@article{20,
  author = {Aravind Balakrishnan and Jaeyoung Lee and Ashish Gaurav and Krzysztof Czarnecki and Sean Sedwards},
  title = {Transfer Reinforcement Learning for Autonomous Driving: From WiseMove to WiseSim},
  abstract = {<p>Reinforcement learning (RL) is an attractive way to implement high-level decision-making policies for autonomous driving, but learning directly from a real vehicle or a high-fidelity simulator is variously infeasible. We therefore consider the problem of transfer reinforcement learning and study how a policy learned in a simple environment using\&nbsp;WiseMove\&nbsp;can be transferred to our high-fidelity simulator, WiseMove.\&nbsp;WiseMove\&nbsp;is a framework to study safety and other aspects of RL for autonomous driving. WiseMoveaccurately reproduces the dynamics and software stack of our real vehicle.\&nbsp;</p>

<p>We find that the accurately modelled perception errors in WiseMove\&nbsp;contribute the most to the transfer problem. These errors, when even naively modelled in\&nbsp;WiseMove, provide an RL policy that performs better in WiseMove\&nbsp;than a hand-crafted rule-based policy. Applying domain randomization to the environment in\&nbsp;WiseMove\&nbsp;yields an even better policy. The final RL policy reduces the failures due to perception errors from 10\% to 2.75\%. We also observe that the RL policy has significantly less reliance on velocity compared to the rule-based policy, having learned that its measurement is unreliable.</p>
},
  year = {2021},
  journal = {ACM Transactions on Modeling and Computer Simulation},
  volume = {31},
  pages = {Article No. 15, pp 1~26},
  month = {07/2021},
  issn = {1049-3301},
  doi = {10.1145/3449356},
}