移动机器人长期自主环境适应研究进展和展望

针对移动机器人在未知环境中自主探索及建图存在盲目性的问题, 提出了一种基于贝叶斯优化评估多重信息增益的探索策略。在候选点提取方法上采用融合前沿点聚类与可通行区域的方式综合衡量提取, 相较于传统的前沿点检测方法有效解决了候选点集合过大及环境信息缺失等问题; 在候选点评估方法上利用贝叶斯优化计算多重信息增益, 综合考虑地图熵值与距离成本, 相较于仅考虑地图熵值选取最佳候选点的方法, 有效改进了机器人在环境中的冗余路径。该算法在机器人操作系统(robot operating system, ROS)中采用Gazebo进行仿真实验验证, 构建环境地图。结果表明, 该方法可以使移动机器人快速有效地探索未知环境, 高质量完成建图任务。

Research progress and prospect of long-term autonomous environment adaptation for mobile robots

Aiming at solving the problem of blindness in autonomous exploration and mapping by mobile robots in unknown environments, an exploration strategy based on Bayesian optimization to evaluate multiple information gains was proposed. In the candidate point extraction method, the method of integrating frontier point clustering and passable area to comprehensively measure and extract was adopted. Compared with the traditional frontier point detection method, it effectively solved the problems of excessive candidate point sets and missing environmental information. The Bayesian optimization was used to calculate multiple information gains considering both map entropy and distance costs. Compared with the method of selecting the best candidate point based solely on map entropy, this method effectively improved the redundancy path of the robot in the environment. Gazebo was used to verify the algorithm in robot operating system (ROS) and build environment map. Results show that the proposed method can enable the mobile robot to explore the unknown environment quickly and efficiently and complete the mapping task with high quality.