Академический Документы
Профессиональный Документы
Культура Документы
Abstract—The Physical Downlink Control Channel (PDCCH) is often the case in a densely populated system with numerous
is used to signal dynamic resource assignment information in the small-packet active users, the overall system performance in
Long Term Evolution (LTE) system. In the presence of numerous terms of blocking probability is likely to be hampered by the
active users, the system performance is likely to be hindered by
shortage of control channel resource. In this paper, several simple passive manner of scheduling in PDCCH manager.
algorithms are brought forward for making efficient PDCCH In order to efficiently pack users into limited amount of
resource allocation. We first propose a minimum aggregation control resource, this paper allows more flexibility for PDCCH
level algorithm (Min-AL) to maximize the total number of resource manager to find its best way to assign control resource
scheduled users by exploiting multi-user diversity gain, followed to users chosen by data scheduler. Several types of PDCCH
by algorithms with the purpose of improving the blocking
performance of cell-edge area through co-channel interference resource allocation algorithms with feasible complexities are
avoidance (CCI-A) and priority boosting (PB). Simulation results introduced in this work. We start with one aggressive algorithm
have shown that the Min-AL algorithm achieves the best system attempting to maximize the number of UEs served by PDCCH
performance at the expense of cell-edge performance, while both by taking advantage of multi-user diversity, i.e. allocating
CCI-A and PB algorithms are effective in reaching a compromise resource to UEs in an order determined by the size of their
between system and cell-edge performance.
required resource. However, a large group of cell-edge users
could be rejected by this aggressive algorithm since more
I. Introduction
resource is needed to compensate for serious signal attenuation
In OFDM-based UTRAN Long Term Evolution (LTE) spec- and Co-Channel Interference (CCI). We then go farther by
ified by Third Generation Partnership Project (3GPP), dynamic looking for means to decrease the blocking probability of cell-
scheduling and resource allocation has been proven to greatly edge users without causing much system loss. Two algorithms
enhance spectral efficiency [1][2]. However, these gains are are subsequently presented on basis of aggressive algorithm
achieved at the cost of high signaling overhead which is for this purpose, one by improving cell-edge user channel
needed to inform scheduled users of necessary information quality through CCI avoidance, and the other by increasing
such as assigned Physical Resource Blocks (PRBs) and se- the likelihood of cell-edge users to get control resource.
lected Modulation and Coding Scheme (MCS). This paper is organized as follows: in the next section
In LTE systems, the Downlink Control Information (DCI) we describe basic assumptions and constraints involved in
messages for both uplink grant and downlink resource assign- PDCCH resource allocation; several algorithms for PDCCH
ment are signaled on the Physical Downlink Control Channel resource allocation are proposed in Section III; in Section IV
(PDCCH) [3]. During each 1ms Time Transmission Interval simulation results are shown to compare the performance of
(TTI), only the first 1 to 3 OFDM symbols are reserved to be these algorithms; we conclude this paper in Section V.
used by PDCCH and several other downlink control channels,
II. Assumptions and Constraints
e.g. Physical Control Format Indicator Channel (PCFICH).
The key objective of PDCCH resource allocation is to max- A. Basic Assumptions
imize the supportable number of scheduled User Equipments We consider a LTE cellular system featuring typical hexag-
(UEs) in each TTI while satisfying user Quality of Service onal cell layouts [5], in which a site consists of one base
(QoS) requirements, given such small subset of time and station (BS) and three cells, and the overall J cells operate on
frequency resource available. Several algorithms for effective the same frequency bandwidth. A large number of UEs are
PDCCH scheduling are proposed in [4]. According to previous uniformly distributed across the system, and serving BS/cell
work, data scheduler is assumed to have higher priority over each UE connected to is selected based on Reference Signal
PDCCH scheduler, i.e., the priorities of UEs in acquiring Received Power (RSRP). UEs bound to the inner 2/3 area of
control resource follow those generated by data scheduler. It is a site are categorized as cell-center users, the remaining UEs
true that maintenance of priorities chosen by data scheduler is near site border as cell-edge users. Since UE density per area
of great importance for dynamic resource allocation. However, is equal over the entire system, cell-center and cell-edge UEs
when there is an inadequate amount of control resource, which approximately comprise 2/3 and 1/3 of total users in each cell.
