The Busload Prediction Model of CANFD Based on Mathematical Statistics

2018 International Conference on Computer, Communication and Network Technology (CCNT 2018) ISBN: 978-1-60595-561-2

The Busload Prediction Model of CANFD

Based on Mathematical Statistics

Bo SUN

and Dong-yang MA

Jilin University, No. 2699 Qianjin Street, Chang Chun Jilin, China 130012

2

CHINA FAW Corporation LIMITED, No.1063 Chuangye Street, Chang Chun Jilin, China 130011

*_{Corresponding author}

Keywords:CANFD, FBFF, CBFF, Busload.

Abstract. A busload prediction models of four kind of frame formats has been established in the paper, the mathematical expectation of the busload for FBFF (FD Base Frame Format), CBFF (Classical Base Frame Format), FEFF(FD Extended Frame Format)and CEFF(Classical Extended Frame Format)is obtained based on the model in order to predict the busload of each message, it can be used for network design.

The comparative analysis shows that when the message length is 8 bytes or less, the transmission efficiency of the CBFF is slightly higher than that of the FBFF, while when the message length is between 9 up to 64 bytes, the transmission efficiency of the FBFF is far higher than that of CBFF. The communication bandwidth is greatly increased with flexible data rate, and the longer the message is, the larger the bandwidth will be.

The research results have been successfully applied to a variety of vehicle CANFD network design to predict the busload of each message, and application results show that the simulation accuracy is high and the efficiency of CANFD network design is improved effectively.

Preface

The development trend of low carbon, intelligent and electric vehicle makes the proportion of automotive electronics cost increasing year by year. The Development Status and the Development Trend Analysis of China's Automotive Electronics Industry in 2017 from China Industry Information Network[1] reported that the cost of automobile electronic vehicle accounted for more than 40% cost of vehicle in 2015, as shown in Figure 1.

Figure 1. The proportion of automobile electronic cost to the whole vehicle cost.

Compared with the traditional CAN network, the busload prediction of CANFD is more complex. Therefore, we establish a busload prediction model of CANFD, and carry out simulation research to improve the accuracy of busload forecasting.

Establish the Busload Prediction Model of CANFD

Establish the Busload Prediction Model of the FBFF

According to ISO 11898-1[2], FBFF consists of 7 fields. The field definition of the FBFF with data length of 20-64 bytes is shown in Figure 2.

Figure 2. The field definition of the FBFF with data length of 20-64 bytes.

The busload prediction model of the FBFF is established according to the following 6 steps. Step1. According to the field definition of the FBFF, the data stream of SOF, arbitration field, control field and data field is established and stored in the character array variable MESSAGE1.

The main program are described as follows.

MESSAGE1[0]='0'; //SOF=0

MESSAGE1[1]∼MESSAGE1[11]= random(2); //ID is a random number of 11 bits MESSAGE1[12]='0'; //RRS＝0

MESSAGE1[13]='0'; //IDE＝0 MESSAGE1[14]='1'; //FDF＝1 MESSAGE1[15]='0'; //res＝0 MESSAGE1[16]='1'; //BRS＝1 MESSAGE1[17]='0'; //ESI＝0

MESSAGE1[18]∼MESSAGE1[21]= DLC; //Converting DLC to len

Step2. MESSAGE1 is stuffed according to the fulfilling rules specified by CANFD protocol, the stuffed binary data stream has been formed and stored in the character array variable MESSAGE2.

Step3. CRC sequence calculation

Two CRC generator-polynomials are used for FBFF. CRC_17, shown in Formula (1) , is used for FBFF with a data field up to 16 bytes long and CRC_21, shown in Formula(2), is used for FBFF with a data field longer than 16 bytes.

CRC_17＝X17 + X16 +X14 + X13 + X11 + X6 + X4 + X3 + X1 + 1 (1)

CRC_21＝X21+ X20 +X13 + X11 + X7 + X4 + X3 + 1 (2)

The program flow chart for calculating CRC sequence is shown in Figure 3. Where M3 represents the dividers and the remainder of the division of binary polynomials. P is the divisor and D is quotient.

Add the CRC field, ACK field and EOF to the end of the MESSAGE2 to form the whole binary data stream of FBFF.

Step4. 1000 simulations are carried out for each FBFF in order to improve the accuracy of simulation, and mathematical expectation of busload is calculated according to the formula (3).

E = ∑ _× ＋

×

(3)

E — — mathematical expectation of busload of FBFF.

L — — the bit length of arbitration phase in FBFF

L — — the bit length of data phase in FBFF

B — — the communication rate of arbitration phase in FBFF (Bit/s)

B — — the communication rate of data phase in FBFF (Bit/s)

C — — transmission cycle of FBFF(s)

Start

End

Add 21"0"to the end of MESSAGE2 to form a divisor M3

Delete all 0 before the first 1 in M3 .

P= "1100000010100010011001"

Z = the length of M3 -the length of P

D = "0……0"（the length of D is Z）

the length of M3>the length of P?

Z = the length of M3 -the length of P

Tem1 = (1 shl z）；D = D xor Tem1

Tem1 = (P shl z)；M3 = M3 xor Tem1

Delete all 0 before the first 1 in M3 . Y

N

Figure 3. The program flow chart for calculating CRC sequence for CRC_21.

Establish the Busload Prediction Model of the CBFF , FEFF and CEFF

The method to establish the busload prediction model of the CBFF, FEFF and CEFF is very similar to that of EBFF. The mathematical expectation of busload of FEFF is also calculated according to the formula (3). The mathematical expectation of busload of CBFF/CEFF is calculated according to the formula (4).

E= ∑

×

(4)

Where:

E — — mathematical expectation of busload of CBFF/CEFF.

