Department of Systems and Control Engineering, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8552, Japan.
Sensors (Basel). 2023 Jan 19;23(3):1140. doi: 10.3390/s23031140.
Odor source localization (OSL) robots are essential for safety and rescue teams to overcome the problem of human exposure to hazardous chemical plumes. However, owing to the complicated geometry of environments, it is almost impossible to construct the dispersion model of the odor plume in practical situations to be used for probabilistic odor source search algorithms. Additionally, as time is crucial in OSL tasks, dynamically modifying the robot's balance of emphasis between exploration and exploitation is desired. In this study, we addressed both the aforementioned problems by simplifying the environment with an obstacle region into multiple sub-environments with different resolutions. Subsequently, a framework was introduced to switch between the Infotaxis and Dijkstra algorithms to navigate the agent and enable it to reach the source swiftly. One algorithm was used to guide the agent in searching for clues about the source location, whereas the other facilitated the active movement of the agent between sub-environments. The proposed algorithm exhibited improvements in terms of success rate and search time. Furthermore, the implementation of the proposed framework on an autonomous mobile robot verified its effectiveness. Improvements were observed in our experiments with a robot when the success rate increased 3.5 times and the average moving steps of the robot were reduced by nearly 35%.
气味源定位（OSL）机器人对于安全和救援团队克服人类暴露于危险化学烟雾的问题至关重要。然而，由于环境的复杂几何形状，几乎不可能在实际情况下构建气味羽流的扩散模型，以便用于概率气味源搜索算法。此外，由于在 OSL 任务中时间至关重要，因此希望动态地修改机器人在探索和开发之间的平衡。在这项研究中，我们通过将障碍物区域简化为具有不同分辨率的多个子环境来解决上述两个问题。随后，引入了一个框架，在 Infotaxis 和 Dijkstra 算法之间切换，以引导代理并使其迅速到达源。一个算法用于引导代理搜索有关源位置的线索，而另一个算法则促进代理在子环境之间的主动移动。所提出的算法在成功率和搜索时间方面都有所改进。此外，在自主移动机器人上实现了所提出的框架，验证了其有效性。当机器人的成功率提高了 3.5 倍且机器人的平均移动步数减少了近 35%时，我们在机器人的实验中观察到了改进。
