20 / 2025-03-29 08:46:00

Research on the SLAM Method for Piggery Mobile Robots Based on Multi-Sensor Fusion

multi-sensor fusion; SLAM; feature extraction; loop-closure detection; navigation

3. 智慧牧场与动物福利 > （3）智慧养殖的传感器、智能感知和机器人

The piggery mobile robots can navigate flexibly through complex piggery layouts and perform various tasks, provided that it possesses high-precision positioning and environmental perception capabilities. The SLAM method based on multi-sensor fusion has emerged as an effective solution. It integrates the strengths of multi-sensor data, overcomes the degradation issues commonly encountered in single-sensor SLAM methods, and prevents a decline in positioning accuracy and map quality, thus exhibiting greater adaptability and robustness. This article proposes a multi-sensor fusion SLAM method that enhances the LVI-SAM framework. Firstly, the SuperPoint algorithm which is based on deep learning, is employed to extract feature points from visual-inertial system (VIS) data, replacing the traditional Shi-Tomasi algorithm, which enhances the detection capability of feature points in complex scenarios such as low texture and varying lighting conditions. Secondly, to enhance the performance of loop-closure detection in complex scenarios, the Scan Context (SC) algorithm is employed to optimize the loop-closure detection of the Lidar-inertial System (LIS), thereby further enhancing the robot's mapping effect. Finally, the results of the piggery scenario experiments show that, compared to the original LVI-SAM algorithm, the multi-sensor fusion SLAM method proposed in this study significantly reduces the degradation phenomenon of the system. The error of this method at the starting and ending points of the trajectory is less than that of LVI-SAM, as well. Therefore, This method proposed in this paper can achieve high-precision and high-robustness localization and mapping in complex piggery scenarios, and provide an important guarantee for realizing autonomous navigation and obstacle avoidance.