L — — the bit length in CBFF/CEFF

B — — the communication rate of CBFF/CEFF (Bit/s)

C — — transmission cycle of CBFF/CEFF (s)

Analyze Simulation Results of Busload Prediction Model of CANFD

Analyze Simulation Results of Busload of FBFF and CBFF

When the DLC is 8 bytes or less, 1 frame of CBFF or FBFF is OK. Because the CBFF has fewer control bits to ensure the reliable transmission of the message, the busload used for the CBFF is less than that for FBFF.

When the DLC is 9 up to 64, it can still be transmitted by 1 frame of FBFF, but it can’t be transmitted by 1 frame of CBFF. For example, the CBFF is used to send a data with 48 bytes. At least 1 frame of CBFF is need to set up a connection between send node and receive node, then this data is broken into at least 6 frames of CBFF. A total of 7 frames of CBFF with DLC of 8 bytes are required. The busload of 1 frame of CBFF is 4.54%, and the total busload is 31.78%. And if the FBFF is used to send a data with 48 bytes, 1 frame of FBFF is OK, the busload is only 18.37%. The details are shown in Table 1, as shown in Figure 4.

Table 1. The busload comparison between the CBFF and the FBFF (example).

data length(byte)

busload%

FBFF CBFF

0 2.5 1.81

1 2.78 2.15

2 3.11 2.49

3 3.45 2.83

4 3.78 3.17

5 4.11 3.52

6 4.44 3.86

7 4.78 4.20

8 5.11 4.54

12 6.42 12.25

16 7.74 13.14

20 9.11 16.31

24 10.43 18.16

32 13.08 22.7

48 18.37 31.78

64 23.67 40.86

Figure 4. The busload comparison between the CBFF and the FBFF (example).

Analyze Simulation Results of Busload of FBFF with the Flexible Data Rate

Run the prediction model of FBFF which transmission cycle(C) is 10ms and the communication rate of arbitration phase(B) is 250Kbit/s, and get the busload used by the FBFF for each Data Length Code(DLC) and each communication rate of data phase(B),the details are shown in Table 2, as shown in Figure 5.

Table 2. The busload used by the FBFF(C＝0.01s, B=250Kbit/s).

DLC

the busload used by the FBFF (%)

=250Kbit/s =500Kbit/s =1Mbit/s =2Mbit/s =5Mbit/s

0 2.5 1.76 1.39 1.2 1.09

1 2.78 1.90 1.46 1.24 1.10

2 3.11 2.06 1.54 1.28 1.12

3 3.45 2.23 1.62 1.32 1.14

4 3.78 2.40 1.71 1.36 1.15

5 4.11 2.56 1.79 1.40 1.17

6 4.44 2.73 1.87 1.44 1.19

7 4.78 2.90 1.96 1.49 1.20

8 5.11 3.06 2.04 1.53 1.22

12 6.42 3.72 2.37 1.69 1.29

16 7.74 4.38 2.70 1.86 1.35

20 9.11 5.06 3.04 2.03 1.42

24 10.43 5.72 3.37 2.19 1.49

32 13.08 7.05 4.03 2.52 1.62

48 18.37 9.69 5.35 3.18 1.88

64 23.67 12.34 6.68 3.85 2.15

Figure 5. The busload used by the FBFF(C＝0.01s, B=250Kbit/s).

Similarly, an example of FBFF with transmission cycle(C) of 10ms and and DLC of 64 bytes is shown in Table 3, the busload used by a FBFF not using flexible data rate can transmit 11.01frame of FBFF with communication rate of data phase(B) of 5Mbit/s.

Table 3. Busload comparison (example1).

FBFF not using

flexible data rate FBFF using flexible data rate

B=250Kbit/s B=500Kbit/s B=1Mbit/s B=2Mbit/s B=5Mbit/s

busload(%) 5.11 3.06 2.04 1.53 1.22

busload saved by flexible

data rate (%) 0 40.12 60.08 70.06 76.13

ratio of the FBFF number 1 1.67 2.50 3.34 4.19

Table 3. Busload comparison (example2).

FBFF not using

flexible data rate FBFF using flexible data rate

B=250Kbit/s B=500Kbit/s B=1Mbit/s B=2Mbit/s B=5Mbit/s

busload(%) 23.67 12.34 6.68 3.85 2.15

busload saved by flexible

data rate (%) 0 47.87 71.78 83.73 90.92

ratio of the FBFF number 1 1.92 3.54 6.15 11.01

In summary, the communication bandwidth is greatly increased with flexible data rate, and the longer the message is, the larger the bandwidth will be.

Conclusion

A busload prediction models of four kind of frame formats has been established in the paper, the mathematical expectation of the busload for FBFF (FD Base Frame Format), CBFF (Classical Base Frame Format),FEFF(FD Extended Frame Format)and CEFF(Classical Extended Frame Format) is obtained based on the model in order to predict the busload of each message, it can be used for network design.

The simulation results are as follows.

(1) When the DLC is 8 bytes or less, the transmission efficiency of the CBFF is slightly higher than that of the FBFF, while when the DLC is 9 up to 64 bytes, the transmission efficiency of the FBFF is far higher than that of the CBFF.

(2)The communication bandwidth is greatly increased with flexible data rate, and the longer the message is, the larger the bandwidth will be.

The research results have been successfully applied to a variety of vehicle CANFD network design to predict the busload of each message, and application results show that the simulation accuracy is high and the efficiency of CANFD network design is improved effectively.

References

[1]The report “the Development Status and the Development Trend Analysis of China's Automotive Electronics Industry in 2017” from China industry information network (http://www.chyxx.com/industry)