Humanoid robot falling prediction method

Abstract

The invention relates to a humanoid robot falling prediction method. The method comprises the steps that within a set prediction interval time, according to the number of sampling times, a real falling result, inertial sensor collection data and mass center position data corresponding to a humanoid robot are sequentially obtained, and zero moment points ZMP of the humanoid robot in the x directionand the y direction are obtained through calculation; then, data preprocessing is carried out, and a robot motion feature matrix containing eight-dimensional data is obtained; the above steps are repeated, and real falling results and robot motion feature matrixes within multiple prediction interval times are obtained; the multiple robot motion feature matrixes are sequentially used as input of aneural network model, the real falling results are combined, the neural network model is trained, and a humanoid robot falling prediction model is obtained; and the falling prediction model is used for carrying out falling prediction. Compared with the prior art, whether falling happens or not and the falling direction can be accurately predicted, and the method can be well suitable for falling prediction of different robots.

Description

technical field [0001] The invention relates to the technical field of robot fall detection, in particular to a fall prediction method of a humanoid robot. Background technique [0002] Humanoid robots are flexible in movement and can be widely applied to various grounds and scenes; and are similar to humans in shape and structure, and are easy to generate a sense of intimacy when interacting with humans, easily adapt to the existing environment of humans, and replace humans to complete many mechanized or dangerous situations. Humanoid robots have great application potential and research value. The basis for humanoid robots to complete various tasks is that their feet need to ensure stable walking and avoid robot falls. Therefore, it is necessary to predict the fall of humanoid robots in order to make corresponding decisions and actions based on the prediction results, such as making Emergency gait or slow action, so as to effectively prevent the robot from falling. [000...

AI Technical Summary

Problems solved by technology

However, this type of method is not robust to the distortion and error of sensor data, and it is easy to predict errors; and, generally, it can only predict whether the robot will fall, and the error rate of predicting the direction of the fall is high

Images