Refine
Is part of the Bibliography
- no (1)
Has Fulltext
- yes (1)
Year of publication
- 2021 (1)
Document Type
- Bachelor Thesis (1)
Institute
Language
- English (1)
To enable modern vehicle access systems, algorithms for the localization of the car key are under continuous development. At BMW, two studies developing such algorithms were carried out recently. One optimized the hyperparameters of an existing k-nearest-neighbor and k-means algorithm while the other one realized the localization of the car key via an artificial neural network. However, as the algorithms of these studies were only optimized for one car model each and differing results are to be expected for different car models due to their various body forms, the results of the two algorithms cannot be compared directly. In this thesis, this limitation is removed and comparison across all car models is made possible, by the creation of a comparison framework.
The analysis of this comparison shows that the existing algorithm with optimized hyperparameters delivers a slightly better and more consistent result in terms of accuracy across the range of models. Other factors concerning the duration of the training processes, the integration of the algorithms onto the automotive electronic control unit and the influence of the recording process of the RSS values in and around the car were taken into account as well.