search

Shuijing725 / CrowdNav_Prediction_AttnGraph

[ICRA 2023] Intention Aware Robot Crowd Navigation with Attention-Based Interaction Graph
https://sites.google.com/view/intention-aware-crowdnav/home
MIT License
163 stars 30 forks source link

part of the preprocessing module #15

Closed bsxlb6 closed 5 months ago

bsxlb6 commented 5 months ago

Excuse me，may I ask what part of the crowd preprocessing module is that?

Shuijing725 commented 5 months ago

If you mean the robot policy network, it is in https://github.com/Shuijing725/CrowdNav_Prediction_AttnGraph/blob/main/rl/networks/selfAttn_srnn_temp_node.py

If you mean the human simulation, it is in https://github.com/Shuijing725/CrowdNav_Prediction_AttnGraph/blob/main/crowd_sim/envs/crowd_sim_var_num.py

bsxlb6 commented 5 months ago

Thank you very much for your answer

bsxlb6 commented 4 months ago

Hi， I have an idea. I would like to gather pedestrians who are closer and have the same direction of movement, treating them as a group. So the robot only needs to avoid this group. May I ask if this idea is feasible?If I want to modify the obstacle avoidance code of the robot to avoid crowds, where do I need to modify it?

Shuijing725 commented 4 months ago

Yes, you can access the human velocities by envs.humans[i].vx, envs.humans[i].vy. Is your policy based on neural network? If no, you just need to add a new class in https://github.com/Shuijing725/CrowdNav_Prediction_AttnGraph/tree/main/crowd_nav/policy. If yes, in addition, you also need to add the new network in https://github.com/Shuijing725/CrowdNav_Prediction_AttnGraph/tree/main/rl/networks.