Scribd
Upload
68 views8 pages

System Model: - How To Schedule Mmwave

The document proposes a system model for scheduling mmWave backhaul transmissions from a base station to multiple small cell access points to minimize latency. It formulates the problem as a constrained Markov decision process that optimizes the power allocation policy across multiple access points. Reinforcement learning algorithms are suggested to learn the optimal transmission policy without requiring full knowledge of state transition probabilities.

Uploaded by

Sơn Đinh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
68 views8 pages

System Model: - How To Schedule Mmwave

The document proposes a system model for scheduling mmWave backhaul transmissions from a base station to multiple small cell access points to minimize latency. It formulates the problem as a constrained Markov decision process that optimizes the power allocation policy across multiple access points. Reinforcement learning algorithms are suggested to learn the optimal transmission policy without requiring full knowledge of state transition probabilities.

Uploaded by

Sơn Đinh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 8

1

System Model

mmWave backhaul with massive


MIMO
K small cell APs connects to a BS,
BS sends data to K small cell Aps
simultaneously through mmWave
massive MIMO links.
Base station contains antennas
At time slot t, traffic arrival to AP k
is (), assuming Markov
Modulated Poisson Process
(MMPP) How to schedule mmWave
backhaul transmission to
minimize the latency at the
BS.

Center Proprietary - Terms of Center Membership Agreement Apply


2
Beamforming (1)
Transmitter Receiver
Phase shifter elements
Mixer PA Phase shifter
DAC LNA Mixer
ADC
Baseband Processor

Baseband Processor
MIMO Precoder

MIMO Combiner
RF chains
beamformer , , RF chains
beamformer

DAC ADC

RF beamformer RF beamformer

BS antenna arrays contain antennas, shared by RF chains


BS sends , data streams to AP k.

, RF chains , , used to send , data streams
Each RF chain connects to antennas.

AP k is assumed to support , chains, , , and
, antennas
Center Proprietary - Terms of Center Membership Agreement Apply
3
Beamforming (2)

The BS applies an , , baseband precoder matrix P, followed by an

, RF beamformer weight matrix W on , 1 data signal vector
The received , 1 signal vector on , antennas at AP k
= +

is the , channel matrix
Assume the transmitted signals to different APs are orthogonal due to
beamforming and zero forcing cancellation.
The processed signal considering hybrid beamforming at both the transmitter and
the receiver can be expressed in a generic form as follows:
= +
: , 1 the processed signal at the receiver

: , , receive RF beamformer weight matrix

: , , baseband combining matrix.
For simplicity, assume only one data stream is sent to each AP and an AP has one
RF chain,
= +

Center Proprietary - Terms of Center Membership Agreement Apply


4
Beamforming (3)

Joint optimization of hybrid beamforming in the RF domain and in the digital domain
Maximize the channel capacity

argmax log 2 (, + 2
( , )( , ) )
,

, =
= 2 is the ratio between transmit power and noise power, that is, transmit signal
to noise ratio (SNR).
For the case of single stream transmission and single receiving RF chain, perform a joint
optimization of the transmitter/receiver (Tx/Rx) RF beamformer and the Tx baseband
precoder.
The hybrid beamforming optimization problem:
= +

, = argmax log 2 (1 +
)
2

=
The data rate achieved with massive MIMO beamforming.

= log 2 (1 + 2 ,
, )

Center Proprietary - Terms of Center Membership Agreement Apply


5
Formulation (1)

At time slot t, is the queue size at the BS for traffic to


AP k.
( + 1) = + min[ , ]
is the data rate achieved with massive MIMO
beamforming.

= log 2 (1 + , ()
,

())
2

Center Proprietary - Terms of Center Membership Agreement Apply


6
Formulation (2)

The state of the BS at time slot t is


= { , } = { t , , }
At each time slot t, BS will determine the optimal transmission policy, that
is, the power allocation to transmit each data stream to each AP, based on
the observed state, that is, = { , }, which determines
the amount of data sent to each AP in time slot t.
The optimal transmission policy minimizes the long-term average delay of
the data to multiple APs, which can be formulated as
1
= argmin lim =1 [( , k )]

s.t. () : total power


The expectation is over the queue at each time slot t
( , k ) = : delay cost function which depends
on both the random state and the action performed by the BS.

Center Proprietary - Terms of Center Membership Agreement Apply


7
Formulation (3)

Constrained Markov decision process (MDP) problem with the average cost
criterion.
Let be the value function for a state under policy
The optimal state value function = min () satisfies the Bellmans

optimality equation
= min [( , , k ) + (| , , k ) ]
,

1
= lim [( , k )]

=1
Solving the above needs the full knowledge of state transition probabilities.
Develop reinforcement learning algorithms for achieving the optimal transmission
policy, which do not require such statistical state information.

Center Proprietary - Terms of Center Membership Agreement Apply


8
Extension

Consider multihop
transmission and
queuing at the
intermediate APs and
relays.

Center Proprietary - Terms of Center Membership Agreement Apply

You might also like