# US 9,001,686 B2 · Interference estimation with TDM

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 9,001,686 B2
- Document ID: US9001686B2
- Status: Granted
- Kind: B2
- Application number: 13/853,369
- Filed: 2013-03-29
- Published: 2014-10-02
- Granted: 2015-04-07
- Earliest claimed: 2013-03-29
- Assignees: Telefonaktiebolaget L M Ericsson (publ)
- Applicants: Telefonaktiebolaget L M Ericsson (publ)
- Inventors: Torbjörn Wigren
- Web page: https://patentgrep.com/patent/US9001686B2

## CPC

- H04W 24/08: Testing, {supervising or monitoring} using real traffic
- H04B 17/345: Interference values ({signal-to-interference ratio [SIR] or carrier-to-interference ratio [CIR]} H04B17/336)
- H04J 11/0023: {Interference mitigation or co-ordination (direct sequence spread spectrum [DSSS] systems H04B1/7097; frequency hopping H04B1/713; allocation criteria for ingress interference avoidance H04L5/0062; frequency allocation criteria for requirements on out-of-channel emissions H04L5/0066; arrangements for removing intersymbol interference or baseband equalisers H04L25/03006; peak power aspects in multicarrier modulation H04L27/2614; power management H04W52/00; traffic scheduling H04W72/54, H04W72/541)}
- H04J 11/005: {of intercell interference}
- H04L 1/1887: {Scheduling and prioritising arrangements}
- H04L 5/0073: {Allocation arrangements that take into account other cell interferences}
- H04W 72/52: based on load
- H04L 1/1845: {Combining techniques, e.g. code combining}
- H04W 24/00: Supervisory, monitoring or testing arrangements
- H04W: WIRELESS COMMUNICATION NETWORKS (broadcast communication H04H; communication systems using wireless links for non-selective communication, e.g. wireless extensions H04M1/72)
- H04B: TRANSMISSION
- H04J: MULTIPLEX COMMUNICATION (transmission in general H04B; peculiar to transmission of digital information H04L5/00; systems for the simultaneous or sequential transmission of more than one television signal H04N7/08; in exchanges H04Q11/00; stereophonic systems H04S)
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Mobile broadband traffic has been exploding in wireless networks resulting in an increase of interferences and reduced operator control. Networks are also becoming more heterogeneous putting additional demand in interference management. Scheduler schedules uplink transmissions from UEs based on a load prediction algorithm that typically assumes worst case. However, UEs do not always use full power granted, and thus, much of granted radio resources are wasted. To address these and other issues, technique(s) to accurately predict/estimate other cell interferences and thermal noise separately and to accurately predict/estimate load utilization probability and variance is(are) described. Inventive estimation technique(s) can be used to schedule UEs to more fully utilize available radio resources, e.g., in TD-scheduling operations. Extended Kalman filtering can be adapted for use in estimation providing low order computational complexity.

## Claims

### Claim 1

A method performed at a radio network node corresponding to a cell of interest in a wireless network, the method comprising:
estimating an interference-and-noise sum Pother(t1)+PN(t1) based at least on a load utilization probability estimate {circumflex over (p)}load(t0) and an interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0) applicable at a time t0 to obtain an interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) applicable at a time t1, wherein t1−t0=T>0; and
estimating an other cell interference Pother(t1) based at least on the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) and a thermal noise estimate {circumflex over (P)}N(t1) to obtain an other cell interference estimate {circumflex over (P)}other(t1) applicable at the time t1,
wherein the interference-and-noise sum Pother(t)+PN(t) expresses a sum of undesired signals, other than an own cell load Pown(t), applicable at the time t, and the interference-and-noise sum estimate {circumflex over (P)}other(t)+{circumflex over (P)}N(t) being an estimate thereof, the own cell load Pown(t) expressing a sum of signals due to wireless activities in the cell of interest applicable at the time t,
wherein the other cell interference Pother(t) expresses a sum of interferences present in the cell of interest due to wireless activities applicable at the time t in one or more cells other than in the cell of interest, and the other cell interference estimate {circumflex over (P)}other(t) being an estimate thereof,
wherein a thermal noise PN(t) expresses a sum of undesired signals present in the cell of interest at the time t other than the own cell load Pown(t) and other than the other cell interference Pother(t), and the thermal noise estimate {circumflex over (P)}N(t) being an estimate thereof, and
wherein the interference-and-noise sum estimate {circumflex over (P)}other(t)+{circumflex over (P)}N(t) and the other cell interference estimate {circumflex over (P)}other(t1) are based on a state space model comprising at least a measurement matrix C(t1).

