Merge branch 'master' of https://github.com/BerkeleyAutomation/Urban_…

…Driving_Simulator

README.md

@@ -1,10 +1,10 @@
 [![Documentation Status](https://readthedocs.org/projects/urban-driving-simulator/badge/?version=latest)](http://urban-driving-simulator.readthedocs.io/en/latest/?badge=latest)
-
 [![Build Status](https://travis-ci.org/jerryz123/Urban_Driving_Simulator.svg?branch=master)](https://travis-ci.org/jerryz123/Urban_Driving_Simulator)
 
+[**Documentation**](urban-driving-simulator.rtfd.io)
+
 # Urban Driving Simulator
 This repository contains a first order driving simulator for autonomous driving in urban environments. The package implements a OpenAI Gym environment describing a city scene with cars and pedestrians.
 
 
 We are currently under heavy development.
-

docs/source/observations.rst

@@ -0,0 +1,18 @@
+Observations
+============
+Observations are generated for each controlled car in the scene. The type of observation can be specified in the config file as 'raw', 'Q-LIDAR', or 'bitmap'. The step function in the environment returns an observation for each controlled car in the scene as a list. 
+
+Raw
+^^^
+A pointer to the raw environment, giving agents full access to the scene, and all other objects in the scene.
+
+Q-LIDAR
+^^^^^^^
+A representation based on features a autonomus vehicle might extract from LIDAR sensors, which is the relative distance to collideable objects in the scene. This is a numpy array of distances produced by a Featurizer. The density and range of the Q-LIDAR beams can be configured in the featurizer.
+
+.. autoclass:: gym_urbandriving.utils.featurizer.Featurizer
+   :members: featurize
+
+Bitmap
+^^^^^^
+Returns a Numpy image array as generated by the visualizer, for vision-based control agents. Image is a top-down view of the intersection.

gym_urbandriving/utils/featurizer.py

@@ -9,18 +9,29 @@
 import os
 
 
-N_ARCS = 9
-ARC_DELTAS = [-1, -0.5, -0.25, -0.20, -0.15, -0.1, -0.05, 0, 0.05, 0.1, 0.15, 0.20, 0.25, 0.5, 1]
-ARC_DELTAS = [i*np.pi/2 for i in ARC_DELTAS]
-BEAM_DISTANCE = 300
+
 LIGHT_ARC = np.pi / 16
 LIGHT_DISTANCE = 300
 
 def distance(p1, p2):
     return ((p1[0] - p2[0])**2 + (p1[1] - p2[1])**2)**0.5
 
 class Featurizer(object):
-    def __init__(self):
+    """
+    Object to convert a state observation into a Q-LIDAR observation.
+
+    Attributes
+    ----------
+    beam_distance : int
+        How far each "LIDAR" beam will project into the scene
+    n_arcs :
+        How many "LIDAR" beams to project around the car
+    """
+    def __init__(self, beam_distance=300, n_arcs=9):
+
+        self.arc_deltas = np.arange(n_arcs + 1) / (float(n_arcs) / 2) - 1
+        self.arc_deltas = [i*np.pi/2 for i in self.arc_deltas]
+        self.beam_distance = beam_distance
         pass
 
 
@@ -59,12 +70,12 @@ def featurize(self, current_state, controlled_key):
         features = [vel]
         #print(goalx, goaly)
 
-        for arc_delta in ARC_DELTAS:
+        for arc_delta in self.arc_deltas:
             arc_angle = angle + arc_delta
-            xd = x + np.cos(arc_angle)*BEAM_DISTANCE
-            yd = y - np.sin(arc_angle)*BEAM_DISTANCE
+            xd = x + np.cos(arc_angle)*self.beam_distance
+            yd = y - np.sin(arc_angle)*self.beam_distance
             linestring = shapely.geometry.LineString([(x, y), (xd, yd)])
-            min_coll_d = BEAM_DISTANCE
+            min_coll_d = self.beam_distance
             min_coll_type = None
             min_coll_vel = 0
             min_coll_angle = 0

gym_urbandriving/visualizer/pygame_vis.py

@@ -126,9 +126,9 @@ def render_collisions(self, state, valid_area):
                                      pygame.SRCALPHA)
 
 
-        for obj1id, obj2id, k in dynamic_collisions:
-            obj1 = state.dynamic_objects[k][str(obj1id)]
-            obj2 = state.dynamic_objects[k][str(obj2id)]
+        for obj1id, obj2id, k1, k2 in dynamic_collisions:
+            obj1 = state.dynamic_objects[k1][str(obj1id)]
+            obj2 = state.dynamic_objects[k2][str(obj2id)]
             pygame.draw.circle(new_surface, (255, 0, 255), obj1.get_pos().astype(int), 5)
             pygame.draw.circle(new_surface, (255, 0, 255), obj2.get_pos().astype(int), 5)
         for obj1id, obj2id, k, _ in static_collisions: