3rd. Maps and navigation
After discussing about the world and its actors, it is time to put everything into place and understand the map and how do the actors navigate it.
- The map
- Navigation in CARLA
- CARLA maps
A map includes both the 3D model of a town and its road definition. Every map is based on an OpenDRIVE file describing the road layout fully annotated. The way the OpenDRIVE standard 1.4 defines roads, lanes, junctions, etc. is extremely important. It determines the possibilities of the API and the reasoning behind decisions made.
The Python API makes for a high level querying system to navigate these roads. It is constantly evolving to provide a wider set of tools.
Changing the map
To change the map, the world has to change too. Everything will be rebooted and created from scratch, besides the Unreal Editor itself. There are two ways to do so.
reload_world()creates a new instance of the world with the same map.
load_world()changes the current map and creates a new world.
world = client.load_world('Town01')
The client can get a list of available maps. Each map has a
name attribute that matches the name of the currently loaded city, e.g. Town01.
The traffic signs defined in the OpenDRIVE file are translated into CARLA as landmark objects that can be queried from the API. In order to facilitate their manipulation, there have been several additions to it.
- carla.Landmark objects represent the OpenDRIVE signals. The attributes and methods describe the landmark, and where it is effective.
- A carla.Waypoint can get landmarks located a certain distance ahead of it. The type of landmark can be specified.
- The carla.Map retrieves sets of landmarks. It can return all the landmarks in the map, or those having an ID, type or group in common.
- The carla.World acts as intermediary between landmarks, and the carla.TrafficSign and carla.TrafficLight that embody them in the simulation.
The lane markings surrounding a lane can be accessed through carla.LaneMarking. These are defined with a series of variables.
- carla.LaneMarkingType are enum values according to OpenDRIVE standards.
- carla.LaneMarkingColor are enum values to determine the color of the marking.
- width to state thickness of the marking.
- carla.LaneChange to state permissions to perform lane changes.
Waypoints use these to aknowledge traffic permissions.
# Get the lane type where the waypoint is. lane_type = waypoint.lane_type # Get the type of lane marking on the left. left_lanemarking_type = waypoint.left_lane_marking.type() # Get available lane changes for this waypoint. lane_change = waypoint.lane_change
A carla.Junction represents an OpenDRIVE junction. This class provides for a bounding box to state whereas lanes or vehicles are inside of it.
The most remarkable method of this class returns a pair of waypoints per lane inside the junction. Each pair is located at the starting and ending point of the junction boundaries.
waypoints_junc = my_junction.get_waypoints()
A carla.Waypoint is a 3D-directed point. These are prepared to mediate between the world and the openDRIVE definition of the road.
Each waypoint contains a carla.Transform. This states its location on the map and the orientation of the lane containing it. The variables
s translate this transform to the OpenDRIVE road. These combined, create the
id of the waypoint.
Due to granularity, waypoints closer than 2cm within the same road share the same
A waypoint also contains some information regarding the lane containing it. Specifically its left and right lane markings, and a boolean to determine if it is inside a junction.
# Examples of a waypoint accessing to lane information inside_junction = waypoint.is_junction() width = waypoint.lane_width right_lm_color = waypoint.right_lane_marking.color
Navigation in CARLA
Navigating through waypoints
Waypoints have a set of methods to connect with others and create a road flow. All of these methods follow traffic rules to determine only places where the vehicle can go.
next(d)creates a list of waypoints at an approximate distance
din the direction of the lane. The list contains one waypoint for each deviation possible.
previous(d)creates a list of waypoints waypoint at an approximate distance
don the opposite direction of the lane. The list contains one waypoint for each deviation possible.
previous_until_lane_start(d)returns a list of waypoints a distance
dapart. The list goes from the current waypoint to the end and start of its lane, respectively.
get_left_lane()return the equivalent waypoint in an adjacent lane, if any. A lane change maneuver can be made by finding the next waypoint to the one on its right/left lane, and moving to it.
# Disable physics, in this example the vehicle is teleported. vehicle.set_simulate_physics(False) while True: # Find next waypoint 2 meters ahead. waypoint = random.choice(waypoint.next(2.0)) # Teleport the vehicle. vehicle.set_transform(waypoint.transform)
Generating a map navigation
The instance of the map is provided by the world. It will be useful to create routes and make vehicles roam around the city and reach goal destinations.
map = world.get_map()
- Get recommended spawn points for vehicles pointed by developers. There is no ensurance that these spots will be free.
spawn_points = world.get_map().get_spawn_points()
- Get the closest waypoint to a specific location or to a certain
# Nearest waypoint on the center of a Driving or Sidewalk lane. waypoint01 = map.get_waypoint(vehicle.get_location(),project_to_road=True, lane_type=(carla.LaneType.Driving | carla.LaneType.Sidewalk)) #Nearest waypoint but specifying OpenDRIVE parameters. waypoint02 = map.get_waypoint_xodr(road_id,lane_id,s)
- Generate a collection of waypoints to visualize the city lanes. Creates waypoints all over the map, for every road and lane. All of them will be an approximate distance apart.
waypoint_list = map.generate_waypoints(2.0)
- Generate road topology. Returns a list of pairs (tuples) of waypoints. For each pair, the first element connects with the second one and both define the starting and ending point of each lane in the map.
waypoint_tuple_list = map.get_topology()
- Convert simulation point to geographical coordinates. Transforms a certain location to a carla.GeoLocation with latitude and longitude values.
my_geolocation = map.transform_to_geolocation(vehicle.transform)
- Save road information. Converts the road information to OpenDRIVE format, and saves it to disk.
info_map = map.to_opendrive()
So far there are seven different maps available. Each one has unique features and is useful for different purposes. Hereunder is a brief sum up on them.
|Town||Summary||Town01||A basic town layout with all "T junctions".|
|Town02||Similar to Town01, but smaller.|
|Town03||The most complex town, with a 5-lane junction, a roundabout, unevenness, a tunnel, and much more. Essentially a medley.|
|Town04||An infinite loop with a highway and a small town.|
|Town05||Squared-grid town with cross junctions and a bridge. It has multiple lanes per direction. Useful to perform lane changes.|
|Town06||Long highways with many highway entrances and exits. It also has a Michigan left.|
|Town07||A rural environment with narrow roads, barely non traffic lights and barns.|
|Town10||A city environment with with different environments such as an avenue or a promenade, and more realistic textures.|
That is a wrap as regarding maps and navigation in CARLA. The next step takes a closer look into sensors types, and the data they retrieve.
Keep reading to learn more or visit the forum to post any doubts or suggestions that have come to mind during this reading.