### Claim 2 (depends on 1)

The method of claim 1, wherein the method is performed in a TD-scheduling environment such that the other cell interference Pother(t1) is estimated only for each TD slot in a subset of TD slots, the subset being determined as a union of TD slots in which a same set of wireless terminals are scheduled.

### Claim 3 (depends on 1)

The method of claim 1, further comprising:
estimating a load utilization probability pload(t1) based at least on the load utilization probability estimate {circumflex over (p)}load(t0) and the interference-and-noise sum estimate {circumflex over (P)}other(t0))+{circumflex over (P)}N(t0) applicable at the time t0 to obtain a utilization probability estimate {circumflex over (p)}load(t1) applicable at the time t1,
wherein the load utilization probability pload(t) expresses a relationships between radio resource grants scheduled to one or more cell terminals and radio resource grants used by the same cell terminals applicable at a time t, each cell terminal being a wireless terminal in the cell of interest, and the load utilization probability estimate {circumflex over (p)}load(t) being an estimate thereof, and
wherein the load utilization probability estimate {circumflex over (p)}load(t) is estimated based on the state space model comprising at least the measurement matrix C(t1).

### Claim 4 (depends on 3)

The method of claim 3, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) comprise:
calculating a scheduled load factor Lown(t1−TD);
obtaining a used load factor Lown(t1−TD);
measuring a load utilization
L
_
own
⁡
(
t
1
-
T
D
)
L
own
⁡
(
t
1
-
T
D
)
based on the scheduled and measured load factors;
obtaining the utilization probability estimate {circumflex over (p)}load(t1); and
obtaining the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1),
wherein the scheduled load factor Lown(t−TD) expresses an amount of the radio resource grants scheduled to be used by the cell terminals for uplink transmissions at the time t, and
wherein the used load factor Lown(t−TD) expresses an amount of the scheduled radio resource grants used by the cell terminals for the uplink transmissions at the time t, and
wherein TD represents a delay between calculation of the scheduled load factor and a time the schedule takes effect on an air interface.

### Claim 5 (depends on 4)

The method of claim 4, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) comprise:
determining whether a HARQ process is active at the time t1,
wherein the load utilization probability estimate {circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) are obtained based on the measured load utilization
L
_
own
⁡
(
t
1
-
T
D
)
L
own
⁡
(
t
1
-
T
D
)
when it is determined that the HARQ process is active, and
wherein the load utilization probability estimate {circumflex over (p)}load(t1) is obtained from the previous load utilization probability estimate {circumflex over (p)}load(t0) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) is obtained from the previous interference-and-noise sum estimate {circumflex over (P)}other(t0))+{circumflex over (P)}N(t0) when it is determined that the HARQ process is not active.

### Claim 6 (depends on 3)

The method of claim 3, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) comprise:
measuring a total wideband power yRTWP(t1);
obtaining the load utilization probability estimate {circumflex over (p)}load(t1); and
obtaining the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1).

### Claim 7 (depends on 6)

The method of claim 6, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) further comprise:
determining whether a HARQ process is active at the time t1,
wherein the load utilization probability estimate {circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) are obtained based on the measured total wideband power yRTWP(t1) when it is determined that the HARQ process is active at the time at the time t1, and
wherein when it is determined that the HARQ process is not active at the time at the time t1, the load utilization probability estimate {circumflex over (p)}load(t1) is obtained from the previous load utilization probability estimate {circumflex over (p)}load(t0) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) is obtained from the previous interference-and-noise sum estimate {circumflex over (P)}other(t0))+{circumflex over (P)}N(t0).

