GLS 554

ADJUSTMENT COMPUTATIONS

LECTURE 9:
LEAST SQUARE ADJUSTMENT (NON-LINEAR
FUNCTION- TRILATERATION)

Nurhanisah Hashim
Room: B612
H/p: 019-6684587
Centre of Studies for Surveying Science and Geomatics,
Faculty of Architecture, Planning and Surveying,
Universiti Teknologi MARA.
 AX = L + V Observation equation for linear function

A system of nonlinear equations that are linearized by a Taylor series approximation can be

written as (Eq 1)

Where the Jacobian matrix J contains the coefficients of the linearized observation equations. The individual
matrices in Equation (Eq 1) are:
The vector of least squares corrections in the equally weighted system of Equation (Eq 1) is given by:

𝑋 = 𝐽𝑇 𝐽 −1 (𝐽𝑇 𝐾)

The system of weighted equations is:

𝑋 = 𝐽𝑇 𝑊𝐽 −1
(𝐽𝑇 𝑊𝐾)

Where W is Weight matrix as discuss in previous topic.

Notice that the least squares solution of a nonlinear system of equations is similar to the linear case.

In fact, the only difference is the use of the Jacobian matrix (J) rather than the coefficient matrix (A) and the
use of the K matrix rather than the observation matrix, L.

Many authors use the same nomenclature for both the linear and nonlinear cases. In these cases, the
differences in the two systems of equations are stated implicitly.
ADJUSTMENT OF HORIZONTAL SURVEYS
(TRILATERATION): DISTANCE OBSERVATION EQUATION
In adjusting trilateration surveys using the parametric least squares method, observation equations are written that
relate the observed quantities and their inherent random errors to the most probable values for the x and y
coordinates (the parameters) of the stations involved.
Referring to Figure 1, the following distance equation can be written for any observation lij.

(Eq 2)

Observed
Distance ij
Residual of
Distance ij

Figure 1
Equation (Eq 2) is a non-linear function involving the unknown variables Xi, Yi, Xj, Yj that can be rewritten as:

(Eq 3)

Where,

(Eq 4)

A system of a non-linear equations such as Equation (Eq 4) can be linearized and solve using a first-order Taylor
series approximation. The linearized form of equation (Eq 4) is:

𝑥𝑖 − 𝑥𝑗 𝑦𝑖 − 𝑦𝑗 𝑥𝑗 − 𝑥𝑖 𝑦𝑗 − 𝑦𝑖
𝑑𝑥𝑖 + 𝑑𝑦𝑖 + 𝑑𝑥𝑗 + 𝑑𝑦𝑗 = 𝑘𝑙𝑖𝑗+𝑣𝑙
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝑖𝑗

Where,

Observed Computed
distance distance
How to linearized this function….

to this function?????

𝑥𝑖 − 𝑥𝑗 𝑦𝑖 − 𝑦𝑗 𝑥𝑗 − 𝑥𝑖 𝑦𝑗 − 𝑦𝑖
𝑑𝑥𝑖 + 𝑑𝑦𝑖 + 𝑑𝑥𝑗 + 𝑑𝑦𝑗 = 𝑘𝑙𝑖𝑗+𝑣𝑙
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝑖𝑗

Take a look at this, and try to

understand 1
2 2 2 2 2
𝑥𝑗 − 𝑥𝑖 + 𝑦𝑗 − 𝑦𝑖 = 𝑥𝑗 − 𝑥𝑖 + 𝑦𝑗 − 𝑦𝑖

1
𝜕𝐹 1 2 2 −2
= 𝑥 − 𝑥𝑖 + 𝑦𝑗 − 𝑦𝑖 2 𝑥𝑗 − 𝑥𝑖 −1 𝜕𝐹
𝜕𝑥𝑖 2 𝑗 =? ? ?
𝜕𝑦𝑖
− 𝑥𝑗 − 𝑥𝑖
= 𝜕𝐹
𝑥𝑗 − 𝑥𝑖
2
+ 𝑦𝑗 − 𝑦𝑖
2 =? ? ?
𝜕𝑥𝑗

𝜕𝐹
𝑥𝑖 − 𝑥𝑗 =? ? ?
= 𝜕𝑦𝑗
𝐼𝐽
 Example1: LSA Using Non-linear Function (Distance
Observation)
The distance between known point A, B and C to unknown point U were observed and recorded in table 1. Known
coordinates of point A, B and C, as well as the approximate coordinate of unknown point U, also given in the Table
1. (The coordinate of point U is approximate coordinate which scaled from a topographic map). Find the Most
Probable Value (MPV) of Coordinates point U.

