A method and system for robot dynamic obstacle avoidance based on visual attention

Abstract

The invention discloses a robot dynamic obstacle avoidance method based on visual attention. The method comprises the following steps: S1, acquiring video information of two paths acquired by two cameras on a robot; S2, processing the video information of two paths to obtain a visual attention map; S3, determining a dynamic object, of which a visual attention value is higher than a dangerous threshold value, in the visual attention map as a dangerous dynamic object; S4, estimating speed and acceleration of the current moment of the dangerous dynamic object, and obtaining predicted dynamic quantity of the dangerous dynamic object through prediction via a b-spline function; S5, performing feature matching for the video information of two paths through an ICP algorithm to obtain robot dynamic quantity of the robot; S6, in combination with the robot dynamic quantity and the predicted dynamic quantity, under a preset constraint condition, solving a cost function to obtain optimal robot control quantity. The invention also discloses a robot dynamic obstacle avoidance system based on visual attention. By method and the system provided by the technical scheme of the invention, a mobile robot is enabled to avoid a dynamic obstacle object more efficiently and accurately.

Description

technical field [0001] The invention belongs to the field of robot visual detection, and in particular relates to a method and system for avoiding dynamic obstacles of a robot based on visual attention. Background technique [0002] The function of avoiding dynamic obstacles in the environment plays an important role in the application of mobile robots, which enables some key applications of robots, such as automatic detection of the robot environment and navigation, to obtain good results. With the development of robot technology, it has broad development prospects in robot applications, and its research and development also occupies an increasingly important position. [0003] At present, known methods for dynamic obstacle avoidance of robots mainly include artificial potential field-based methods (Artificial Potential Field), probability-based obstacle avoidance methods, and control theory-based obstacle avoidance methods. However, the method based on the artificial pote...

AI Technical Summary

Problems solved by technology

However, the method based on the artificial potential field does not consider the dynamics of obstacles, lacks motion estimation for dynamic obstacles, and there are still certain safety hazards when the robot works in a complex environment.

The probability-based obstacle avoidance method takes into account the uncertainty of the dynamic environment, and the obstacle avoidance efficiency of the robot is high, but the disadvantage is that the calculation is complicated, the memory consumption is huge, and a high-performance CPU processor is required.

The obstacle avoidance method based on control theory can consider more constraints, but in the complex environment where the robot works, too many constraints (for example, there are many dynamic obstacles in the environment) lead to an exponential increase in the amount of calculation of the system, which makes Reduced computational efficiency and increased resource consumption

[0004] It can be seen from the above that the existing dynamic obstacle avoidance methods for robots generally have the problems of low efficiency and large resource consumption.

Images