### Claim 8 (depends on 6)

The method claim 6, wherein when it is determined that the HARQ process is active, the step of obtaining the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) comprises:
determining a gain factor g(t1) based on the load utilization probability estimate {circumflex over (p)}load(t1) and the scheduled load factor Lown(t0);
modeling the measured total wideband power yRTWP(t1) as a combination of the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) weighted by the gain factor g(t1) and a measurement uncertainty eRTWP(t1); and
obtaining the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) based on the measured total wideband power yRTWP(t1) and the model thereof.

### Claim 9 (depends on 8)

The method of claim 8, wherein the step of determining the gain factor g(t1) comprises:
determining a load factor bias ΔLown(t1); and
determining the gain factor g(t1) based on the based at least on the load utilization probability estimate {circumflex over (p)}load(t1) the scheduled load factor Lown(t0), and the load factor bias ΔLown(t1),
wherein the load factor bias ΔLown(t) expresses an error of the scheduled load factor Lown(t).

### Claim 10 (depends on 3)

The method of claim 3, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) comprise:
modeling in a state vector x(t) of the state space model a first state x1(t)=pload(t) as a load utilization probability, a second state x2(t)=Pother(t)+PN(t) as the interference-and-noise sum, a third state x3 (t)=Δ Lown(t) as a load factor bias expressing an error of a scheduled load factor, and a fourth state x4(t)=x1(t−T) reflecting that the load utilization probability measurement is subject to a delay corresponding to the sampling period T;
modeling a measured total wideband power yRTWP(t) and a measured load utilization yloadUtilization(t) in an output measurement vector y(t) of the state space model;
obtaining a predicted state vector {circumflex over (x)}(t1|t0) which includes therein first and second predicted states {circumflex over (x)}1(t1|t0), {circumflex over (x)}2(t1|t0) whose values are based on the load utilization probability estimate {circumflex over (p)}load(t0) and the interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0);
updating the predicted state vector {circumflex over (x)}(t1|t0) based on one or more measurements included in an output measurement vector y(t1) applicable at the time t1 to obtain an estimated state vector {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1); and
obtaining first and second estimated states {circumflex over (x)}1(t1), {circumflex over (x)}2(t1) from the estimated state vector {circumflex over (x)}(t1) respectively as the load utilization probability estimate {circumflex over (x)}1(t1)=pload(t1) and the interference-and-noise sum estimate {circumflex over (x)}2(t1)={circumflex over (P)}other(t1)+{circumflex over (P)}N(t1),
wherein modeling errors and measurement errors are incorporated in the state space model as a model error vector w(t) and a measurement error vector e(t),
wherein the predicted state vector {circumflex over (x)}(t|t−T) denotes a prediction of the state vector x(t) based on information available up to a time t−T, and
wherein the estimated state vector {circumflex over (x)}(t|t)={circumflex over (x)}(t) denotes an estimate of the state vector x(t) based on information available up to the time t.

### Claim 11 (depends on 10)

The method of claim 10, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) further comprise:
determining whether a HARQ process is active at the time t1,
wherein the steps of modeling the states in the state vector x(t), modeling the outputs in the output measurement vector y(t), obtaining the predicted state vector {circumflex over (x)}(t1|t0), and updating the predicted state vector {circumflex over (x)}(t1|t0) are performed only when it is determined that the HARQ process is active, and
wherein the first and second estimated states {circumflex over (x)}1(t1), {circumflex over (x)}2(t1) are obtained from the previous the previous states {circumflex over (x)}1(t0), {circumflex over (x)}2(t0) when it is determined that the HARQ process is not active at the time at the time t1.

