RLbench_for_vit / readme.md
Louis0411's picture
add description
0e024f0

项目说明

本项目中,data 变量是一个 numpy 浮点数组,其形状为 (n, 29),其中:

  • n:观测帧数
  • 29:每一帧的低维观测数据长度

每一行(长度为 29 的 1D 数组)代表一帧观测,其内部结构如下:

序号 长度 字段名 说明
1 1 gripper_open 抓手张开度(浮点数)
2 7 joint_velocities 7 个关节的速度
3 7 joint_positions 7 个关节的位置
4 7 joint_forces 7 个关节受到的力
5 7 gripper_pose 抓手在空间中的位姿(7 维表示方式)

Task description: ['put rubbish in bin', 'drop the rubbish into the bin', 'pick up the rubbish and leave it in the trash can', 'throw away the trash, leaving any other objects alone', 'chuck way any rubbish on the table rubbish']

示例

import numpy as np

# 假设 data.npy 存储了 shape=(100,29) 的观测数据
data = np.load("data.npy")    # data.shape == (100, 29)

# 取第 i 帧数据
i = 0
frame = data[i]               # frame.shape == (29,)

# 拆分各个字段
idx = 0
gripper_open     = frame[idx : idx+1]        # shape (1,)
idx += 1
joint_velocities = frame[idx : idx+7]        # shape (7,)
idx += 7
joint_positions  = frame[idx : idx+7]        # shape (7,)
idx += 7
joint_forces     = frame[idx : idx+7]        # shape (7,)
idx += 7
gripper_pose     = frame[idx : idx+7]        # shape (7,)

print("gripper_open:", gripper_open)
print("joint_velocities:", joint_velocities)
# … 以此类推