University of Twente Student Theses
Predicting car movement for autonomous driving through traffic using neural networks
Delfgaauw, Camilio (2020) Predicting car movement for autonomous driving through traffic using neural networks.
|
PDF
861kB |
| Abstract: | The technology of autonomous driving is becoming a more common technology. These autonomous cars have the ability to communicate with other autonomous cars. This makes it possible to use the data of the autonomous cars to make predictions on the traffic around the cars. Because of this this research tries to determine what deep learning method can be used in the prediction of vehicle movement. And from the method that can make predictions which method works best. The goal of the research is to answer the following research questions: Q1 Which deep learning method is best to use in predicting traffic movement for autonomous driving? Q2 What machine learning techniques can be used in the trajectory and speed prediction of traffic? Q3 Which method creates the smallest error in our simulated setup? To answer these question the literature study determined to use a CNN method where you feed the network spatial data through the use of images and an LSTM network that uses coordinate data from each vehicle individually to determine the next coordinate of the vehicle. The generate data than can be used in the training and testing of these deep learning techniques, SUMO was used. With the help of SUMO two data sets where created, one with a simplistic network to test the different techniques and one more realistic and complicated to determine which technique preforms best. The CNN was moderately accurate with an accuracy of around 77% after training for 150 epochs. But the CNN also produced a lot of false positives making the results diluted. The LSTM network preformed slightly worse with an average accuracy of 58%. But these results where not influenced by any false positive values, because of how the network works. The above mentioned values together with some more variables, the conclusion was made that the LSTM network is more suitable for the predicting of traffic movement for autonomous driving. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Business & IT BSc (56066) |
| Link to this item: | https://purl.utwente.nl/essays/82216 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page