### Claim 12 (depends on 10)

The method of claim 10, wherein the step of updating the predicted state vector {circumflex over (x)}(t1|t0) comprises:
modeling the measured total wideband power yRTWP(t1) applicable at the time t1 as
y
RTWP
⁡
(
t
1
)
=
x
2
⁡
(
t
1
)
1
-
L
own
⁡
(
t
1
-
T
D
)
⁢
x
1
⁡
(
t
1
)
+
x
3
⁡
(
t
1
)
+
e
RTWP
⁡
(
t
1
)
,
TD representing a delay between calculation of the schedule and a time the schedule takes effect on an air interface;
modeling the load utilization yloadUtilization(t1) applicable at the time t1 as yloadUtilization(t1)=x4(t1)+eloadUtilization(t1);
obtaining the measurement matrix C(t1) around the predicted state vector {circumflex over (x)}(t1|t0), the predicted state vector {circumflex over (x)}(t1|t0) including first, second, third, and fourth predicted states {circumflex over (x)}1(t1|t0), {circumflex over (x)}2(t1|t0), {circumflex over (x)}3(t1|t0), {circumflex over (x)}4(t1|t0) predicted based on data upto the time t0;
obtaining a Kalman gain matrix Kf(t1) based on at least the measurement matrix C(t1), the measurement error vector e(t1), and a predicted covariance matrix P(t1|t0) corresponding to the predicted state vector {circumflex over (x)}(t1|t0);
updating the predicted state vector {circumflex over (x)}(t1|t0) based on at least the Kalman gain matrix Kf(t1), the output measurement vector y(t1), and the measurement vector c(x(t1)) to obtain the estimated state vector {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1), the estimated state vector {circumflex over (x)}(t1) including the first, second, third, and fourth estimated states {circumflex over (x)}1(t1), {circumflex over (x)}2(t1), {circumflex over (x)}3(t1), {circumflex over (x)}4(t1); and
updating the predicted covariance matrix P(t1|t0) based on at least the Kalman gain matrix Kf(t1) and the measurement matrix C(t1) to obtain an updated covariance matrix P(t1|t1) corresponding to the estimated state vector {circumflex over (x)}(t1).

### Claim 13 (depends on 10)

The method of claim 10, wherein the step of obtaining the measurement matrix C(t1) comprises:
determining whether a HARQ process is active at the time t1;
determining the measurement matrix C(t1) linearized around the predicted state vector {circumflex over (x)}(t1|t0) such that
C
⁡
(
t
)
=
∂
c
⁡
(
x
)
∂
x
⁢
❘
x
=
x
^
⁡
(
t
⁢
⁢
1
❘
t
⁢
⁢
0
)
when it is determined that the HARQ process is active at the time t1; and
settings the measurement matrix C(t1) to a predetermined value when it is determined that the HARQ process is not active at the time t1,
wherein the step of obtaining the Kalman gain matrix Kf(t1) comprises determining Kf(t1)=P(t1|t0)CT(t1)(C(t1)P(t1|t0)CT(t1)+R2(t1))−1 in which CT(t) is a transpose of the measurement matrix C(t) and (R2(t)) is a measurement covariance matrix corresponding to the measurement error vector e(t),
wherein the step of updating the predicted state vector {circumflex over (x)}(t1|t0) to obtain the estimated state vector {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1) comprises determining {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1|t0)+Kf(t1)(y(t1)−c({circumflex over (x)}(t1|t0))), and
wherein the step of updating the predicted covariance matrix P(t1|t0) to obtain the updated covariance matrix P(t1|t1) comprises determining P(t1|t1)=P(t1|t0)−Kf(t1)C(t1)P(t1|t0).

### Claim 14 (depends on 10)

The method of claim 10, wherein the step of estimating the load utilization probability pload(t1) and the step of estimating the interference-and-noise sum Pother(t1)+PN(t1) further comprise:
projecting the estimated state vector {circumflex over (x)}(t1) based on at least the system matrix A(t1) to obtain a predicted state vector {circumflex over (x)}(t2|t1) which includes the first, second, third, and fourth predicted states {circumflex over (x)}1(t2|t1), {circumflex over (x)}2(t2|t1), {circumflex over (x)}3(t2|t1), {circumflex over (x)}4(t2|t1), in which the first and second predicted states {circumflex over (x)}1(t2|t1) and {circumflex over (x)}2(t2|t1) values are based on the load utilization probability estimate {circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1); and
projecting the updated covariance matrix P(t1|t1) to obtain a predicted covariance matrix P(t2|t1) based on at least the system matrix A(t1) and a system noise covariance matrix R1(t1),
wherein the state space model is characterized through equations x(t+T)=A(t)x(t)+B(t)u(t)+w(t) and y(t)=c(x(t))+e(t), in which x(t) represents the state vector, u(t) represents an input vector, y(t) represents the output measurement vector, w(t) represents the model error vector, e(t) represents the measurement error vector, A(t) represents a system matrix describing dynamic modes of the system, B(t) represents an input gain matrix, c(x(t)) represents a measurement vector which is a function of the states of the system, t represents the time and T represents a sampling period.

