Wait one tick before creating carla actors to ensure the transforms of
the newly spawned actors have been updated by the server

and fix task_queue assignment

object_type attribute is empty, therefore guess the object class based
on the blueprint name as a workaround

Fill the message header of CarlaEgoVehicleControl messages with header
received from odometry

Some robustness increases:
- make vehicle_pid_controller robust against missing speed/pose input

Don't execute planner step if no odometry message was received

The number of vegetation bouunding boxes can explode due to the
duplication of small vegetation pieces. Millions of bounding boxes are
not useful, taking minutes to load and consuming gigabytes of memory.

… into berndgassmann/dev

…rndgassmann/dev

…ssmann/dev

…o berndgassmann/dev

…mann/dev

This allows for using multiple carla_ad_agents within a scenario.

Since Carla native interface directly accepts ackermann commands, this
is not required anymore.

To replace /carla/markers published by the old ros bridge.

- make local_planner subscription to CarlaVehicleStatus best effort and
extend ros compatibility qos setting
- adapt cala_walker_agent to carla native ros interface
- fix derived_objects_visualizer

This provides the pseudo.speedometer and pseudo.odom sensors for better
backward compatibility and usability.

…nario runner

allowing to select the vehicle to be manually controlled