Passive Entry to the Car Using Machine Learning - Improving the User Experience

Abstract

The automatic unlocking of the car, also referred to as passive entry, is a feature
which simplifies the process of getting into the car. A current and future problem
with the passive entry is that it needs to unlock fast enough so that the user does
not need to pull the handle more than once. In this thesis it was investigated if
machine learning can be used in the passive entry process to improve the user
experience by making predictions about whether the user intends to use the car
or not. If the user intends to use the car it should unlock, if not, the vehicle
should remain locked and in sleep mode.

Today machine learning is being used within areas such as search engines, speech
recognition and presence detection in smart homes. It relies on the collection
of data that can be useful input to a machine learning model. In this thesis it
was investigated if it is possible to collect data and use machine learning as a
solution for the passive entry feature.

The purpose was to develop a proof of concept containing an application collecting
data and a machine learning model containing an algorithm which can
predict user intentions based on the data. It also aimed to investigate further
development and integration of this logic into Volvo Car Corporation's existing
application, Volvo On Call.

The training and testing of the machine learning model has shown a positive
trend. The accuracy of the predictions indicate that a solution using machine
learning should be investigated further.

The inner workings of machine learning are brought up in this thesis but the
overall process has been carried out from a user perspective.

Popular Abstract

Is it possible to use machine learning to determine the user’s intentions? More specifically, can it be used to determine if the user intends to use the car? To avoid a scenario where the user intends to use the car but it does not unlock as it should, in this thesis machine learning is investigated as a possible solution.

Today machine learning is being used within areas such as search engines, speech recognition and presence detection in smart homes. It relies on the collection of data that can be useful input to a machine learning model.

The thesis writer has developed a proof of concept to a possible future solution. The data collected on the user’s smartphone will be sent to the cloud where the machine learning model will be trained and then sent back to the user. Depending on the prediction of the model an unlock signal will be sent to the car where the system performing the final authentication will be triggered and the car will unlock.

By collecting data from different sensors on the users smartphone a prediction can be made about whether the user intends to use the car or if the user is moving the lawn. The improvement of the user experience is the elimination of the situations where the car is not unlocked when it should be.

In this thesis studies have shown a positive trend when collecting data from the user’s smartphone and using it as input to a machine learning model. The collected data was time, date, GPS coordinates and RSSI strength, the input data was mapped to the correct output data and then the model was trained. When the training was done the model was able to predict a value between 0 and 1 where 0 meant that the car should remain locked and 1 meant that the car should unlock.

• author: Kareliusson, Martin
• supervisor: Joakim Eriksson
• organization: Ergonomics and Aerosol Technology
• course: MAMM01 20191
• year: 2019
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: User Experience, Articial Intelligence, Machine Learning, TensorFlow, Keras, Passive Entry, Volvo On Call
• language: English
• id: 8989332
• date added to LUP: 2019-07-02 09:24:07
• date last changed: 2019-07-02 09:24:07