1500
Next, estimate the worst-case SINR per CCE by taking Power Site 1
where h̄ j,n and h̄l,n respectively denotes u j,n ’s average CCE Index
û j,1 , . . . , û j,N Σ j,n ≤ Σ j,n , 1 ≤ n ≤ n ≤ N , so that users 1) Power shaping: Load power for every CCE in SCS by
with lower ALs are given higher priorities to choose
control resource within their search space than users
S
PCCE = θ j · P̄CCE , (6)
with higher ALs. If two users have identical ALs, their where θ j ∈ (0, 1] is a configurable parameter chosen
relative order is kept the same as in U j . independently in each site. CCEs in PSCs are loaded
4) Physical resource allocation: Allocate CCEs to UEs with nominal transmit power per CCE P̄CCE (see (3)).
following an order defined by Û j , one UE at a time. Calculate worst-case SINR γ̃ Pj,n corresponding to S Pj for
For each user u j,n (1 ≤ n ≤ N ) of Σ j,n 0, an attempt is
each cell-edge user u j,n ∈ Uedge j according to power
made to find an empty PDCCH candidate based on (1).
pattern in neighboring cells (refer to (4)). Note that two
If vacant PDCCH candidates are found available in its
CCE segments in S Sj may experience different CCI, a
search space, the first one is assigned to u j,n and then
marked “occupied”; otherwise, u j,n is blocked. Since set of SINRs γ̃ sj,n = γ̃ s,1 s,2
j,n , γ̃ j,n are then calculated for
blocking of one user does not necessarily mean blocking each cell-center user u j,n ∈ Ucenter j .
for another under PDCCH constraints, every user in Û j 2) AL selection: Determine AL Σ j,n for cell-edge user
should be given a try until all CCEs are exhausted. u j,n as the minimum one whose PDCCH candidates
(at least one) locate within S Pj and with which the
The AL-based scheduling (step 3) in Min-AL) exploits the target BLER can be met by γ̃ Pj,n (see (5)). Supportable
benefits of multi-user diversity by allocating the current CCEs ALs Σ1j,n and Σ2j,n for cell-center user u j,n are chosen
to UEs who can best utilize them. In general, UEs assigned likewise on SCS-related CCE segments, and then let
with low ALs on PDCCH are likely to have good downlink Σ j,n = min(Σ1j,n , Σ2j,n ). Set Σ j,n and Σ j,n 0 if no suitable
channel for high data rate. The Min-AL, therefore, is also able aggregation level is found.
to improve system throughput by passing a large number of 3) User scheduling: Use AL-based scheduling given in 3)
“good” UEs. However, Min-AL dose not treat all users equally, of Min-AL to obtain updated scheduling sublists Ûedge j
especially those cell-edge UEs suffering from serious CCI. In and Ûcenter
j respectively with Σ j,n and Σ j,n .
order to find a good compromise between system and cell-edge 4) Physical resource allocation: The procedure resembles 4)
performance, two algorithms are proposed afterwards. in Min-AL except for an additional restriction imposed
1501
TABLE I
Cell-center UE list ˆ center
U Simulation Parameters and Assumptions
j
... Selector
Updated UE list ˆ
U Parameter Values and Assumptions
j
W1 64 sites-3cells/site (only central 16 sites are
Cell layouts
observed), ISD =500m
Cell-edge UE list ˆ edge
U ...
Propagation model Urban Area [5]
j
W2
Tail Head Shadowing deviation 10dB
... 2Tx*2Rx SFBC (antenna radiation pattern
Antenna configuration
Tail Head and gain for BS and UE in [5] )
Multi-path model Extended Pedestrian A 5Hz (EPA5)
BS transmit power 46dBm
Fig. 3. Priority boosting algorithm System bandwidth 10MHz (50RB)
System loading Full buffer services
Number of downlink UEs 50 UEs per cell-or otherwise stated
Data scheduling Round Robin
on available resource for different types of UEs, that is, Number of CCEs per cell 30 (3 OFDM symbols)
S Pj for Ûedge
j and S Sj for Ûcenter
j . CCE Aggregation levels 1, 2, 4, 8 CCE
PDCCH modulation QPSK
Note that the UE-specific search space changes over time, PDCCH BLER target 1% BLER
a temporarily declined cell-edge/center UE due to a lack PDCCH payload 46bits
of suitable PDCCH candidates within PCS/SCS could still
be a competitive applicant for control resource in next TTI. (a) RS (b) Min-AL
y (m)
y (m)
4000 8 16 4000 8 16
1 9 1 9
8
C. Priority Boosting Algorithm (PB)
2 10 2 10
3 11 3 11
2000 4 12
2000 4 12 7
An alternative for cutting down blocking probability of cell- 1000 2000 3000 4000 5000 1000 2000 3000 4000 5000
6
edge users is to raise their ranks in the order list produced by x (m) x (m) 5
(c) CCI-A, T = 0.5 (d) PB, P = 2
AL-based scheduling, thus permitting them to choose their 6000
4
6000
favorable resource with higher priorities. This idea can be 6
5
14
13
6
5
14
13
3
y (m)
4000 8 16 4000 8 16
1 9 1 9
4
3
10
12
11
2
4
3
10
12
11
1
2000 2000 0
turning on each of the two prioritized lists with a certain
probability (W1 versus W2 for cell-center and cell-edge UE 1000 2000 3000 4000 5000
x (m)
1000 2000 3000 4000 5000
x (m)
list) to build the final UE list for resource allocation. Fig. 4. Aggregation level distribution over 16 central statistical sites, active
Algorithm: PB users are marked by dots of different color corresponding to various ALs
(blocked UEs with 0-AL are marked by blue dots).
