Abstract
Road tracking (RT) can be considered as one of trendy research topics, as it is essential in self-driving car applications. In this paper, we are proposing two main contributions. First of all, we propose a new Deep Reinforcement Support vector machine Learning for Road Tracking (DRSL-RT) approach. Secondly, we combine the DRSL-RT with the novel Deep Reinforcement Learning for Road Tracking (DRL-RT) that was suggested in [1], to further enhance the RT performances. Databases from the SYNTHIA-SEQS-05 are employed in this work, where databases from four environments are considered. These are the rain, heavy-rain, fog and spring environments. The four environments are employed and evaluated separately and all together. In addition, various comparisons with other networks are constructed. Here, final results are significantly increased to 91.79%, 92.74%, 96.48%, 98.48% and 97.75% for rain, heavy-rain, fog, spring and all environments, respectively.
Original language | English |
---|---|
Article number | 040004 |
Journal | AIP Conference Proceedings |
Volume | 2839 |
Issue number | 1 |
Early online date | 29 Sept 2023 |
DOIs | |
Publication status | Published - 2023 |
Event | 2022 International Conference on Innovations in Science, Hybrid Materials, and Vibration Analysis, IC-ISHVA 2022 - Virtual, Online, India Duration: 16 Jul 2022 → 17 Jul 2022 |