simulations [57], and introduces novel algorithms for seamless state switching, including strategies for convoy formation, disbandment, and reordering, ensuring accurate navigation in varied scenarios such as overpass transits.…”

Section: State Of the Art Reviewmentioning

confidence: 99%

See 1 more Smart Citation

Deep Reinforcement Learning for Autonomous Driving in Amazon Web Services DeepRacer

Petryshyn,

Postupaiev,

Ben Bari

et al. 2024

Information

View full text Add to dashboard Cite

The development of autonomous driving models through reinforcement learning has gained significant traction. However, developing obstacle avoidance systems remains a challenge. Specifically, optimising path completion times while navigating obstacles is an underexplored research area. Amazon Web Services (AWS) DeepRacer emerges as a powerful infrastructure for engineering and analysing autonomous models, providing a robust foundation for addressing these complexities. This research investigates the feasibility of training end-to-end self-driving models focused on obstacle avoidance using reinforcement learning on the AWS DeepRacer autonomous race car platform. A comprehensive literature review of autonomous driving methodologies and machine learning model architectures is conducted, with a particular focus on object avoidance, followed by hands-on experimentation and the analysis of training data. Furthermore, the impact of sensor choice, reward function, action spaces, and training time on the autonomous obstacle avoidance task are compared. The results of the best configuration experiment demonstrate a significant improvement in obstacle avoidance performance compared to the baseline configuration, with a 95.8% decrease in collision rate, while taking about 79% less time to complete the trial circuit.

show abstract

Section: State Of the Art Reviewmentioning

confidence: 99%

“…where π θ k is the policy at the previous step and k, s, and a are the states and actions defined in the state and action spaces of the environment. The optimized policy is a policy that maximizes the expected value of L(s, a, θ k , θ) which is defined as follows [56]:…”

Section: Proximal Policy Optimisation Algorithmmentioning

confidence: 99%