Wireless Ad Hoc and Sensor Networks
Wireless Ad Hoc and Sensor Networks Wireless Ad Hoc and Sensor Networks
Distributed Power Control and Rate Adaptation 273Now substituting a new state of Equation 6.41 in Equation 6.49, andincluding it in the last Nth iteration, a corresponding outgoing trafficcomponent w n, we get2 2J ( z( k )) =Q ( z( k )) +R ( u( k )) +J 1( z(k+ 1))k i k i k i k+ i(6.51)Applying the DP approach to Equation 6.42 we haveJ (( z N)) = Q ( z( N))N N iJ (()) z k = min E{ Q ( z()k ) 2 2+ R ( u( k )) +J ( z( k ) +u( k))}k2u kk ik i k+ 1 i i(6.52)First, we expand the one before the last iteration2JN−1(( z N− 1)) = min E{ QN−1( zi( N− 1)) + RN−1( u i ( N − 1))uN−12+Q ( z ( N −1) +u ( N −1)) }N−1N i i22= Q ( z ( N−1)) + min ER { ( u( N −1)) +Q ( z( N−1))iuN−1N−1i N i22+Qz ( N−1) u( N −1) +Q ( u ( N −1)) 2 }u NN i i N i(6.53)The minimization of Equation 6.44 with respect to i( − 1)is performedby differentiating Equation 6.44 and equating it to zero which yields( )ui * ( N− 1) =−zi( N−1)⋅ QN QN + RN−1(6.54)By substitutingu i( N− 1)in Equation (6.44) with Equation 6.45 we get,22J (( z N− 1)) = Q ( z ( N− 1)) + R ( z ( N−1)) (− Q /[ Q + R ])N−1 N−1i N−1iN N N−1222+ Q ( z ( N−1)) ( 1 − Q /[ Q + R ]) = G ( z ( N−1))N i N N N−1N−1i2(6.5)where2G = Q + Q R −R Q ( R Q )N−1( ) +2N−1 N−1 N N−1N−1 N N−1N= Q + Q [ ( )]N 1−QN QN −RN−1(6.56)
274 Wireless Ad Hoc and Sensor NetworksFollowing the preceding calculations we can calculate optimal inputfor k = N−2, N−3,....,0. In such a case, an optimal law for every k is equaltowhere( )u * ( k) =−z( k)⋅ G G + Ri i N+ 1 k+1 k(6.57)G =QnG =GNk k+ 1⎛ Gk+1 ⎞1−⎝⎜G +G +Qk+ 1 k ⎠⎟k(6.58)However, because the transmission duration is unknown, it is desirableto calculate a steady-state solution by assuming an infinite flow. Thesteady-state solution is more useful in implementation, because most ofthe calculations can be performed offline before network deployment andonly limited calculations have to be done online. In such a case, Equation6.58 becomesu * ( k) =−z( k) ⋅ G ( G+R )i i k(6.59)where G is stable state solution ( k →∞)of Equation 6.49, Rk = α ⋅P 0( k)is aparameter of a cost function with P0( k)being the transmission powervalue calculated by the DPC for the next transmission.By substituting Equation 6.33 and Equation 6.41 in Equation 6.59 wecan calculate the stationary law that depends directly on the queue utilizationasu*( k ) = −G( q( k) −q+E{ w ( k)})/( G+R )i ideal i k(6.60)This control law is applied before each transmission to calculate adesired burst size u*. Next, the optimal modulation rate is selected asin Subsection 6.9.2.2, as the lowest rate that can support a given burstsize.6.9.4 Additional Conditions for Modulation SelectionTransmission power is physically limited by the node’s hardware.Hence, the rate adaptation has to exclude modulation rates that requiretransmission power higher than the maximum available value. Thisthreshold is calculated according to Equation 6.24. The maximum powerwill dictate the maximum modulation rate that can be applied for agiven channel state. If the modulation rate is reduced, then the burstsize will be reduced as well.
