carla VS AirSim

It sounds like you're looking for something like the CARLA simulator.

If you are looking for a different simulator, I would suggest using (Carla)[https://carla.org/] with ROS bridge and it also has an inbuilt support for OSM which worked flawlessly (you have to install it from source to get the OSM plugin).

(3) I would suggest you use CARLA or TORCS for self-driving cars in RL as they are common test beds.

carla env

Link to carla

There are a few ways you could go with this. A fun recent trend has been to make simulators in the unreal engine for photorealistic training. If you wanted to spend your whole project on the simulator part you could make your own environment, but I highly recommend using an open-source sim package. If you don't care too much about photorealism, you could use gazebo just fine. I've also made small visual worlds in blender, then simulated with RVIZ. Here's a good one with a focus on aerial robotics: https://theairlab.org/tartanair-dataset/ You could also give CARLA a spin for autonomous vehicles: https://carla.org/

Have created all widgets, code compile successfully but have problem with replay spectator. Im using plugin AirSim https://github.com/microsoft/AirSim it is plugin that simulate drone. It has own gamemode AirsimGamemode (here is c++ file https://github.com/microsoft/AirSim/tree/main/Unreal/Plugins/AirSim/Source) that spawn drone at start. Problem is that I cant assign BP_PC_Spectator there. So I created own gamemode that I thought will override AirSimGameMode. It kinda did, it spawned drone on start but replay spectator widget show record screen, but it is not shown on play replay screen and I cannot move there as well.

Hi, I implemented path planner as node in ROS. Now I want to try the route planner in the AirSim simulator. During path execution, I want to get outputs from sensors such as GPS, IMU and Lidar. AirSim come with build in wrapper (https://github.com/microsoft/AirSim/tree/main/ros/src/airsim_ros_pkgs) that create topic and services once launched. Wrapper create two nodes, one to obtain sensors data and one to control drone. Wrapper is build in AirSim directory and path planer is in another. Is it possible to share msg so I can call and subscribe to wrapper topics in my planner node ? Or do I have to write msg for every topic, service I want to use ?

And there are also some open-source projects you could join, instead of starting a new one. This one looks interesting, haven't tried it though: https://github.com/Microsoft/AirSim

In 2017 Microsoft created AirSim, an open-source simulation platform for AI research and experimentation with drones and cars. This July, Microsoft announced the upcoming release of a new simulation platform and the archive of the original 2017 AirSim.

AirSim: https://github.com/microsoft/AirSim

AirSim

I've been trying for a while now and I started thinking it may not be possible. If anyone has managed to train a self-driving car simulator using openai gym on google colab(preferably), or on any remote server (AWS, GCP, ...) please let me know. So far, I tried carla, airsim, svl, deepdrive and they are all equally useless unless run locally with a gui. I'd really appreciate if someone suggests some way that actually can make it possible.