1) Power shaping: This is the same as 1) for Min-AL.
2) AL selection: Use the same procedure as for Min-AL.
3) User scheduling: First, implement AL-based scheduling IV. Simulation Results
for Uedge
j and Ucenter
j and get two updated prioritized sub-
In this section, simulation results are shown for various
lists, i.e., cell-edge sublist Ûedge
j and cell-center sublist algorithms presented above to evaluate their impact on the
center
Û j . A bi-directional selector is then used to merge downlink system performance. The quasi-static system-level
these two sublists into a new UE list Û j , in a manner simulation is done in accordance with PDCCH descriptions
of picking UEs one-by-one from the top of either Ûedge j
in 3GPP LTE Release 8 and link-to-system model given
or Ûcenter with probability W versus W . One simple in Section II. More details of simulation parameters and
j 1 2
way to carry out this process is through Bernoulli trials assumptions can be found in Tab. 1. To witness the perfor-
which takes value 1 (cell-center list) with probability W1 mance gain yielded by PDCCH-oriented scheduling, random
and value 0 (cell-edge list) with probability W2 . scheduling (RS), an approach whose UE priorities are decided
4) Physical resource allocation: Same as 4) for Min-AL. by data scheduling instead of PDCCH scheduling, is adopted
for comparisons in the following discussion. Other resource
To quantitatively analyze how much the cell-edge users are
allocation procedures of RS are the same with Min-AL.
favored, define preferential factor as follows
Fig. 4 plots aggregation level distribution pattern of dif-
W2 /W1 ferent algorithms, i.e., each user’s geographic location and
μ = , s.t.W1 + W2 = 1, (7)
Ûedge Ûcenter its assigned number of CCE(s). Compared with RS, whose
j j
scheduled UEs spreading over the entire system in a random
where the function of |·| denotes the size of the corresponding manner, the Min-AL favors more the 1- and 2-CCE cell-center
vector. In the regime of μ > 1, the likelihood that cell- users, leaving the majority of cell-edge users unattended. By
edge UEs are picked by selector is greater in proportion to contrast, CCI-A and PB have relative larger coverage than
that of cell-center UEs. The impact of such biased selection Min-AL since more users in cell edge area are accommodated.
will become more conspicuous with a larger μ. Once the Fig. 5 shows the Probability Density Functions (PDFs) of
preferential factor is determined towards the desirable system the number of total and cell-edge scheduled users for each
performance, probability W1 and W2 can be derived from (7). algorithm. All our proposed algorithms, i.e., Min-AL, CCI-
1502
(a) 10 1.4
Probability Density Function
(Mbps)
0.1 0.8
5 12.8 UEs
2.7 UEs
4 0.6
0
5 10 15 20 25 30 3 0.4
(b) 2
Probability Density Function
0.2
0.4
1
RS
0 0
0.3 Min-AL
CCI-A, T =0.5 RS Min-AL CCI-A, ș PB, ȝ
0.2
PB, P =2
0.1 Fig. 7. Average downlink cell throughput (left) and average cell-edge
throughput (right), with the average number of cell/ cell-edge scheduled UEs
0
0 2 4 6 8 10 12 given on top of each bar corresponding a specific algorithm.
Number of scheduled users
Fig. 5. PDFs of the number of (a) cell scheduled UEs, and (b) cell-edge
scheduled UEs. exhibits higher cell-edge throughput than RS although it has
fewer scheduled users. This is mainly contributed to AL-based
1
scheduling used in CCI-A which picks a larger percentage of
0.95
UEs with good channel quality.
0.9
0.85
V. Conclusion
In this paper we consider PDCCH resource allocation prob-
CCE Utilization
0.8
1503