Point X- cood (m) Y- cood (m)

A 865.40 4527.15
B 2432.55 2047.25
C 2865.22 27.15
U 6861.325 3727.596
Observed Distance (m)
AU 6049.00
BU 4736.83
CU 5446.49
 Step 1: Create the linearized observation equation for each line

𝑥𝑖 − 𝑥𝑗 𝑦𝑖 − 𝑦𝑗 𝑥𝑗 − 𝑥𝑖 𝑦𝑗 − 𝑦𝑖
For any line IJ, 𝑑𝑥𝑖 + 𝑑𝑦𝑖 + 𝑑𝑥𝑗 + 𝑑𝑦𝑗 = 𝑘𝑙𝑖𝑗+𝑣𝑙
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝑖𝑗

Thus, linearized observation equation for line AU, BU and CU,

𝑥𝐴 − 𝑥𝑈 𝑦𝐴 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐴 𝑦𝑈 − 𝑦𝐴
Line AU, 𝑑𝑥𝐴 + 𝑑𝑦𝐴 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐴𝑈 +𝑣𝑙
𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈

𝑥𝐵 − 𝑥𝑈 𝑦𝐵 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐵 𝑦𝑈 − 𝑦𝐵
Line BU, 𝑑𝑥𝐵 + 𝑑𝑦𝐵 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐵𝑈 +𝑣𝑙
𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈

𝑥𝐶 − 𝑥𝑈 𝑦𝐶 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐶 𝑦𝑈 − 𝑦𝐶 I J
Line CU, 𝑑𝑥𝐶 + 𝑑𝑦𝐶 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐶𝑈+𝑣𝑙 A U
𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈

B U
C U
 𝑥𝐴 − 𝑥𝑈 𝑦𝐴 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐴 𝑦𝑈 − 𝑦𝐴
Line AU, 𝑑𝑥𝐴 + 𝑑𝑦𝐴 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐴𝑈 +𝑣𝑙
𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈

𝑋 = 𝐽𝑇 𝐽 −1
(𝐽𝑇 𝐾)
𝑥𝐵 − 𝑥𝑈 𝑦𝐵 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐵 𝑦𝑈 − 𝑦𝐵
Line BU, 𝑑𝑥𝐵 + 𝑑𝑦𝐵 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐵𝑈 +𝑣𝑙
𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈

𝑥𝐶 − 𝑥𝑈 𝑦𝐶 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐶 𝑦𝑈 − 𝑦𝐶
Line CU, 𝑑𝑥𝐶 + 𝑑𝑦𝐶 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐶𝑈+𝑣𝑙
𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈

[J] [K]

In order to compute matrix J and K, following items need to be computed earlier:

Observed Computed
1. Computed Distance of line AU, BU and CU using the known and approximate coordinates distance distance
given in the Table 1.
 Step 2: Compute the distance between points using known and approximate coordinate

2 2
Distance of line AU, BU and CU can be computed using the equation, 𝑥𝑗 − 𝑥𝑖 + 𝑦𝑗 − 𝑦𝑖

Thus, the computed distance are as follows,

Line AU, 𝑥𝑈 − 𝑥𝐴 2 + 𝑦𝑈 − 𝑦𝐴 2
= 6861.325 − 865.40 2 + 3727.596 − 4527.15 2 = 6049.000

Line BU, 𝑥𝑈 − 𝑥𝐵 2 + 𝑦𝑈 − 𝑦𝐵 2 2 2
= 6861.325 − 2432.55 + 3727.596 − 2047.25 = 4736.836

Line CU, 𝑥𝑈 − 𝑥𝐶 2 + 𝑦𝑈 − 𝑦𝐶 2 = 6861.325 − 2865.22 2 + 3727.596 − 27.15 2 = 5446.297

 Step 3: Compute the J matrix (or A matrix)

The J matrix can be formed using the following steps.

