search

CollaborativePerception / V2Xverse

79 stars 10 forks source link

Agent control issues #22

Open ryhnhao opened 2 months ago

ryhnhao commented 2 months ago

Hello! Thanks for your great contribution!

I have two questions regarding to the agent control.

  1. The figure 1 indicates "customized autonomous agent" has "RL Driving Policy". However, the main text did not mention this and the code seems not including this part (I'm not sure about this, sorry for the possible misunderstanding)? image

  2. Based on your wonderful GitHub repo, could you please give me more instructions on how to generate these flexible scenarios?

    From the agent perspective, we set up some crazy pedestrians that ignore traffic rules and surrounding vehicles. We also include more large vehicles and roadside obstacles for diverse simulations. Additionally, we design several trigger-based scenarios extended from Carla Autonomous Driving Leaderboard setting [44], e.g. pedestrians suddenly appearing from occluded regions to cross the road, unexpected agents from occluded buildings, and unprotected turns at intersections, see Figure 2. We provide highly maneuverable APIs for these safety-critical scenarios.

Thank you very much! Looking forward to hearing from you. Best regards

gjliu9 commented 1 month ago

Hello! Thanks for your great contribution!

I have two questions regarding to the agent control.

  1. The figure 1 indicates "customized autonomous agent" has "RL Driving Policy". However, the main text did not mention this and the code seems not including this part (I'm not sure about this, sorry for the possible misunderstanding)? image
  2. Based on your wonderful GitHub repo, could you please give me more instructions on how to generate these flexible scenarios?

From the agent perspective, we set up some crazy pedestrians that ignore traffic rules and surrounding vehicles. We also include more large vehicles and roadside obstacles for diverse simulations. Additionally, we design several trigger-based scenarios extended from Carla Autonomous Driving Leaderboard setting [44], e.g. pedestrians suddenly appearing from occluded regions to cross the road, unexpected agents from occluded buildings, and unprotected turns at intersections, see Figure 2. We provide highly maneuverable APIs for these safety-critical scenarios.

Thank you very much! Looking forward to hearing from you. Best regards

Sorry for the late response and thank you for your attention. You can set the attribute of scenarios in simulation/leaderboard/leaderboard/scenarios/scenario_parameter.yaml, the position of trigger-based scenarios in simulation/leaderboard/data/scenarios. I hope this is helpful to you.