Table 1 - uploaded by Ashit Talukder
Content may be subject to copyright.
Source publication
This paper describes a method for learning coordination policies in body sensor networks. The learning of a compact coordination policy is important for implementing the policy in sensor nodes with limited memory. We present a novel algorithm, Reinforcement Learning Average Approximation (RLAA), to learn local coordination policies for each sensor...
Contexts in source publication
Context 1
... 5 shows photographs of the ISF alcohol sensor and the internal circuitry of the microcontroller interface. The energy capacity of the battery used for the sensors and the energy consumption rate of the system components are shown in Table 1. The vacuum pump operates independently of the sensor (for periods of 5 seconds when the pressure falls below a predefined threshold) but shares the battery with the sensor. ...
Context 2
... action level is defined by a period of time of operating the sensor followed by the sensor entering a sleep mode, i.e., the duty cycle of the sensor varies for different action levels. The expected energy consumption rate can then be computed for each action level from the values in Table 1. ...
Similar publications
Small basestations (SBs) equipped with caching units have potential to handle the unprecedented demand growth in heterogeneous networks. Through low-rate, backhaul connections with the backbone, SBs can prefetch popular files during off-peak traffic hours, and service them to the edge at peak periods. To intelligently prefetch, each SB must learn w...
Citations
... However, these do not solve the fundamental scalability problem. In [8], we proposed the RLAA learning algorithm to learn the coordination policies. In this paper, we describe an alternate method to approximate a global policy with a set of local policies which are learned during offline simulations. ...
This paper presents the Average-Max Reinforcement Learning (AMRL) algorithm that can be used to approximate a global policy of a Markov Decision Process (MDP) as a set of local policies that can be executed in a partially observable environment. The local policies are obtained by reinforcement learning and averaging state-action tables under a stochastic process model. This approach overcomes the scalability problem that arises when a large MDP has to be solved exactly. The approach is motivated by the problem of computing coordination policies for correlated but distributed sensors. We demonstrate the performance of this learning scheme on a simulation of a wireless body sensor network. These results show that the performance of the AMRL algorithm is significantly better than a random policy and is close to the optimal policy that can be obtained from solving a global MDP. The results also show that the AMRL algorithm is scalable to networks represented by large state spaces.
... Decrease sampling rates [8][9][10][11] Select the sensor closest to the base station [12][13][14] Control transmission power at the lower bound of successful transmission [15] Data aggregation to reduce the trade volume [16,17,21] Refine circuit [18,19] Tree based or hierarchical architecture to reduce external transmission [14,20,21] servers. The advantage is that sensors only deliver data to the MBS with short-range transmissions. ...
... The Body-Posture-based Dynamic Link Power Control (BDLPC) has been proposed in [15]. The BDLPC sets the transmission power to the lower threshold of transmission power that can successfully transmit data. ...
In order to set up WSN in various rigorous environments, the size and power constraints are stricter due to the high demands for convenience and reliability. Therefore, power efficiency is very important for a WSN. For this, a novel architecture is presented in this paper. The proposed architecture categorizes sensors into different clusters by events. In each cluster, a minimum spanning tree is constructed for intracluster routing. The hierarchical architecture is useful in reducing the power consumption. In each intracluster routing tree, only leaf nodes are responsible for periodical detection. Data transmissions only occur when abnormal events are detected. An abnormality will be reported to the data center only if the majority of cluster members sense the same event. By reducing unnecessary data transmissions and shortening transmission distances, the proposed mechanism significantly reduces the power consumption and prolongs the network lifetime without influencing the accuracy of event response. The simulations show that the proposed architecture has about an 18-fold improvement rate in the device lifetime and avoids the false positive caused by the erroneous alarm of a single sensor. The proposed architecture is feasible, practical, and highly applicable to many applications.
The problem of controlling the operation of the multiple sensors in a monitoring network is one of jointly optimizing for both information requirements and energy constraints. These goals conflict with each other — increasing the sampling rate collects more data but this consumes energy and hence reduces system lifetime. On the other hand, sampling at a higher rate when the environment is not changing does not generate more useful information. This work presents a general data-centric framework for adaptive sensing that brings together both information requirements and energy constraints in an optimization problem for generating sensor control actions. Clustering of correlated sensors is proposed as a solution for this framework where the collected data is periodically used to learn the relationship between sensors and energy parameters. Control is divided into periods of full sensing and periods of approximation. During periods of full sensing, measurements collected from all sensors are used to compute pair-wise correlations which are then used as the distance metric for clustering. A regression function is learned to approximate the measurement of a sensor given its past sensor measurements and the measurements of another sensor in the cluster. The control framework is evaluated using a publicly-available dataset of approximately 2.4 million sensor readings collected from 54 sensors over a period of approximately 35 days. The parameter space of the clustering-based adaptive sensing policy is systematically explored using this dataset. The framework enables a point on the energy-information trade-off curve to be quantified for different choices of policy parameters.
Wireless sensor network nowadays has been applied to a wide range of applications. However, there are still some challenges to apply sensor networks to practical applications, such as size constraint, cost constraint, and power constraint. All those constraints would become stricter in applying to Body Sensor Network (BSN), which motivates scholars to develop more compact sensors with reasonable maintenance fees so as to carry by users or even embed into users' bodies. Under the constraints of size and cost, the battery size is strictly restricted accordingly. Thereupon power consumption becomes more and more essential to make BSN realistic. Therefore, we propose a novel architecture of BSN in this paper, which reduces the regular power consumption of BSN so that the device lifetime and network reliability can be significantly improved. The proposed architecture has following advantages: (1) it reduces redundant data transmissions without influencing the accuracy of the system. (2) By reducing unnecessary data transmissions and optimizing routing path, it simultaneously reduces the power consumption and prolongs the network lifetime. (3) It is feasible and highly applicable to many applications.
The paper describes a method for adaptive transmission of data over error-prone wireless links and suited to the limitations of embedded pervasive sensing systems. The technique performs a multi-resolution wavelet transform on the data followed by interleaving the resulting coefficients among the transmitted packets. The interleaving scheme is designed to facilitate estimation of coefficients lost during transmission by minimizing the correlation between elements in one packet. A correlation model between detail coefficients obtained from a wavelet transform of Gauss-Markov processes is used to estimate the optimal distribution of coefficients to packets. The interleaving ensures that packet loss results in non-contiguous "gaps" in the reconstructed wavelet tree. The original data is reconstructed by polynomial interpolation of the missing coefficients from correlated neighboring coefficient. We present simulation results that show the performance of our scheme on both real-world and simulated datasets.
