A simple offline multi-Ego vehicle data generation script for the CARLA simulator.
To use the script you need to run multi_data_generator.py. The configuration variables are listed inside the Class SimulationParams in configuration.py.
From there you can control various aspects of the simulation, like the number of pedestrians, vehicles, select starting town, etc.
For each Ego vehicle in the simulation that gathers data (minimum 1) it needs to have an associated spawn point inside the simulation and also an associated
json file that describes sensors and their extrinsic and intrinsic parameters (for example, see sensors.json in the Config folder). Both the spawn point and the json file path must be defined at ego_vehicle_spawn_point
and sensor_json_filepath respectively in Class SimulationParams in configuration.py.
My implementation uses ques and synchronous communication with the server such that the sensor output is always exactly syncrhonized with the world clock
python3 multi_data_generator.py and a folder _out will be created with the respective Ego and sensor data.