SUMMIT: A Simulator for Urban Driving in Massive Mixed Traffic

Cai, Panpan; Lee, Yiyuan; Luo, Yuanfu; Hsu, David

doi:10.48550/arxiv.1911.04074

Cited by 1 publication

(2 citation statements)

References 31 publications

(47 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The task of driving in an urban crowd is formulated as a POMDP with hidden states encoding the intentions and attentions of exo-agents. The formulation is modified from the Context-POMDP model described in [4] by replacing the steering actions with lane decisions (changing / keeping). This action space better leverages the structure of urban roads and thus reduce the computational complexity.…”

Section: Pomdp For Driving In An Urban Crowdmentioning

confidence: 99%

“…In our experiments, we have tested both open-loop and close-loop integration of planning and learning as well as both supervision and reinforcement learners for crowd-driving in urban environments. Due to the difficulties in testing crowddriving in the real-world, we validate LeTS-Drive in a realistic simulator, SUMMIT [4]. SUMMIT simulates dense, unregulated urban traffic at worldwide locations supported by the OpenStreetMap.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Closing the Planning-Learning Loop with Application to Autonomous Driving

Cai¹,

Hsu²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Imagine an autonomous robot vehicle driving in dense, possibly unregulated urban traffic. To contend with an uncertain, interactive environment with many traffic participants, the robot vehicle has to perform long-term planning in order to drive effectively and approach human-level performance. Planning explicitly over a long time horizon, however, incurs prohibitive computational cost and is impractical under realtime constraints. To achieve real-time performance for large-scale planning, this paper introduces Learning from Tree Search for Driving (LeTS-Drive), which integrates planning and learning in a close loop. LeTS-Drive learns a driving policy from a planner based on sparsely-sampled tree search. It then guides online planning using this learned policy for real-time vehicle control. These two steps are repeated to form a close loop so that the planner and the learner inform each other and both improve in synchrony. The entire algorithm evolves on its own in a self-supervised manner, without explicit human efforts on data labeling. We applied LeTS-Drive to autonomous driving in crowded urban environments in simulation. Experimental results clearly show that LeTS-Drive outperforms either planning or learning alone, as well as open-loop integration of planning and learning

show abstract

Section: Pomdp For Driving In An Urban Crowdmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%