| 申传艳,牛晶,高光浩,郑佳豪,张利鹏,刘世锋.基于空间多尺度连续性特征提取的轮式机器人经验地图构建[J].电子测量与仪器学报,2025,39(9):87-98 |
| 基于空间多尺度连续性特征提取的轮式机器人经验地图构建 |
| Experience map construction for wheeled robots based onspatial multi-scale continuity feature extraction |
| |
| DOI： |
| 中文关键词: 轮式机器人 经验地图 空间多尺度连续性特征 PSA ASPP 转向角度预测模型 |
| 英文关键词:wheeled robots experience map spatial multi-scale continuity features PSA ASPP steering angle prediction model |
| 基金项目:甘肃省重点研发计划(23YFFE0001)、甘肃省高校教师创新基金(2023A-1143)、天水师范大学产业支撑引导项目(CYZ2023-05)资助 |
|
|
| Author | Institution |
| Shen Chuanyan | School of Mechanical, Electrical and Automotive Engineering, Tianshui Normal University, Tianshui 741001, China |
| Niu Jing | School of Mechanical, Electrical and Automotive Engineering, Tianshui Normal University, Tianshui 741001, China |
| Gao Guanghao | School of Mechanical, Electrical and Automotive Engineering, Tianshui Normal University, Tianshui 741001, China |
| Zheng Jiahao | School of Mechanical, Electrical and Automotive Engineering, Tianshui Normal University, Tianshui 741001, China |
| Zhang Lipeng | School of Vehicle and Energy, Yanshan University, Qinhuangdao 066004, China |
| Liu Shifeng | School of Mechanical, Electrical and Automotive Engineering, Tianshui Normal University, Tianshui 741001, China |
|
| 摘要点击次数: 576 |
| 全文下载次数: 493 |
| 中文摘要: |
| 针对传统同步定位与地图构建（SLAM）算法建图过程复杂、调参难度大和泛化能力差等问题，提出了一种基于空间多尺度连续性特征提取的轮式机器人经验地图构建方法。首先，在ResNet18网络模型中添加金字塔切分注意力模块（PSA）和空洞空间卷积池化金字塔（ASPP）模块，PSA对中间层的特征进行分组处理，通过计算不同通道的注意力权重捕获多尺度信息，提高特征的表达能力，ASPP利用不同扩张率的空洞卷积和全局平均池化整合全局上下文信息，进一步强化空间多尺度连续性特征的表征；其次，利用改进的ResNet-PSA-ASPP网络，在donkey_sim仿真模拟器和机器人实际运行跑道上对采集的数据集进行深度学习训练，获取优化后的机器人转向角度预测模型；最后，利用donkey_sim仿真模拟器和机器人操作系统（ROS）分别在仿真环境和实际场景下进行模型性能测试实验。实验结果表明，提出的模型对转向角度预测的误差分别减少了38.47%、44.34%、35.51%，相比经典的ResNet18、ResNet50、VGGNet等网络模型在特征提取能力、计算效率和建图准确度上均获得显著提升。 |
| 英文摘要: |
| To address the issues of complex mapping processes, difficult parameter tuning, and poor generalization in traditional simultaneous localization and mapping (SLAM) algorithms, this paper proposes a wheeled robot experience map construction method based on spatial multi-scale continuity feature extraction. First, PSA and ASPP modules are integrated into the ResNet18 architecture. PSA groups intermediate features and calculates attention weights across channels to capture multi-scale information, thereby enhancing feature representation. ASPP incorporates dilated convolutions with varying dilation rates and global average pooling to aggregate global contextual information, further strengthening the representation of spatial multi-scale continuity features. Then, the improved ResNet-PSA-ASPP model is trained on datasets collected in both the donkey_sim simulator and real-world robot racetrack scenarios. Finally, model performance is evaluated in both simulated and real-world environments using the donkey_sim simulator and the robot operating system （ROS）. Experimental results show that the proposed model reduces steering angle prediction errors by 38.47%, 44.34%, and 35.51%, respectively, and significantly outperforms classical networks such as ResNet18, ResNet50, and VGGNet in feature extraction capability, computational efficiency, and mapping accuracy. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
