Janet-xujing-1216 / SafeCrowdNav

基于深度强化学习的机器人视觉导航安全性评估
MIT License
8 stars 0 forks source link

problem #1

Open hubeiluotian opened 8 months ago

hubeiluotian commented 8 months ago

excuse me, where is the code of aemcarl? thank you very much!

Janet-xujing-1216 commented 8 months ago

It is based on https://github.com/SJWang2015/AEMCARL

hubeiluotian commented 7 months ago

yes,I know, but it seems that the author didn't provide these three paramers,do you know about them?thank you! image

hubeiluotian commented 7 months ago

大佬,老实说真的很需要这个aemcarl代码,原作者那里的总感觉哪里的参数有问题导致我跑出来的效果不好,想问下能不能把你跑的amecarl包括参数设置啥的发我一份,邮箱是1654638902@qq.com,感激不尽!!!大佬好人!!

Janet-xujing-1216 commented 7 months ago

很抱歉,因为我之前用的服务器出现故障损坏了,里面所有的东西都没有了。AEMCARL的权重我并没有保存在本地,但我看之前的记录AEMCARL大概需要在GeForceRTX3090上训练三天左右,论文里的对比结果是在第 5400episode 的训练权重得到的。我记得之前并没有修改AEMCARL原文中的参数设置,直接训练就能得到。我把我另外一个collision-avoidance仓库设置为public了,不知道那里面的代码会不会对你有帮助。

hubeiluotian commented 7 months ago

好的好的,感谢大佬!!!

------------------ 原始邮件 ------------------ 发件人: "Janet-xujing-1216/SafeCrowdNav" @.>; 发送时间: 2024年3月2日(星期六) 下午3:10 @.>; @.**@.>; 主题: Re: [Janet-xujing-1216/SafeCrowdNav] problem (Issue #1)

很抱歉,因为我之前用的服务器出现故障损坏了,里面所有的东西都没有了。AEMCARL的权重我并没有保存在本地,但我看之前的记录AEMCARL大概需要在GeForceRTX3090上训练三天左右,论文里的对比结果是在第 5400episode 的训练权重得到的。我记得之前并没有修改AEMCARL原文中的参数设置,直接训练就能得到。我把我另外一个collision-avoidance仓库设置为public了,不知道那里面的代码会不会对你有帮助。

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

hubeiluotian commented 4 days ago

大佬不好意思,我又来了,想请问障碍物碰撞概率 和速度变化的评估是在代码里的哪个位置呢?