A minimalist environment for decision-making in autonomous driving
Huge thanks to contributors @zerongxi, @TibiGG, @KexianShen, @lorandcheng
TupleObservation
, which is a union of several observation typesLidarObservation
PolyLane
, and methods to save/load road networks from a configGoalEnv
interface which was removed from gymThis release introduces additional content:
racetrack-v0
, where the agent must learn to steer and follow the tracks, while avoiding other vehicles"on_road"
layer in the OccupancyGrid
observation type, which enables the observer to see the drivable space"align_to_vehicle_axes"
option in the OccupancyGrid
observation type, which renders the observation in the local vehicle frameDiscreteAction
action type, which discretizes the original ContinuousAction
type. This allows to do low-level control, but with a small discrete action space (e.g. for DQN). Note that this is different from the DiscreteMetaAction
type, which implements its own low-level sub-policies.This release contains
highway-fast-v0
: a faster variant of highway-v0 to train/debug models more quicklyMinor update with