Font Size: a A A

Energy-efficient Intelligentsensing Strategy

Posted on:2017-03-11Degree:MasterType:Thesis
Country:ChinaCandidate:X C LiFull Text:PDF
GTID:2308330503951207Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
The function of mobile devices based on user state detection is more and more powerful, for example, mobile phones can achieve the tracking and positioning of the user through the GPS. Meanwhile, mobile devices are integrated with a variety of wireless networks and high quality sensor with the development of mobile Internet. Such as gravity sensor, acceleration sensor, gyroscope, Bluetooth, GPS, Wi Fi, etc.. The energy consumption of these technologies is very large. However, the development of battery technology has been slow and has entered the bottleneck now, the limited capacity of the battery has been unable to meet the needs of mobile equipment and energy consumption. However, continuously capturing this contextual information consumes significant amount of energy in the application based on user state recognition. However, the user behavior has a certain regularity. It is significant to reduce energy consumption through adjusting the sampling frequency of sensors.The sampling frequency of the sensor is fixed. In the application of user state detection, define user’s behavior as the state in the application of user state detection. The energy consumption is large when the sampling frequency is too large, the state estimation error is too large when the sampling frequency is sparse.This dissertation propose a computationally efficient algorithm to obtain the optimal sensor sampling policy under the assumption that the user state transition is Markovian. This Markovoptimal policy minimizes user state estimation error while satisfying a given energy consumption budget, the user state is modeled as a N-state discrete time Markov chain. To obtain the optimal sensing policy for Markovian user state, by formulating the constrained optimization problem as a infinite-horizon CMDP and solving its corresponding LP. Since the relationship between the states can be measured by entropy, the entropy strategy is obtained by re defining the optimization objective as H(x). We compare the Markov-optimal policy with uniform periodic sensing and then we compare the Markov-optimal policy with Entropy policy. Considering the uncertainty of the real state in the sampling point, a partial observability strategy is proposed.Monitoring the energy consumption of sensors with power monitor. For Triaxial accelerometer, we program language of Android, collect energy consumption data about different frequencies under different user behaviors, analysis of data distribution characteristics and verify the feasibility of the optimal sampling strategy.
Keywords/Search Tags:energy-efficient, Markov-optimal policy, sampling frequency, Android
PDF Full Text Request
Related items