基于Nelder-Mead算法的机器人主动嗅觉室内时变污染源定位

Locating Indoor Time-Variant Contaminant Sources Based on Nelder-Mead Algorithm Using Robot Active Olfaction Method

摘要:

将Nelder-Mead（NM）算法与机器人主动嗅觉相结合，对室内衰减型和周期型两种时变污染源开展定位研究。首先通过计算流体动力学（CFD）模拟得到这两种时变污染源的浓度场，然后利用NM算法对其定位。结果表明，两种时变污染源的定位成功率均在80 %以上。对机器人数量、响应时间和最大搜索步数这三个影响因素进行讨论，通过分析发现当室内面积为100 m²左右时，机器人数量为5个、响应时间为2 s、最大搜索步数为50步定位效果最好。

Abstract:

In this paper， the Nelder-Mead （NM） algorithm was combined with the robot active olfaction method to locate two types of time-variant contaminant sources， which are attenuated and periodic sources， respectively. First， the concentration fields of the contaminant sources were simulated through computational fluid dynamics （CFD） simulation. Then， the time-variant contaminant sources were located using the NM algorithm. The results show that the success rates of locating the two time-variant contaminant sources are both above 80 %. Three influencing factors such as the number of robots， robot response time， and the maximum number of robot search steps were discussed. The analysis indicates that in the case of 100 m² indoor space， time-variant sources can be best located when 5 robots are working on the site with a setting of 2 seconds in robot response and a maximum number of 50 robot search steps.