Method and apparatus for detecting ground attribute of legged robot
AssigneeBeijing Xiaomi Robot Technology Co., Ltd.
InventorHaiwang Xu
Life of this patent
Estimating the term…
Abstract
A method for detecting a ground attribute of a legged robot includes obtaining a collision audio of a foot of the legged robot with a ground; and detecting a workable level attribute of the ground in a working environment of the legged robot according to the collision audio. The sound of the collision between the foot of the robot and the ground is collected, and the workable level attribute of the ground in the working environment of the legged robot is detected based on the sound, so that the operable level attribute can be effectively used to control the legs of the legged robot. On the one hand, the motion noise of the legged robot can be reduced, and on the other hand, the power consumption of the legged robot can be reduced, thereby increasing its range of motion.
Claims16
A method for detecting a ground attribute of a legged robot, comprising:
- obtaining a collision audio caused by acting of a foot of the legged robot on a ground of the legged robot;
- generating a first collision image by sampling the collision audio;
- collecting a plantar force value of the foot;
- generating a second collision image according to the plantar force value and the first collision image;
- and
- generating a workable level attribute corresponding to the ground by inputting the second collision image into a first neural network model, wherein the workable level attribute is indicated by a value determined based on a degree of hardness and rigidity of the ground.
A method for controlling a legged robot, comprising:
- obtaining a collision audio caused by acting of a foot of the legged robot on a ground of the legged robot;
- generating a first collision image by sampling the collision audio;
- collecting a plantar force value of the foot;
- generating a second collision image according to the plantar force value and the first collision image;
- generating a workable level attribute corresponding to the ground by inputting the second collision image into a first neural network model, wherein the workable level attribute is indicated by a value determined based on a degree of hardness and rigidity of the ground;
- and
- controlling the legged robot according to the workable level attribute.
An apparatus for detecting a ground attribute of a legged robot, comprising:
- one or more processors;
- a memory storing instructions executable by the one or more processors;
- wherein the one or more processors are configured to:
- obtain a collision audio caused by acting of a foot of the legged robot on a ground of the legged robot;
- generate a first collision image by sampling the collision audio;
- collect a plantar force value of the foot generate a second collision image according to the plantar force value and the first collision image;
- and
- generate a workable level attribute corresponding to the ground by inputting the second collision image into a first neural network model, wherein the workable level attribute is indicated by a value determined based on a degree of hardness and rigidity of the ground.
Dependencies are inferred from the claim text.