Developed a robot that can look to the future

Developed A Robot That Can Look To The Future

Researchers have developed a robot that can accurately predict what you will do based on your movements and then provide timely help. it works!

You may be thinking: Why must the robot foresee what I am going to do? But it is very convenient. For example: You have just been shopping and you are in a hurry and need to put the product in the refrigerator. Then, it is very convenient if the robot opens the refrigerator door for you. However, such a robot must know that you intend to put items in the refrigerator.

Database camera
But how do we make robots understand what we want, and then take action to achieve our wishes?Scientist Cornell University Some suggestions are made for this. They equipped the robot with a 3D camera and provided the robot with a 3D video database. With the help of these tools, the robot can recognize the human actions it sees and then try to predict the next step. An example: When a robot sees a person take an object from a table, it will recognize it. He found that it was a pack of butter. The robot can then choose from a series of follow-up actions taken by humans with the pack of butter. For example: eating, drinking, cleaning, tidying, etc. The robot chooses the most obvious option. At the same time, people keep moving forward. For example, after he picked up a pack of butter, he got up. The robot adjusts its predictions based on all these movements. For example, if people start to walk with butter packets, then people are less likely to eat butter packets. He is more likely to clean up the bag of butter. The robot knows where the butter box is and can then open the refrigerator door for human use.

Pieces
What the robot actually does is a simple action-for example: cleaning a box of butter-cutting it into small pieces. The robot analyzes each part and can make increasingly accurate predictions based on multiple parts. Experiments show that the robot can predict human behavior within one second of 82% of the time. If the robot must observe in the next two to three seconds, then 71% of the time is correct. Looking into the next ten seconds will be more difficult, but the robot is still correct 57% of the time.

Although the robot has performed well, the researchers do have greater ambitions. Researcher Ashutosh Saxena (Ashutosh Saxena) said: “Although humans are foreseeable, but at some point only foreseeable.” Researchers are currently “teaching” robots every response to human actions. In the future, researchers hope that this method will no longer be needed, but that robots can learn appropriate responses to human behavior through accumulated experience.

Researchers have developed a robot that can accurately predict what you will do based on your actions and then provide timely help. it works!

You may be thinking: Why must the robot foresee what I am going to do? But it is very convenient. For example: You have just been shopping and you are in a hurry and need to put the product in the refrigerator. Then, it is very convenient if the robot opens the refrigerator door for you. However, such a robot must know that you intend to put items in the refrigerator.

Database camera
But how do we make robots understand what we want, and then take action to achieve our wishes?Scientist Cornell University Some suggestions are made for this. They equipped the robot with a 3D camera and provided the robot with a 3D video database. With the help of these tools, the robot can recognize the human actions it sees and then try to predict the next step. An example: When a robot sees a person take an object from a table, it will recognize it. He found that it was a pack of butter. The robot can then choose from a series of follow-up actions taken by humans with the pack of butter. For example: eating, drinking, cleaning, tidying, etc. The robot chooses the most obvious option. At the same time, people keep moving forward. For example, after he picked up a pack of butter, he got up. The robot adjusts its predictions based on all these movements. For example, if people start walking with butter packets, then people are less likely to eat butter packets. Instead, he is more likely to clean up a pack of butter. The robot knows where the butter box is and can then open the refrigerator door for human use.

Pieces
What the robot actually does is a simple action-for example: cleaning a box of butter-and cutting it into small pieces. The robot analyzes each part and can make increasingly accurate predictions based on multiple parts. Experiments show that the robot can predict human behavior within one second of 82% of the time. If the robot has to observe in the next two to three seconds, then 71% of the time is correct. Looking into the next ten seconds will be more difficult, but the robot is still correct 57% of the time.

Although the robot has performed well, the researchers do have greater ambitions. Researcher Ashutosh Saxena (Ashutosh Saxena) said: “Although humans are predictable, but at some point only predictable. Researchers are currently “teaching” robots every response to human behavior. In the future, Researchers hope that this method is no longer needed, but that robots can learn appropriate responses to human behavior by accumulating experience.