### Claim 15 (depends on 1)

The method of claim 1, wherein the step of estimating the other cell interference Pother(t1) comprises:
obtaining a thermal noise floor estimate {circumflex over (N)}(t1) corresponding to the cell of interest as the thermal noise estimate {circumflex over (P)}N(t1); and
subtracting the thermal noise estimate {circumflex over (P)}N(t1) from the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) to obtain the other cell interference estimate {circumflex over (P)}other(t1).

### Claim 16

A radio network node of a wireless network, the radio network node corresponding to a cell of interest and being structured to determine an other cell interference Pother(t), the radio network node comprising:
a transceiver structured to transmit and receive wireless signals via one or more antennas from and to one or more cell terminals located within the cell of interest;
a communicator structured to communicate with other network nodes; and
an interference estimator structured to:
estimate an interference-and-noise sum Pother(t1+PN(t1) based at least on a load utilization probability estimate {circumflex over (p)}load(t0) and an interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0) applicable at a time t0 to obtain an interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) applicable at the time t1, and
estimate an other cell interference Pother(t1) based at least on the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) and a thermal noise estimate {circumflex over (P)}N(t1) to obtain an other cell interference estimate {circumflex over (P)}other(t1) applicable at the time t1,
wherein the interference-and-noise sum Pother(t)+PN(t) expresses a sum of undesired signals, other than an own cell load Pown(t), applicable at the time t, and the interference-and-noise sum estimate {circumflex over (P)}other(t)+{circumflex over (P)}N(t) being an estimate thereof, the own cell load Pown(t) expressing a sum of signals due to wireless activities in the cell of interest applicable at the time t,
wherein the other cell interference Pother(t) expresses a sum of interferences present in the cell of interest due to wireless activities applicable at the time t in one or more cells other than in the cell of interest, and the other cell interference estimate {circumflex over (P)}other(t) being an estimate thereof,
wherein a thermal noise PN(t) expresses a sum of undesired signals present in the cell of interest at the time t other than the own cell load Pown(t) and other than the other cell interference Pother(t) and the thermal noise estimate {circumflex over (P)}N(t) being an estimate thereof, and
wherein the interference-and-noise sum estimate {circumflex over (P)}other(t)+{circumflex over (P)}N(t) and the other cell interference estimate {circumflex over (P)}other(t1) are based on a state space model comprising at least a measurement matrix C(t1).

### Claim 17 (depends on 16)

The radio network node of claim 16, wherein the scheduler is a TD-scheduler with M processors, each processor being structured to estimate the other cell interference Pother(t1) only for each TD slot in a subset of TD slots, the subset being determined as a union of TD slots in which a same set of wireless terminals are scheduled.

### Claim 18 (depends on 16)

The radio network node of claim 16, wherein the interference estimator further structured to:
estimate a load utilization probability pload(t1) based at least on the load utilization probability estimate {circumflex over (p)}load(t0) and the interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0) applicable at the time t0 to obtain a utilization probability estimate {circumflex over (p)}load(t1) applicable at the time t1,
wherein the load utilization probability pload(t) expresses a relationships between radio resource grants scheduled to one or more cell terminals and radio resource grants used by the same cell terminals applicable at a time t, each cell terminal being a wireless terminal in the cell of interest, and the load utilization probability estimate {circumflex over (p)}load(t) being an estimate thereof, and
wherein the load utilization probability estimate {circumflex over (p)}load(t) is estimated based on the state space model comprising at least the measurement matrix C(t1).

