Research On Robot Odor Tracing Based On Extremum Seeking Control

The phenomenon of odor tracking is widespread in nature. Many organisms usesmell to escape invaders, communicate, find the similar, search for food and so on.Inspired by biological odor tracking, from the beginning the1990s, a number ofscholars started to use robot(s) to accomplish the task of tracking odors. Extremumseeking control algorithm is targeted at the real-time optimization method for dynamicsystems like the robot based odor tracking, in which very little systematic informationis known except for knowing the existence of local extrema. In this thesis, disturbancebased extremum seeking controllers are designed for robot(s) to search forconcentration extremum in time-varying odor/gas fields by using only gas sensors.This thesis focuses on robots based odor source localization using extremumseeking control algorithm, and the following research work has been carried out.1. The state-of-the-art odor tracking and extremum seeking algorithm have beenreviewed, and the existing problems in the robot based on odor tracking havebeen explained in detail. Four kinds of extremum seeking algorithms and theassociated characteristics are summarized. The theoretic knowledge for provingconvergence of extremum seeking algorithm, i.e., mean theory and singular valueperturbation, is introduced.2. This thesis elaborated four concentration fields: the second symmetrical analyticconcentration field, the standard Gaussian concentration field, asymmetricstandard Gaussian concentration field, numerical concentration field. Theconcentration extremum seeking is realized by a control strategy which regulatesthe linear velocity of the robot while keeps its angular velocity constant in fourdifferent time-varying concentration fields. The convergence of the proposedalgorithm is proved by using averaging theory in the time-varying symmetricalquadratic concentration field.3. The single robot extremum seeking control algorithm framework has been given.At the same time, MATLAB simulations are carried out in the secondsymmetrical analytic concentration field and asymmetric standard Gaussianconcentration field, which approximately correspond to the underground diffusionmodel and that in the air with relatively constant wind speed and direction, respectively. Simulations on MATLAB platforms show that, in the two analyticfields, the size of the convergent region is proportional to the amplitude of thedisturbance signal and inversely proportional to the disturbance frequency; andthe change of the robot’s initial position does not affect the convergence.4. Odor/gas plume numerical simulation platform and robot-based platform areintroduced. The results in an odor/gas plume numerical simulation platform androbot-based platform in the artificial wind tunnel experiments preliminaryexperimental results show that the extremum search algorithm can be used inodor tracking. The numerical field approximately simulates the odor/gasdispersion in a ventilated indoor environment.5. The advantages and characteristics of multi-robot systems are introduced withrespect to the single-robot system. The multi-robot extremum seeking controlalgorithm framework has been given. At the same time, MATLAB simulationsare carried out in one-dimensional and two-dimensional environments. Throughthe analysis of simulation results, we can obtain that extremum seeking controlalgorithm in multi-robot system is equally applicable, and multi-robot systemcompared to a single robot has excellent speed and efficiency.