Changeset 2781

Show
Ignore:
Timestamp:
10/30/09 14:44:14 (4 weeks ago)
Author:
jaein
Message:
 
Files:
1 modified

Legend:

Unmodified
Added
Removed

  • HydroWatch/Tim/doc/ipsn10/sec_related.tex

    r2774 r2781  
    7272where the duty cycles of node components adapt according to  
    7373the energy available in the battery and through harvesting mechanisms.   
    74 While the energy architecture in Section~\ref{sec:system} supports  
    75 duty cycle adaptation of any node component, this paper has focused  
    76 mainly on adaptation of sensor sampling rates, which has been overlooked  
    77 as a major energy sink by most of the related work. In addition, while  
    78 our energy architecture supports energy harvesting, the main premise of  
    79 our work is that node batteries will eventually expire, even with  
    80 harvesting, because of the limited number of recharge cycles for the  
     74These algorithms maximizes system utility for the next time slot and 
     75maintains ENO by adjusting the duty-cycling based on the recent change 
     76of the system such as battery energy level \cite{jiang05,vigorito07}  
     77or the energy supply and consumption \cite{hsu06,kansal07}.  
     78Our work is different from these earlier works in two aspects.   
     79First, our work not only optimized the energy utility but also articulated 
     80how current system energy utility is materialized in terms of communication 
     81and sensing, whereas previous algorithms only focused on the abstract notion  
     82of duty-cycle, not articulating how such duty-cycle is translated in a real system. 
     83Second, our work uses high-level environment prediction model while previous 
     84works estimate environment using exponentially averaged history of the solar 
     85energy supply or battery level increase. Our work can predict the energy flow  
     86using atmoshperic and weather models, thus achieving better adaptability even  
     87under harvested solar energy variation under severe weather changes.   
     88In addition, while our energy architecture supports energy harvesting,  
     89the main premise of our work is that node batteries will eventually expire,  
     90even with harvesting, because of the limited number of recharge cycles for the  
    8191batteries. As a result, the architecture considers a lifetime target  
    8292for the node and determines sampling strategies accordingly.   
    … …  
    8696 
    8797Many researches have been done on adaptive sampling to provide  
    88 higher fidelity and save communication bandwidth and these  
    89 previous works can be categorized to either \textit{spatial  
    90 adaptive sampling}, \textit{temporal adaptive sampling} or 
    91 \textit{energy adaptive delivery}.   
    92 Spatial adaptive sampling adjusts sampling density depending 
    93 on how abruptly a physical phenomenon changes, and it can be  
    94 divided into \textit{mobile adaptive sampling} 
    95 \cite{batalin04sensys,zhang04iros,singh06ipsn}, which adjusts  
    96 sampling density by physically moving a sensor over a range of space,  
    97 or \textit{sampler selection} \cite{willett04ipsn,gedik07tpds}, 
    98 which selectively chooses representative sensor nodes among the  
    99 sensor nodes deployed. Temporal adaptive sampling adjusts sampling  
    100 rate over time, and it can be divided into \textit{adaptive  
    101 sensor sampling} \cite{batalin04sensys} or \textit{adaptive sensor  
    102 filtering} \cite{jain04dmsn,santini06inss} depending on where the  
    103 decision is made. Our work is similar to temporal adaptive sampling in that 
     98higher fidelity and save communication bandwidth. 
     99These previous works can be categorized to either  
     100\textit{spatial adaptive sampling} that adjusts sampling 
     101density depending on how abruptly a physical phenomenon changes 
     102\cite{batalin04sensys,zhang04iros,singh06ipsn,willett04ipsn,gedik07tpds}, 
     103\textit{temporal adaptive sampling} that adjusts sampling rate 
     104over time \cite{batalin04sensys,jain04dmsn,santini06inss}, or 
     105\textit{energy adaptive delivery}. 
     106Our work is similar to temporal adaptive sampling in that 
    104107it adjusts sampling rate over time. The differences are that our work  
    105108provides two-level services, where low frequency summary is provided  
    106109as a guaranteed service and high frequency report is provided as a  
    107 variable rate service depending on energy availability.   
    108 Lance \cite{lance08sensys} is a protocol that optimizes the data 
    109 collection rate with the constraints of varying battery energy and 
    110 lifetime target. Our work has a similarity with Lance in that 
    111 it optimizes utility (high-resolution data report rate) with 
    112 the varying energy availability and the given constraints, 
    113 but it is different in energy model and data filtering method.   
    114 As for the energy model, our work considers a solar energy 
    115 harvesting system, which requires estimation on energy  
    116 variability due to daily, seasonal and meteorological variance.  
     110variable rate service depending on energy availability.  
     111Our work is also similar to Lance \cite{lance08sensys}, a energy-aware 
     112data collection protocol in that it adjusts sampling rate depending 
     113on energy availability, but it has a few differences. First, our work 
     114considers a solar energy harvesting system where energy supply, 
     115energy storage and energy consumption varies over time. Our work 
     116uses solar energy atmospheric model and weather predictions to  
     117predict the energy trend and control the system accordingly. 
    117118Whereas, Lance assumes a simple model of non-rechargeable battery  
    118 and linear energy consumption rate. 
    119 As for the data filtering method, Lance selects data values 
    120 higher than preset threshold rather than evenly selecting the values.  
    121 This method can be effective with the case of Lance but is not  
    122 applicable in general. 
     119and linear energy consumption rate. Second, our work samples data 
     120evenly with varying sampling rate depending on energy availability. 
     121Whereas, as for data filtering, Lance selects those data that are  
     122higher than preset threshold. This method can be effective with the  
     123case of Lance but is not applicable in general. 
    123124 
    124125