### Claim 19 (depends on 18)

The radio network node of claim 18, further comprising a load meter structured to calculate a scheduled load factor Lown(t1−TD), obtain a used load factor Lown(t1−TD) and measure a load utilization
L
_
own
⁡
(
t
1
-
T
D
)
L
own
⁡
(
t
1
-
T
D
)
based on the scheduled and measured load factors,
wherein in order to estimate the load utilization probability pload(t1) and estimate the interference-and-noise sum Pother(t1)+PN(t1) the interference estimator is structured to obtain a utilization probability estimate {circumflex over (p)}load(t1) and obtain the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1),
wherein the scheduled load factor Lown(t−TD) expresses an amount of the radio resource grants scheduled to be used by the cell terminals for uplink transmissions at the time t, and
wherein the used load factor Lown(t−TD) expresses an amount of the scheduled radio resource grants used by the cell terminals for the uplink transmissions at the time t, and
wherein TD represents a delay between calculation of the scheduled load factor and a time the schedule takes effect on an air interface.

### Claim 20 (depends on 19)

The radio network node of claim 19, wherein the interference estimator is further structured to:
determine whether a HARQ process is active at the time t1,
obtain the load utilization probability estimate {circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) based on the measured load utilization
L
_
own
⁡
(
t
1
-
T
D
)
L
own
⁡
(
t
1
-
T
D
)
when it is determined that the HARQ process is active, and
obtain the load utilization probability estimate {circumflex over (p)}load(t1) from the previous load utilization probability estimate {circumflex over (p)}load(t0) and to obtain the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) from the previous interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0) when it is determined that the HARQ process is not active.

### Claim 21 (depends on 18)

The radio network node of claim 18, further comprising a wideband meter structured to measure a total wideband power yRTWP(t1),
wherein in order to estimate the load utilization probability {circumflex over (p)}load(t1) and estimate the interference-and-noise sum Pother(t1)+PN(t1), the interference estimator is structured to obtain the load utilization probability estimate {circumflex over (p)}load(t1) and obtain the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1).

### Claim 22 (depends on 21)

The radio network node of claim 21, wherein in order to estimate the load utilization probability pload(t1) and estimate the interference-and-noise sum Pother(t1)+PN(t1) the interference estimator is further structured to:
determine whether a HARQ process is active at the time t1,
obtain the load utilization probability estimate {circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) based on the measured total wideband power yRTWP(t1) when it is determined that the HARQ process is active at the time at the time t1, and
obtain the load utilization probability estimate {circumflex over (p)}load(t1) from the previous load utilization probability estimate {circumflex over (p)}load(t0) and to obtain the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) from the previous interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0) when it is determined that the HARQ process is not active at the time at the time t1.

### Claim 23 (depends on 21)

The radio network node of claim 21, wherein in order to obtain the interference-and-noise sum Pother(t1)+PN(t1), the interference estimator is further structured to:
determine a gain factor g(t1) based on the load utilization probability estimate {circumflex over (p)}load(t1) and the scheduled load factor Lown(t0),
model the measured total wideband power yRTWP(t1) as a combination of the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) weighted by the gain factor g(t1) and a measurement uncertainty eRTWP(t1), and
obtain the interference-and-noise sum estimate {circumflex over (P)}other(t1)+PN(t1) based on the measured total wideband power yRTWP(t1) and the model thereof.

### Claim 24 (depends on 23)

The radio network node of claim 23, wherein in order to determine the gain factor g(t1), the interference estimator is structured to:
determine a load factor bias ΔLown(t1), and
determine the gain factor g(t1) based on the based at least on the load utilization probability estimate {circumflex over (p)}load(t1) the scheduled load factor Lown(t0), and the load factor bias ΔLown(t1),
wherein the load factor bias ΔLown(t) expresses an error of the scheduled load factor Lown(t).

