exoticdft / stanford-carla-toolbox

A toolbox for the Carla Simulator.

stanford-carla-toolbox

A toolbox for the Carla Simulator.

This toolbox is a collection of useful C++, Python and Julia wrappers for working with the Carla Simulator and other autonomous driving libraries and packages.

Prerequisites

These are the required prerequisites necessary for using the toolbox. A list of general requirements and well as extras for the desired coding language.

• Carla Simulator (0.9.7)
• Python (3.5) - Carla only provides API binaries for Python 3.5.
• Julia (1.2+)

Carla

The code has been verified to work with version 0.9.7 of the Carla simulator. For more information on using the Carla simulator, please visit the Carla documentation, forum, Discord server and GitHub page.

Python

The toolbox extensively utilizes the Carla PythonAPI for interacting with the Carla simulator.

• Carla PythonAPI egg - Found in the PythonAPI/carla/dist/ directory of the Carla simulator binary above. See "Setting Up Your Environment".

Julia

The Julia codebase for this toolbox has only been tested directly for version 1.3 of Julia. Since Carla doesn't provide direct Julia support, this wrapper currently uses their provided Python API. Please make sure your environment is setup to load the Carla provided .egg file.

• AutomotiveSimulator.jl
• PyCall.jl

Using the Toolbox

In order to use the toolbox, you must first have the above prerequisites installed on your system and follow the instructions provided in "Setting Up Your Environment".

Before running any code that uses the toolbox, we must have a instance of the Carla server running. Open a terminal with the appropriate environment and run:

~/Simulators/Carla-0.9.7/./CarlaUE4.sh

Next, open another terminal (or tab) with the appropriate environment and run your code using the toolbox. Below is a simple Julia program that will connect to a running Carla server and spawn a vehicle in the simulator.

julia --project=~/Development/VirtualEnvironments/Julia/sct-testing \
~/Development/stanford-carla-toolbox/examples/spawn-vehicle.jl

This example only provides a brief look at the functionality of the toolbox. For further examples, look under the examples directory.

The same idea can be applied to the Python examples.

python ~/Development/stanford-carla-toolbox/python/examples/spawn_vehicle.py

stanford-carla-toolbox's Issues

Create a visualize helper function

Create a "visualize" helper function that passes the scenes to visualize, a timestep in which the scenes were generated and any mapping from Entity.id to Carla id if necessary. Something like

visualize(scenes:` List, timestep: double, entity_map: Dict{Any, Int})
    for i = 1 : length(scenes)
        start_time = Now()
        scene = scenes[i]
        update_world(carla_world, scene, entity_map, handiness)
        sleep(start_time + timestep - Now())
   end

Convert OpenDrive to ADM roadways

In order for Carla to work with the simulation functionality of the AutomotiveDrivingModels repository we'll need to convert the OpenDrive format into the ADM roadways format. This will require parsing an XML file, or string of all the XML info, using the OpenDrive 1.5 specification.

OpenDrive: http://www.opendrive.org/docs/OpenDRIVEFormatSpecRev1.5M.pdf
Carla: https://carla.readthedocs.io/en/latest/python_api/#carlamap-class
ADM: https://github.com/sisl/AutomotiveDrivingModels.jl/tree/master/src/roadways

Add Python related wrapper functions to the codebase

There are several functions floating around in various scripts made to help with previous research projects. It would be ideal to add these functions into the codebase to start the Python related functionality of the toolbox.

Add functionality to attach spector to an actor

It would be nice to have a utility to attach the spectator (server camera) to an actor.

Add able to use AutomotiveSimulator.Scene objects in Carla.

The AutomotiveSimulator.Scene object stores all the information of the vehicles in AutomotiveSimulator (AS) at a given point in time. It would be great to be able to use this object and convert it directly into a Carla.World object. This would be useful for quick visualization of already run simulations and potentially a quick way to move vehicles manually in the Carla environment after each AS tick, instead of using the Carla Control commands.