app.py

import gradio as gr
from ai2thor.controller import Controller
from ai2thor.platform import CloudRendering
# 初始化全局模拟器对象，启用 headless 模式
controller = Controller(platform=CloudRendering)

# 用于记录所有动作的列表
action_history = []

def initialize_simulation():
    """
    初始化 ai2thor 模拟器，并返回初始场景的截图。
    """
    controller.reset('FloorPlan28')  # 指定场景
    event = controller.step(action='Pass')  # 不执行任何动作，只获取初始图像
    image = event.frame
    action_history.clear()  # 清空历史记录
    action_history.append((image, "初始化"))
    return image, "初始化", action_history

def move_forward():
    """
    在 ai2thor 模拟器中执行"向前移动"操作，并返回当前环境的截图。
    """
    action = "向前移动"
    event = controller.step(action='MoveAhead')
    image = event.frame
    action_history.append((image, action))
    return image, action, action_history

def move_backward():
    """
    在 ai2thor 模拟器中执行"向后移动"操作，并返回当前环境的截图。
    """
    action = "向后移动"
    event = controller.step(action='MoveBack')
    image = event.frame
    action_history.append((image, action))
    return image, action, action_history

def move_left():
    """
    在 ai2thor 模拟器中执行"向左旋转"操作，并返回当前环境的截图。
    """
    action = "向左旋转"
    event = controller.step(action='RotateLeft')
    image = event.frame
    action_history.append((image, action))
    return image, action, action_history

def move_right():
    """
    在 ai2thor 模拟器中执行"向右旋转"操作，并返回当前环境的截图。
    """
    action = "向右旋转"
    event = controller.step(action='RotateRight')
    image = event.frame
    action_history.append((image, action))
    return image, action, action_history

# 使用 Gradio 构建交互界面
with gr.Blocks() as iface:
    gr.Markdown("## 具身模型演示")
    init_button = gr.Button("初始化")
    forward_button = gr.Button("向前移动")
    backward_button = gr.Button("向后移动")
    left_button = gr.Button("向左旋转")
    right_button = gr.Button("向右旋转")
    image_output = gr.Image()
    current_action_output = gr.Textbox(label="当前动作")
    action_history_output = gr.Gallery(label="动作历史")

    init_button.click(fn=initialize_simulation, outputs=[image_output, current_action_output, action_history_output])
    forward_button.click(fn=move_forward, outputs=[image_output, current_action_output, action_history_output])
    backward_button.click(fn=move_backward, outputs=[image_output, current_action_output, action_history_output])
    left_button.click(fn=move_left, outputs=[image_output, current_action_output, action_history_output])
    right_button.click(fn=move_right, outputs=[image_output, current_action_output, action_history_output])

iface.launch()