### Claim 25 (depends on 18)

The radio network node claim 18, wherein in order to estimate the load utilization probability pload(t1) and estimate the interference-and-noise sum Pother(t1)+PN(t1) the interference estimator is structured to:
model in a state vector x(t) of the state space model a first state x1(t)=pload (t) as a load utilization probability, a second state x2(t)=pother(t)+PN(t) as the interference-and-noise sum, a third state x3(t)=Δ Lown(t) as a load factor bias expressing an error of a scheduled load factor, and a fourth state x4(t)=x1(t−T) reflecting that the load utilization probability measurement is subject to a delay corresponding to the sampling period T,
model a measured total wideband power yRTWP(t) and a measured load utilization yloadUtilization(t) in an output measurement vector y(t) of the state space model,
obtain a predicted state vector {circumflex over (x)}(t1|t0) which includes therein first and second predicted states {circumflex over (x)}1(t1|t0), {circumflex over (x)}2(t1|t0) whose values are based on the load utilization probability estimate {circumflex over (p)}load(t0) and the interference-and-noise sum estimate {circumflex over (P)}other(t0)+{circumflex over (P)}N(t0),
update the predicted state vector {circumflex over (x)}(t1|t0) based on one or more measurements included in an output measurement vector y(t1) applicable at the time t1 to obtain an estimated state vector {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1), and
obtain first and second estimated states {circumflex over (x)}1(t1), {circumflex over (x)}2(t1) from the estimated state vector {circumflex over (x)}(t1) respectively as the load utilization probability estimate {circumflex over (x)}1(t1)={circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (x)}2(t1)={circumflex over (P)}other(t1)+{circumflex over (P)}N(t1),
wherein modeling errors and measurement errors are incorporated in the state space model as a model error vector w(t) and a measurement error vector e(t),
wherein the predicted state vector {circumflex over (x)}(t|t−T) denotes a prediction of the state vector x(t) based on information available up to a time t−T, and
wherein the estimated state vector {circumflex over (x)}(t|t)={circumflex over (x)}(t) denotes an estimate of the state vector x(t) based on information available up to the time t.

### Claim 26 (depends on 25)

The radio network node of claim 25, wherein in order to estimate the load utilization probability pload(t1) and estimate the interference-and-noise sum Pother(t1)+PN(t1) the interference estimator is further structured to:
determine whether a HARQ process is active at the time t1,
wherein the interference estimator is structured to model the states in the state vector x(t), model the outputs in the output measurement vector y(t), obtain the predicted state vector {circumflex over (x)}(t1|t0), and update the predicted state vector {circumflex over (x)}(t1|t0) only when it is determined that the HARQ process is active, and
wherein the interference estimator is structured to obtain the first and second estimated states {circumflex over (x)}1(t1), {circumflex over (x)}2(t1) from the previous the previous states {circumflex over (x)}1(t0), {circumflex over (x)}2(t0) when it is determined that the HARQ process is not active at the time at the time t1.

### Claim 27 (depends on 25)

The radio network node of claim 25, wherein in order to update the predicted state vector {circumflex over (x)}(t1|t0), the interference estimator is structured to:
model the measured total wideband power yRTWP(t1) applicable at the time t1 as
y
RTWP
⁡
(
t
1
)
=
x
2
⁡
(
t
1
)
1
-
L
own
⁡
(
t
1
-
T
D
)
⁢
x
1
⁡
(
t
1
)
+
x
3
⁡
(
t
1
)
+
e
RTWP
⁡
(
t
1
)
,
TD representing a delay between calculation of the schedule and a time the schedule takes effect on an air interface,
model the load utilization yloadUtilization(t1) applicable at the time t1 as yloadUtilization(t1)=x4(t1)+eloadUtilization(t1) obtain the measurement matrix C(t1) around the predicted state vector {circumflex over (x)}(t1|t0), the predicted state vector {circumflex over (x)}(t1|t0) including first, second, third, and fourth predicted states {circumflex over (x)}1(t1|t0), {circumflex over (x)}2(t1|t0), {circumflex over (x)}3(t1|t0), {circumflex over (x)}4(t1|t0) predicted based on data upto the time t0/,
obtain a Kalman gain matrix Kf(t1) based on at least the measurement matrix C(t1), the measurement error vector e(t1), and a predicted covariance matrix P(t1|t0) corresponding to the predicted state vector {circumflex over (x)}(t1|t0),
update the predicted state vector {circumflex over (x)}(t1|t0) based on at least the Kalman gain matrix Kf(t1), the output measurement vector y(t1), and the measurement vector c(x(t1)) to obtain the estimated state vector {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1), the estimated state vector {circumflex over (x)}(t1) including the first, second, third, and fourth estimated states {circumflex over (x)}1(t1), {circumflex over (x)}2(t1), {circumflex over (x)}3(t1), {circumflex over (x)}4(t1), and
update the predicted covariance matrix P(t1|t0) based on at least the Kalman gain matrix Kf(t1) and the measurement matrix C(t1) to obtain an updated covariance matrix P(t1|t1) corresponding to the estimated state vector {circumflex over (x)}(t1).