- Page 245 and 246: 222 Wireless Ad Hoc and Sensor Netw
- Page 247 and 248: 224 Wireless Ad Hoc and Sensor Netw
- Page 249 and 250: 226 Wireless Ad Hoc and Sensor Netw
- Page 251 and 252: 228 Wireless Ad Hoc and Sensor Netw
- Page 253 and 254: 230 Wireless Ad Hoc and Sensor Netw
- Page 255 and 256: 232 Wireless Ad Hoc and Sensor Netw
- Page 257 and 258: 234 Wireless Ad Hoc and Sensor Netw
- Page 259 and 260: 236 Wireless Ad Hoc and Sensor Netw
- Page 261 and 262: 238 Wireless Ad Hoc and Sensor Netw
- Page 263 and 264: 240 Wireless Ad Hoc and Sensor Netw
- Page 265 and 266: 242 Wireless Ad Hoc and Sensor Netw
- Page 267 and 268: 244 Wireless Ad Hoc and Sensor Netw
- Page 269 and 270: 246 Wireless Ad Hoc and Sensor Netw
- Page 271 and 272: 248 Wireless Ad Hoc and Sensor Netw
- Page 273 and 274: 250 Wireless Ad Hoc and Sensor Netw
- Page 275 and 276: 252 Wireless Ad Hoc and Sensor Netw
- Page 277 and 278: 254 Wireless Ad Hoc and Sensor Netw
- Page 279 and 280: 256 Wireless Ad Hoc and Sensor Netw
- Page 281 and 282: 258 Wireless Ad Hoc and Sensor Netw
- Page 283 and 284: 260 Wireless Ad Hoc and Sensor Netw
- Page 285 and 286: 262 Wireless Ad Hoc and Sensor Netw
- Page 287 and 288: 264 Wireless Ad Hoc and Sensor Netw
- Page 289 and 290: 266 Wireless Ad Hoc and Sensor Netw
- Page 291 and 292: 268 Wireless Ad Hoc and Sensor Netw
- Page 293 and 294: 270 Wireless Ad Hoc and Sensor Netw
- Page 295: 272 Wireless Ad Hoc and Sensor Netw
- Page 299 and 300: 276 Wireless Ad Hoc and Sensor Netw
- Page 301 and 302: 278 Wireless Ad Hoc and Sensor Netw
- Page 303 and 304: 280 Wireless Ad Hoc and Sensor Netw
- Page 305 and 306: 282 Wireless Ad Hoc and Sensor Netw
- Page 307 and 308: 284 Wireless Ad Hoc and Sensor Netw
- Page 309 and 310: 286 Wireless Ad Hoc and Sensor Netw
- Page 311 and 312: 288 Wireless Ad Hoc and Sensor Netw
- Page 313 and 314: 290 Wireless Ad Hoc and Sensor Netw
- Page 315 and 316: 292 Wireless Ad Hoc and Sensor Netw
- Page 317 and 318: 294 Wireless Ad Hoc and Sensor Netw
- Page 319 and 320: 296 Wireless Ad Hoc and Sensor Netw
- Page 321 and 322: 298 Wireless Ad Hoc and Sensor Netw
- Page 323 and 324: 300 Wireless Ad Hoc and Sensor Netw
- Page 326 and 327: 7Distributed Fair Scheduling in Wir
- Page 328 and 329: Distributed Fair Scheduling in Wire
- Page 330 and 331: Distributed Fair Scheduling in Wire
- Page 332 and 333: Distributed Fair Scheduling in Wire
- Page 334 and 335: Distributed Fair Scheduling in Wire
- Page 336 and 337: Distributed Fair Scheduling in Wire
- Page 338 and 339: Distributed Fair Scheduling in Wire
- Page 340 and 341: Distributed Fair Scheduling in Wire
- Page 342 and 343: Distributed Fair Scheduling in Wire
- Page 344 and 345: Distributed Fair Scheduling in Wire
274 <strong>Wireless</strong> <strong>Ad</strong> <strong>Hoc</strong> <strong>and</strong> <strong>Sensor</strong> <strong>Networks</strong>Following the preceding calculations we can calculate optimal inputfor k = N−2, N−3,....,0. In such a case, an optimal law for every k is equaltowhere( )u * ( k) =−z( k)⋅ G G + Ri i N+ 1 k+1 k(6.57)G =QnG =GNk k+ 1⎛ Gk+1 ⎞1−⎝⎜G +G +Qk+ 1 k ⎠⎟k(6.58)However, because the transmission duration is unknown, it is desirableto calculate a steady-state solution by assuming an infinite flow. Thesteady-state solution is more useful in implementation, because most ofthe calculations can be performed offline before network deployment <strong>and</strong>only limited calculations have to be done online. In such a case, Equation6.58 becomesu * ( k) =−z( k) ⋅ G ( G+R )i i k(6.59)where G is stable state solution ( k →∞)of Equation 6.49, Rk = α ⋅P 0( k)is aparameter of a cost function with P0( k)being the transmission powervalue calculated by the DPC for the next transmission.By substituting Equation 6.33 <strong>and</strong> Equation 6.41 in Equation 6.59 wecan calculate the stationary law that depends directly on the queue utilizationasu*( k ) = −G( q( k) −q+E{ w ( k)})/( G+R )i ideal i k(6.60)This control law is applied before each transmission to calculate adesired burst size u*. Next, the optimal modulation rate is selected asin Subsection 6.9.2.2, as the lowest rate that can support a given burstsize.6.9.4 <strong>Ad</strong>ditional Conditions for Modulation SelectionTransmission power is physically limited by the node’s hardware.Hence, the rate adaptation has to exclude modulation rates that requiretransmission power higher than the maximum available value. Thisthreshold is calculated according to Equation 6.24. The maximum powerwill dictate the maximum modulation rate that can be applied for agiven channel state. If the modulation rate is reduced, then the burstsize will be reduced as well.