Gas Source Localization Using Mobile Olfaction Sensor Network

The so-called mobile sensor network (MSN) in this dissertation is composed bymobile sensor nodes. In other words, a MSN node has the same feature as awireless-sensor-network (WSN) node except that it can move. Owing to the mobility,the sensing range of the MSN node could be much larger than the WSN node, thususing less MSN nodes could achieve the similar result to that obtained using a greatnumber of fixed WSN nodes. The MSN has broad application prospects in the fieldssuch as toxic/harmful gas leak detection, fire detection and post-disaster rescue.Focusing on MSN and its application to gas source localization, this dissertationmainly addresses the following research work.Firstly, considering the demand of small size and low price for MSN, mobile sensornodes suitable for gas source localization are designed. A flexible motion controlalgorithm for the designed nodes is introduced. The self-recovery network ptotocol ispresented for the MSN, and the gas sensor calibration method which can suppress thezero drift is proposed.Secondly, a sound-based relative localization method for distributed MSN systemsis proposed. Theoretical derivation and experiments prove that, as long as any twomobile nodes send sounds successively, the relative localization of the whole MSNsystem can be realized using the proposed method. In the process of estimating theorientation and distance of the sounding node, the adding-window operation onconvolution and the algebra operation between sampling signals efficiently reduce theamount of calculation and the random noises in the environments, respectively.Thirdly, a gas source localization method based on the MSN with fixed geometrictopology is proposed, where no wind information is needed. The method can beregarded as a gas source localizaiton process through dynamic deploying a limitednumber of mobile sensor nodes. In the time-avergaed constant wind environments,each estimation cycle is divided into two steps. At first, the gas source location isestimated using nonlinear least square method based on the measured gasconcentration. Then, the mobile nodes are controlled to move towards a point, definedas the new center of the MSN, in the direction of the estimated gas-source locationusing saturated algorithm. Besides, a new plume tracing method is proposed basedupon the fixed-topology MSN in the time-varying wind environment. The validity of the above two methods are demonstrated using real hardware experiemnts andcomputer simulations, respectively.At last, a conditional entropy based gas source localization method is proposed,where the MSN with constraint connection is employed. Distributed particle filter isutilized to approximately obtain the conditional entropy and its gradient. The gradientcan give rise to the control law which forces the mobile nodes to move along thenegative gradient of the conditional entropy. Thus, the conditional entropy isdecreased with an increase in the certainty of the estimated gas source location. Theproposed method achieves the gas source localization in the environments withtime-varying wind or obstacles. Simulations on several cases validate the proposedmethod.