You are on page 1of 19

End Semester Presentation

for
B. Tech. Project
on
GPS Based Travel Mode Detection
by
Homendra Kumar (20141036)
Shailesh Kumar (20141039)
Dheerendra Kumar Verma (20141017)
Shaurav Shahi (20141011)
Under the guidance of
Dr. Varun Singh
Associate Profesor

Department of Civil Engineering


Motilal Nehru National Institute of Technology Allahabad
Allahabad – 211004 (India)
01 May 2018
introduction
 Smartphones are becoming increasingly popular day-by-day.
 The use of smartphones, however, can be extended from the usual
telecommunication field to applications in other specialized fields
including transportation.
 Transportation planners, policymakers, state departments of
transportation, metropolitan planning organisations, industry
professionals and academic researchers use survey data to better
understand the current demand and performance of the transportation
infrastructure, and to plan in preparation for future growth.
 Sensors embedded in the smartphones like GPS, accelerometer and
gyroscope can collect data passively, which in turn can be processed to
infer the travel mode of the smartphone user.
 This project is focused toward investigating the use of GPS and
accelerometer sensor to detect mode of travel in Allahabad City.
OBJECTIVES
Followings are the objectives of the project:
 To develop the GIS database of collected data.
 To filter the collected data using filtering technique.
 To prepare the method to classify the travel mode
using ANN.
METHODOLOGY
1. Data collection using smart phones with the help of Next GIS
Logger app.
2. Development of GIS database using QGIS software for
visualization of the spatial data.
3. Filtering technique will be used to filter the noises from the
collected data in MATLAB to increase the accuracy of data.
4. ANN techniques will be used on filtering data on the basis of
acceleration and speed to classify the travel mode.
5. Validation of method will be done to ensure it’s applicability.
ROAD NETWORK OF
ALLAHABAD

DATA COLLECTION

GIS DATABASE OF SMART PHONE APP.


SPATIAL DATA (NEXT GIS LOGGER)
Q-GIS
SOFTWARE •VEHICLE SPEED
•ACCELEROMETER DATA
SHAPE FILE •GYROSCOPIC DATA
•GPS DATA
•LATITUDE
•LONGITUDE

FILTERING OF COLLECTED DATA

CLASSIFICATION OF TRANSPORT MODE USING


ANN TECHNIQUE

Fig.1 Flow Chart


DETAILS OF WORK CARRIED OUT
 Study Area:-
 Route is taken from Yamuna Gate of MNNIT Allahabad to Civil
Lines ( Beena CNG Pump ) via Stanley road.

Fig. 2 Adopted Route for Study shown over Google Maps (URL: https://maps.google.com)
DETAILS OF WORK CARRIED OUT (Contd..)

 Data is collected using Next GIS Logger App using


smart phones.
 Data collected by different vehicle is given below.

TRAVEL AUTO CAR BUS


MODE
NUMBER OF 33 27 25
TRIPS
NUMBER OF 20226 18797 14998
SAMPLE
POINTS
DETAILS OF WORK CARRIED OUT (Contd..)

 Collected data has been separated in MS-Excel using Text to


Column Technique.
 This sorted data file is used as input in Q-GIS as .csv file to
produce shape file.
 Filtering of noises from raw data has been carried out using a
developed program in MATLAB.
 Program uses the FIR Low Pass Filtering technique to filter the
data.
MATLAB Program for Filtering Data