### Claim 28 (depends on 25)

The radio network node of claim 25, wherein in order to obtain the measurement matrix C(t1), the interference estimator is structured to:
determine whether a HARQ process is active at the time t1,
determine the measurement matrix C(t1) linearized around the predicted state vector {circumflex over (x)}(t1|t0) such that
C
⁡
(
t
)
=
∂
c
⁡
(
x
)
∂
x
⁢
❘
x
=
x
^
⁡
(
t
⁢
⁢
1
❘
t
⁢
⁢
0
)
when it is determined that the HARQ process is active at the time t1, and
set the measurement matrix C(t1) to a predetermined value when it is determined that the HARQ process is not active at the time t1,
wherein in order to obtain the Kalman gain matrix Kf(t1), the interference estimator is structured to determine Kf(t1)=P(t1|t0)CT(t1)(C(t1)P(t1|t0)CT(t1)+R2(t1))−1 in which CT(t) is a transpose of the measurement matrix C(t) and (R2(t)) is a measurement covariance matrix corresponding to the measurement error vector e(t),
wherein in order to update the predicted state vector {circumflex over (x)}(t1|t0) to obtain the estimated state vector {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1), the interference estimator is structured to determine {circumflex over (x)}(t1|t1)={circumflex over (x)}(t1|t0)+Kf(t1)(y(t1)−c({circumflex over (x)}(t1|t0))), and
wherein in order to update the predicted covariance matrix P(t1|t0) to obtain the updated covariance matrix P(t1|t1), the interference estimator is structured to determine P(t1|t1)=P(t1|t0)−Kf(t1)C(t1)P(t1|t0).

### Claim 29 (depends on 25)

The radio network node of claim 25, wherein in order to estimate the load utilization probability pload(t1) and estimate the interference-and-noise sum Pother(t1)+PN(t1), the interference estimator is further structured to:
project the estimated state vector {circumflex over (x)}(t1) based on at least the system matrix A(t1) to obtain a predicted state vector {circumflex over (x)}(t2|t1) which includes the first, second, third, and fourth predicted states {circumflex over (x)}1(t2|t1), {circumflex over (x)}2(t2|t1), {circumflex over (x)}3(t2|t1), {circumflex over (x)}4(t2|t1), in which the first and second predicted states {circumflex over (x)}1(t2|t1) and {circumflex over (x)}2(t2|t1) values are based on the load utilization probability estimate {circumflex over (p)}load(t1) and the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1), and
project the updated covariance matrix P(t1|t1) to obtain a predicted covariance matrix P(t2|t1) based on at least the system matrix A(t1) and a system noise covariance matrix R1(t1),
wherein the state space model is characterized through equations x(t+T)=A(t)x(t)+B(t)u(t)+w(t) and y(t)=c(x(t))+e(t), in which x(t) represents the state vector, u(t) represents an input vector, y(t) represents the output measurement vector, w(t) represents the model error vector, e(t) represents the measurement error vector, A(t) represents a system matrix describing dynamic modes of the system, B(t) represents an input gain matrix, c(x(t)) represents a measurement vector which is a function of the states of the system, t represents the time and T represents a sampling period.

### Claim 30 (depends on 16)

The radio network node of claim 16, further comprising a noise estimator structured to obtain a thermal noise floor estimate {circumflex over (N)}(t1) corresponding to the cell of interest as the thermal noise estimate {circumflex over (P)}N(t1),
wherein in order to estimate the other cell interference Pother(t1) the interference estimator is structured to subtract the thermal noise estimate {circumflex over (P)}N(t1) from the interference-and-noise sum estimate {circumflex over (P)}other(t1)+{circumflex over (P)}N(t1) to obtain the other cell interference estimate {circumflex over (P)}other(t1).

### Claim 31 (depends on 1)

A non-transitory computer-readable medium which has stored therein programming instructions, wherein when a computer executes the programming instructions, the computer executes a method performed in a radio network node of a wireless network for determining other cell interference, wherein the method is the method of claim 1.