RoboEval reports terminal failures as successes and drops simulator truncations

未关闭
#85 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
3/5
预计耗时
1-2 天
新手友好度
45/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
robotics

调研方向

Start with RoboevalWrapper.step() in environments/roboeval/env.py, where is_success comes from termination and truncated is hardcoded to False. Then read the simulator's outcome contract in roboeval/roboeval_env.py and the dropped-pot case in roboeval/envs/lift_pot.py. Done means scoring uses the simulator's task_success and truncations stop the loop, with tests for both cases in the evaluation loop.

由索引模型根据 Issue 内容生成。

描述

RoboevalWrapper.step() sets is_success from termination and always returns truncated=False. In the pinned simulator, dropping the pot terminates the episode with task_success=0. Rho records that drop as a success. An unhealthy simulation can also truncate while Rho keeps stepping.

Use the simulator’s task_success and truncated fields and cover both outcomes through the evaluation loop. Two related wrapper changes can be reviewed separately. Forward the episode reset seed and support configurable RGB cameras for recordings.

Rho wrapper · Simulator outcome contract · Dropped-pot failure

Pull requests

  • #86 uses the simulator’s task-success signal.
  • #87 stops on simulator truncation.
  • #88 forwards the episode reset seed.
  • #89 supports configurable RGB cameras.
  • #90 tests scoring and seeded policy replay across successive rollouts.
主要语言
Python
星标
38
派生
6
平均合并
2 天 9 小时
30 天内合并 PR
20

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南,环境需要你自己搭建:先看它的 README,通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

microsoft/rhobotics 的其他 Issue

查看 microsoft/rhobotics 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。