1. clc;
2. clear all;
3. close all;
4. %% loading data
5. X=xlsread('car_data.xlsx');
6. time=X(:,1);
7. speed=X(:,2);
8. acc_x=X(:,3);
9. acc_y=X(:,4);
10.acc_z=X(:,5);
11.%% filter
12.Fs = 48000; % Sampling Frequency
13.N = 4; % Order
14.Fpass = 600; % Passband Frequency
15.Apass = 0.5; % Passband Ripple (dB)
16.% Construct an FDESIGN object and call its CHEBY1 method
MATLAB Program for Filtering Data (Contd..)
17.h = fdesign.lowpass('N,Fp,Ap', N, Fpass, Apass, Fs); 44. title('filtered acc_x vs time');
18.Hd = design(h, 'cheby1'); 45. %% acceleration in y
19.num=Hd.sosMatrix(1:3); 46. Z_accy=filter(num,din,acc_y);
20.din=Hd.sosMatrix(4:6); 47. subplot(4,2,5);
21.%% speed 48. plot(time,acc_y);
22.Z_speed=filter(num,din,speed); 49. xlabel('time');
23.subplot(4,2,1); 50. ylabel('acc_y');
24.plot(time,speed); 51. title('acc_y vs time');
25.xlabel('time'); 52. subplot(4,2,6);
26.ylabel('speed'); 53. plot(time,Z_accy);
27.title('speed vs time'); 54. xlabel('time');
28.subplot(4,2,2); 55. ylabel('acc_y');
29.plot(time,Z_speed);
56. title('filtered acc_y vs time');
57. %% acceleration in z
30.xlabel('time');
58. Z_accz=filter(num,din,acc_z);
31.ylabel('speed');
59. subplot(4,2,7);
32.title('filtered speed vs time');
60. plot(time,acc_z);
33.%% acceleration in x
61. xlabel('time');
34.Z_accx=filter(num,din,acc_x);
62. ylabel('acc_z');
35.subplot(4,2,3);
63. title('acc_z vs time');
36.plot(time,acc_x); 64. subplot(4,2,8);
37.xlabel('time'); 65. plot(time,Z_accz);
38.ylabel('acc_x'); 66. xlabel('time');
39.title('acc_x vs time'); 67. ylabel('acc_z');
40.subplot(4,2,4); 68. title('filtered acc_z vs time');
41.plot(time,Z_accx); 69. end
42.xlabel('time');
43.ylabel('acc_x');
results
 Visualization of collected data by Q-GIS software as a shape
file are given below.

MNNIT
ALLAHABAD

BUS - YELLOW
AUTO - RED
CAR - BLUE
SHAPE FILE OF ALL DATA
FIR low pass filter output

Fig. 3 Speed vs time

Fig. 4 Accelerationx vs time


FIR low pass filter output (Contd..)

Fig. 5 Accelerationy vs time

Fig. 6 Accelerationz vs time


Artificial Neural Network (ANN) Model

The training of three layered feed-forward neural network with varying


number of neurons in hidden layer was performed using MATLAB neural net
toolbox called (nntool).

Fig. 7 FFNN architecture within MATLAB interface for 3-3-1.


Artificial Neural Network (ANN) Model (Contd..)

Fig 8. Levenberg-Marquardt training Model in MATLAB

Levenberg–Marquardt algorithm (LMA) is used to solve non-linear least squares


problems.
 The results for the accuracy of collected dataset are given below.

Table 1: Accuracy Assessment of the dataset


S.NO MODE OF TRANSPORTATION AVERAGE ACCURACY PER MODE

1. AUTO 81.48

2. CAR 84.74

3. BUS 78.35

Table 1 shows the accuracy achieved for each individual mode of transportation.
For car highest overall mode detection accuracy is 84.74 percent. These results
are important since they demonstrate that bus trips are suffering from the
highest number of false classifications as 21.65 percent.
CONCLUSIO
NS
 Research study for travel mode detection using smart phone has not been
carried out in prevailing roadway and traffic condition in India.
 It is found that longer monitoring duration and higher sampling frequency
result in higher travel mode detection rates.
 Proposed methodology can be suitably used to detect mode with
satisfactory level of accuracy.
 Quantum of data set effects the accuracy of mode detection.
REFERENCES
1. Xiao, g.; Juan, Z.; Zhang, C. Travel mode detection based on GPS track data and bayesian networks.
Comput. Environ. Urban syst. 2015.
2. Linlin wu, biao yang * and peng jing travel mode detection based on GPS raw data collected by
smartphones: A systematic review of he existing methodologiesc MDPI 2016.
3. H .Gong,c. Chenb, E. Bialostozkya, C. T. Lawson A GPS/GIS method for travel mode detection in New
york city computers environment and urban system march 2012.
4. Sun, Z., & Ban, X. J. (2013). Vehicle classification using GPS data. Transportation Research Part C:
Emerging Technologies, 37, 102-117.
5. Tao feng, harry J.P. Timmermans transportation mode recognition using GPS and accelerometer data,
trasportation reserch 2013.
6. Beale, M. H., Hagan, M. T., & Demuth, H. B. (2010). Matlab Neural networks toolbox
user’s guide; Mathwork.
7. Liao, K. P., & Fildes, R. (2005). The accuracy of a procedural approach to specifying
feedforward neural networks for forecasting. Computers & Operations Research, 32(8),
2151-2169.
8. Reddy, S., Mun, M., Burke, J., Estrin, D., Hansen, M., Srivastava, M., 2010. Using mobile
phones to determine transportation modes.
9. Cooper, A.R., Page, A.S., Wheeler, B.W., Griew, P., Davis, L., Hillsdon, M., Jago, R., 2010.
Mapping the walk to school using accelerometry combined with a Global positioning
system. American Journal of Preventive Medicine 38 (2), 178–183.
THANKING YOU

You might also like