1. Head each column with an unknown value

2. Create a row for every observation

3. Substitute in the appropriate coefficient corresponding to the column into each row

Line Unknown
IJ 𝜕𝑥𝑢 𝜕𝑦𝑢
𝜕𝐹 𝜕𝐹 𝑥𝐴 − 𝑥𝑈 𝑦𝐴 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐴 𝑦𝑈 − 𝑦𝐴
𝑑𝑥𝐴 + 𝑑𝑦𝐴 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐴𝑈 +𝑣𝑙
AU 𝜕𝑥𝑢 𝜕𝑦𝑢 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈

𝜕𝐹 𝜕𝐹 𝑥𝐵 − 𝑥𝑈 𝑦𝐵 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐵 𝑦𝑈 − 𝑦𝐵
BU 𝑑𝑥𝐵 + 𝑑𝑦𝐵 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐵𝑈 +𝑣𝑙
𝜕𝑥𝑢 𝜕𝑦𝑢 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈

𝜕𝐹 𝜕𝐹 𝑥𝐶 − 𝑥𝑈 𝑦𝐶 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐶 𝑦𝑈 − 𝑦𝐶
CU 𝑑𝑥𝐶 + 𝑑𝑦𝐶 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐶𝑈 +𝑣𝑙
𝜕𝑥𝑢 𝜕𝑦𝑢 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈

AU, BU and CU is the computed distance

Inserting all the appropriate value,

6861.325 − 865.40 3727.596 − 4527.15

6049.000 6049.000
6861.325 − 2432.55 3727.596 − 2047.25 0.9912 −0.1322
𝐽= = 0.9345 0.3547
4736.836 4736.836
0.7337 0.6794
6861.325 − 2865.22 3727.596 − 27.15
5446.297 5446.297

Step 4: Compute the K matrix (or L matrix)

Observed Computed
distance distance

Line AU, 6049.00 − 6049.000 0

Line BU, 𝐾 = 4736.83 − 4736.836 = −0.006
Line CU, 5446.49 − 5446.297 0.193
 Step 5: Compute the X matrix (Unknown/ correction value to the unknown)

𝜕𝑥𝑢 Where, 𝑋 = 𝐽𝑇 𝐽 −1 (𝐽𝑇 𝐾)

𝑋 = 𝑢𝑛𝑘𝑛𝑜𝑤𝑛 =
𝜕𝑦𝑢
−1
0.9912 −0.1322 0
0.9912 0.9345 0.7337 0.9961 0.9345 0.7337
𝑋 = 0.9345 0.3547 −0.006
−0.1322 0.3547 0.6794 −0.1328 0.3547 0.6794
0.7337 0.6794 0.193
0.6303 −0.7283 0.136
= ×
−0.7283 2.4948 0.129
−0.008 𝜕𝑥𝑢
= =
0.223 𝜕𝑦𝑢

Thus, the adjusted coordinate of point U:

𝑥𝑢 = 6861.325 + −0.008 = 6861.317

𝑦𝑢 = 3727.596 + 0.223 = 3727.819

Step 6: Iteration process

−0.008 𝜕𝑥𝑢 Considered too large, thus, adjustment have to undergo the iteration process, until
X= =
0.223 𝜕𝑦𝑢 the X value reached < 0.001 meter.
 1st Iteration – Step 2: Compute the distance between points using known and approximate coordinate

Repeat Step 2 until Step 5, using the same observed data for distance AU, BU, CU, known coordinate A, B and
C. For coordinate of Point U, use the adjusted coordinate already computed in Step 5.

Point X- cood (m) Y- cood (m)

A 865.40 4527.15
B 2432.55 2047.25
C 2865.22 27.15
U 6861.317 3727.819
Observed Distance (m)
AU 6019.00
BU 4736.83
CU 5446.49
 2 2
Distance of line AU, BU and CU can be computed using the equation, 𝑥𝑗 − 𝑥𝑖 + 𝑦𝑗 − 𝑦𝑖

Thus, the computed distance are as follows,

Line AU, 𝑥𝑈 − 𝑥𝐴 2 + 𝑦𝑈 − 𝑦𝐴 2
= 6861.317 − 865.40 2 + 3727.819 − 4527.15 2 = 6048.963

Line BU, 𝑥𝑈 − 𝑥𝐵 2 + 𝑦𝑈 − 𝑦𝐵 2 2 2
= 6861.317 − 2432.55 + 3727.819 − 2047.25 = 4736.907

Line CU, 𝑥𝑈 − 𝑥𝐶 2 + 𝑦𝑈 − 𝑦𝐶 2 = 6861.317 − 2865.22 2 + 3727.819 − 27.15 2 = 5446.443

 1st Iteration – Step 3: Compute the J matrix (or A matrix)

