Exploring tasks in unknown environments has become a relevant search and rescue robotics approach. Ground robots are a better alternative to rescuers for first exploration. However, exploration progress is often limited by uneven terrains that exceed the kinematic capabilities of robots, including those with complex locomotion systems. This work proposes an innovative solution based on collaborative behaviours to overcome even terrains. A method employing two collaborative robots designed to operate in a marsupial configuration to surmount uneven terrains has been implemented. These robots, denoted as R1 (enhanced with a mobile ramp) and R2 (serving as an explorer), interact synergistically to expand the explored area autonomously. A state machine has been implemented to manage the progression of the mission, based on a perception (RGB-D) system, for both decision-making and autonomous execution of the process. In the initial stage, the terrain and ascent zones to be explored are characterized using point clouds and unsupervised learning. Subsequently, the second stage manages the interaction between the robots by controlling the R2 ascent through the R1 ramp using artificial vision algorithms and beacons. Outdoor tests have been performed to validate the method. The main results show an effectiveness of 95% in automatically identifying access zones.

中文翻译：

---

用于在不平坦地形中进行探索的基于视觉的协作机器人

在未知环境中探索任务已成为一种相关的搜索和救援机器人方法。地面机器人是救援人员进行首次探索的更好选择。然而，勘探进展往往受到不平坦地形的限制，这些地形超出了机器人的运动能力，包括那些具有复杂运动系统的机器人。这项工作提出了一种基于协作行为的创新解决方案，以克服平坦的地形。已经实施了一种采用两个协作机器人的方法，该机器人设计为以有袋动物的配置运行以克服不平坦的地形。这些机器人被标记为 R1（通过移动坡道增强）和 R2（充当探索者），协同互动以自主扩展探索区域。基于感知 (RGB-D) 系统，已实施状态机来管理任务的进展，用于决策和自主执行过程。在初始阶段，使用点云和无监督学习来表征要探索的地形和上升区域。随后，第二阶段使用人工视觉算法和信标控制 R2 通过 R1 坡道上升，从而管理机器人之间的交互。已经进行了室外测试来验证该方法。主要结果显示，自动识别访问区域的有效性为 95%。