引用本文: 王阳, 孟庆浩, 李腾, 曾明. 室内通风环境下基于模拟退火算法的单机器人气味源定位[J]. 机器人, 2013, 35(3): 283-291.
WANG Yang, MENG Qinghao, LI Teng, ZENG Ming. Single-Robot Odor Source Localization in a Ventilated Indoor Environment Using Simulated Annealing Algorithm[J]. ROBOT, 2013, 35(3): 283-291.
 Citation: WANG Yang, MENG Qinghao, LI Teng, ZENG Ming. Single-Robot Odor Source Localization in a Ventilated Indoor Environment Using Simulated Annealing Algorithm[J]. ROBOT, 2013, 35(3): 283-291.

## Single-Robot Odor Source Localization in a Ventilated Indoor Environment Using Simulated Annealing Algorithm

• 摘要: 针对室内通风环境下的气味源定位问题，提出了一种基于模拟退火策略的单机器人气味源定位算法．受流场的控制，室内气味源释放的烟羽除了具有蜿蜒和间歇特性外， 还会在涡流区域形成局部浓度极大值．本文将气味源定位看作是一种动态函数寻优问题，使用模拟退火策略求取浓度分布函数的最优解，即气味源所在位置． 算法不依赖风信息，从而可以减少流场波动造成的影响．同时通过研究气味浓度与气味源距离之间的关系，提出了一种与气味源距离呈近似线性关系的模拟退火目标函数． 真实室内通风环境下的实验表明，使用本文提出的算法，机器人能够在8m×6m区域内跟踪烟羽并定位气味源，平均定位时间约为10min，且在搜索过程中可以有效地跳出局部极大值．

Abstract: Aiming at the odor source localization in a ventilated indoor environment, a simulated annealing based odor source localization algorithm for a single robot is proposed. Dominated by the flow field, the plume released from an odor source in the room is meandering and intermittent. Moreover, the local concentration maxima may appear in the eddy areas. In this paper, the odor source localization is regarded as a dynamic function optimization problem. The simulated annealing strategy is used to obtain the optimal solution of the concentration distribution function, namely the location of the odor source. The algorithm doesn't need the wind information, so the influence of the flow-field fluctuation is reduced. The relationship between odor concentration and odor-source distance is also studied, and an objective function for simulated annealing that is approximately linear with the odor source distance is presented. The experiments in an actual ventilated indoor environment show that the robot using the proposed method can trace the dynamic odor plume and eventually declare the odor source in an 8m×6m region, and the average localization time is about 10 min. In the search process, the robot is able to jump out of the local maxima effectively.

/

• 分享
• 用微信扫码二维码

分享至好友和朋友圈