Method of designing household sweeping robot based on deep reinforcement learning

Abstract

A method of designing a household sweeping robot based on deep reinforcement learning comprises the following steps: (1) collecting an original image of the horizontal section of a room around a sweeping robot through a laser radar, and transmitting original data back to a computer via a wireless serial port; (2) carrying out SLAM (Simultaneous Localization and Mapping) on the original data, and after real-time localization of the sweeping robot and room mapping, processing the data and generating a 168*168 two-dimensional array; (3) building a CNN+LSTM neural network; and (4) training the sweeping robot, and giving corresponding return to the neural network according to whether the motion of the sweeping robot meets the expected requirement, in order to change the parameters of the neural network, wherein the sweeping robot can make independent decisions after training. Through the steps, the sweeping robot can find a target object in a strange scene in a relatively short period of time, and can avoid obstacles and plan a path automatically. A sweeping robot designed using the method is of certain versatility and task mobility.

Description

technical field [0001] The invention provides a design method of a household sweeping robot based on deep reinforcement learning, which belongs to the field of smart home. Background technique [0002] At present, more and more families use sweeping robots. When the sweeping robot on the market now enters a new environment to work, it must first traverse the entire room according to the algorithm to complete the mapping and positioning. Then build a semantic map, and finally do path planning, and then start cleaning the room. Among them, the method of traversing the room is often that the sweeping robot keeps walking against a wall, forming a closed loop, and then gradually fills in the blank space in the middle. This method takes a long distance and takes a lot of time. Likewise, path planning algorithms are not only complex and require a lot of computation and human engineering, but they are also not perfect enough to make a robot vacuum work on an optimal path. [0003...

AI Technical Summary

Problems solved by technology

Among them, the method of traversing the room is often that the sweeping robot keeps walking against a wall, forming a closed loop, and then gradually fills the blank space in the middle. This method takes a long distance and takes a lot of time

Likewise, the path planning algorithm is not only complex, requiring a lot of calculation and human engineering, but also not perfect enough to make the sweeping robot work on the optimal path

Images