Research on path navigation algorithm of garden robot based on Beidou and vision integration

Abstract

In recent years, garden robots have gained increasing attention for their potential to automate tasks such as lawn maintenance, watering, and monitoring. However, a significant challenge in the development of these robots is achieving precise and reliable autonomous navigation in complex garden environments. This paper addresses this problem by proposing a combined navigation solution that integrates the Beidou satellite positioning system with a machine vision navigation system. The proposed approach aims to enhance the accuracy and robustness of garden robot navigation. The vision system employs an improved grayscale factor and vertical grayscale projection technique to filter out noise interference, leveraging differences in noise characteristics and road surface occupancy. Trapezoidal four-point coordinates are then used to extract the navigation path. For satellite-based positioning, the ultra-core HI600R module is utilized in conjunction with the HI600D reference station to construct a minimal RTK (Real-Time Kinematic) system. The raw data from both systems is processed using the Kalman filter algorithm, and the Gaussian-Krüger projection is applied to transform the coordinate system. The final navigation path is generated through a fusion algorithm that combines data from both the Beidou and vision systems. Experimental results demonstrate that the proposed navigation system achieves high accuracy and excellent stability, with an average positioning error of less than 0.2 meters in static conditions and less than 0.5 meters in dynamic conditions. These findings highlight the potential of the Beidou-vision fusion approach for practical applications in garden robot navigation.