Line Unknown
IJ 𝜕𝑥𝑢 𝜕𝑦𝑢
𝜕𝐹 𝜕𝐹 𝑥𝐴 − 𝑥𝑈 𝑦𝐴 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐴 𝑦𝑈 − 𝑦𝐴
AU 𝑑𝑥𝐴 + 𝑑𝑦𝐴 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐴𝑈 +𝑣𝑙
𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈 0 𝐴𝑈
𝜕𝑥𝑢 𝜕𝑦𝑢
𝜕𝐹 𝜕𝐹 𝑥𝐵 − 𝑥𝑈 𝑦𝐵 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐵 𝑦𝑈 − 𝑦𝐵
BU 𝑑𝑥𝐵 + 𝑑𝑦𝐵 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐵𝑈 +𝑣𝑙
𝜕𝑥𝑢 𝜕𝑦𝑢 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈 0 𝐵𝑈

𝜕𝐹 𝜕𝐹 𝑥𝐶 − 𝑥𝑈 𝑦𝐶 − 𝑦𝑈 𝑥𝑈 − 𝑥𝐶 𝑦𝑈 − 𝑦𝐶
CU 𝑑𝑥𝐶 + 𝑑𝑦𝐶 + 𝑑𝑥𝑈 + 𝑑𝑦𝑈 = 𝑘𝑙𝐶𝑈 +𝑣𝑙
𝜕𝑥𝑢 𝜕𝑦𝑢 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈 0 𝐶𝑈

AU, BU and CU is the computed distance

Inserting all the appropriate value,

6861.317 − 865.40 3727.819 − 4527.15

6048.963 6048.963
6861.317 − 2432.55 3727.819 − 2047.25 0.9912 −0.1321
𝐽= = 0.9349 0.3548
4736.907 4736.907
0.7337 0.6794
6861.317 − 2865.22 3727.819 − 27.15
5446.443 5446.443
 1st Iteration - Step 4: Compute the K matrix (or L matrix)

Line AU, 6049.00 − 6048.963 0.037

Line BU, 𝐾 = 4736.83 − 4736.907 = −0.077
Observed Computed Line CU, 5446.49 − 5446.443 0.047
distance distance

1st Iteration - Step 5: Compute the X matrix (Unknown/ correction value to the unknown)

𝜕𝑥𝑢 Where, 𝑋 = 𝐽𝑇 𝐽 −1 (𝐽𝑇 𝐾)

𝑋 = 𝑢𝑛𝑘𝑛𝑜𝑤𝑛 =
𝜕𝑦𝑢
−1
0.9912 −0.1321 0.037
0.9912 0.9349 0.7337 0.9912 0.9349 0.7337
𝑋 = 0.9349 0.3548 −0.077
−0.1321 0.3548 0.6794 −0.1321 0.3548 0.6794
0.7337 0.6794 0.047
0.6303 −0.7286 −0.00083
= ×
−0.7286 2.4953 −0.00028
−0.00032 𝜕𝑥𝑢
= = Thus, the adjusted coordinate of point U:
−8.3 𝑋 10−5 𝜕𝑦𝑢
𝑥𝑢 = 6861.317 + −0.00032 = 6861.317𝑚
𝑦𝑢 = 3727.819 + (−8.3 𝑋 10−5 ) = 3728.819𝑚
 1st Iteration - Step 6: Iteration process

−0.00032 𝜕𝑥𝑢
𝑋 = =
−8.3 𝑋 10−5 𝜕𝑦𝑢

The iteration process will stop here. No further iteration process needed as the correction value
considered sufficiently small, which is smaller than 0.001m.

If the correction value, which is X, yields large value even after the first iteration process, thus second
iteration process or even third iteration process are needed. Repeat steps 2 to step 6 for the next
iteration process using the adjusted coordinate of unknown point, yields from the previous iteration
process.
Exercise

A linearized observation equation for least squares adjustment of distance between any two points such as I and J can be written as
follows;

𝑥𝑖 − 𝑥𝑗 𝑦𝑖 − 𝑦𝑗 𝑥𝑗 − 𝑥𝑖 𝑦𝑗 − 𝑦𝑖
𝑑𝑥𝑖 + 𝑑𝑦𝑖 + 𝑑𝑥𝑗 + 𝑑𝑦𝑗 = 𝑘𝑙𝑖𝑗 +𝑣𝑙
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝐼𝐽 0
𝑖𝑗
Thank you