{
"role": "system",
"content": "You are a helpful assistant to help me determine how to select objects for robotics manipulation experiments."
},
{
"role": "user",
"content":
"I will first give you a list of objects [OBJECT LIST] as a database. Your task is to return the name of object that matches my query. If there were no object matches my query, you need to return UNKNOWN_OBJECT. You also need to output the initial position (e.g., (2, 3)) of the object in my command if applicable. If I also specific the object's orientation, you also need to output it. The orientation should be either 'standing', 'horizontal', or a quaternion such as [0.5, 0.5, -0.5, 0.5]; otherwise ignore it. Please format your response in the following format:
INPUT: Insert a coke can to (2, 3)
OUTPUT: {"object": "coke_can", "p": "(2, 3)"}
INPUT: ADD a nestea can
OUTPUT: {"object": "UNKNOWN_OBJECT"}
INPUT: Add a pepsi can to (4,5), and the orientation is standing
OUTPUT: {"object": "pepsi_can", "p": "(4, 5)", "q": "standing"}
INPUT: Add a pepsi can to (4,5), and the orientation is [0,5, 0.6, -0.4, 0.5]
OUTPUT: {"object": "pepsi_can", "p": "(4, 5)", "q": "[0.5, 0.6, -0.4, 0.5]"}
Your output should only start with OUTPUT "
}
{
"role": "system",
"content": "You are a helpful assistant to help me adjust the lighting condition in the simulation environment."
},
{
"role": "user",
"content":
"I will give you a command about adjusting the light condition. Your task is to interpret the command. For example, if I say make the scenario brighter, you need to return BRIGHT. If I say make the scenario darker, you need to return DARK. If I say use the default lighting, you need to return DEFAULT. Please format your response in the following format:
INPUT: I need to make the scenario brighter.
OUTPUT: BRIGHT
INPUT: I want to use the default lighting.
OUTPUT: DEFAULT
Your output should only start with OUTPUT"
}
{
"role": "system",
"content": "You are a helpful assistant to help me adjust a robotic arm in the simulation environment."
},
{
"role": "user",
"content":
"I will give you a command about adjusting the robot. Your task is to interpret the command. First, you need to determine it is to adjust the camera. Then, you need to determine how to adjust if by returning the position, and orientation. Finally, you also need to determine if there is any update to the camera intrinsic. For the orientation, you need to return the quaternion (make sure the order is (x,y,z,w)). Please format your response in the following format:
INPUT: Move the camera to (2, 3, 4).
OUTPUT: {"p": "[2, 3, 4]"}
INPUT: Move the camera to (2, 3, 4) and rotate it alongside z axis 60 degrees clockwise.
OUTPUT: {"p": "[2, 3, 4]", "q": "[0, 0, 0.5, 0.8660254]"}
INPUT: Move the camera to (0, 0, 0) and rotate it along size z-axis 60 degrees counter-clockwise.
OUTPUT: {"p": "[0, 0, 0]", "q": "[0, 0, -0.5, 0.8660254]"}
INPUT: Rotate the camera 20 degrees counter-clockwise alongside x-axis.
OUTPUT: {"q": "[-0.1736482, 0, 0, 0.9848078]"}
INPUT: Change the camera to 1920x1080, with a fov of 120.
OUTPUT: {"width": 1920, "height": 1080, "fov": 120}
INPUT: Change the camera to 512*256.
OUTPUT: {"width": 512, "height": 256}
Your output should only start with OUTPUT"
}
{
"role": "system",
"content": "You are a helpful assistant to help me generate language mutations."
},
{
"role": "user",
"content":
"I will first give you a sentence, which is formatted as "Pick up [OBJECT]", [OBJECT] can vary. Your task is to transfer this sentence into exactly {num_mutations} different ones that convey same meanings. These sentences can be in imperative sentences. For instance, "Pick up coke" can be transferred to "grasp up coke", or "select coke", or "can you pick up coke", or "let's pick up coke", etc. Your return should only include the exactly {num_mutations} sentences seperated by the linebreaks. Do not add extra the, a, etc. Do not number them. Now transfer this sentence:
[SENTENCE]
"