Robot behavior pre-judgment system and method based on multi-layer neural network

Abstract

The invention discloses a robot behavior pre-judgment system and method based on a multi-layer neural network. The system comprises a perception preprocessing module, an SOAR logic planning module, anintelligent data fusion module and a multi-layer neural network decision module. The perception preprocessing module is used for acquiring feature information of each object; the SOAR logic planningmodule is used for determining a grabbing probability corresponding to each object according to characteristics related to grabbing in the characteristic information to obtain a logic sequence; the intelligent data fusion module is used for extracting feature information related to a task target in the feature information and fusing the extracted feature information with the logic sequence; and the multi-layer neural network decision module is used for performing robot behavior pre-judgment according to the fused data. According to the invention, the intelligent robot can learn and simulate the cognitive psychology of human processing information so as to have perfect logical reasoning ability, and multi-step behavior pre-judgment or operation control is realized.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a system and method for predicting robot behavior based on a multi-layer neural network. Background technique [0002] Nowadays, robots are often used in multi-object and complex and changeable scenes. When they complete the task of grabbing a specified object, existing robots often perform scene perception and behavior control and output based on image recognition and judgment. , but this process does not really fit the process of human beings to complete this type of task operation, that is, the existing agents or robots lack the inference cognition and logic design of the ontology to complete the behavior and action, the objects to be grasped and the task status Incomplete prediction and other issues. At present, the development of society has higher and higher requirements for intelligent robots. Therefore, the task scenarios and operating states faced by ro...

AI Technical Summary

Problems solved by technology

However, as a connectionist model, ANN has obvious shortcomings in the realization of long-term planning, logical reasoning, multi-step decision-making, etc., and when the internal structure of the network is slightly improved or changed, it will usually cause data costs. index increase

Images