Autonomous driving is a hot topic in the area of artificial intelligence and machine learning where numerous researches is being conducted to make driverless vehicles mainstream. A successful autonomous driving system should contain accurate results, interpretability (for safety and reliability), and cost-effective. The core of autonomous driving is a union of perception (conscious about surroundings- roads, obstacles, etc.) and decision making (making decisions to drive corresponding to the perceived environment). Problems with the current innovations, too complex model architecture to achieve the state-of-the-art results which makes it too expensive and hard to interpret. In this paper, we show that using simple models (vanilla UNet/FCN for perception and linear algebra techniques for decision making) which are easy to interpret and produce a cost-effective system, attains good results. Our idea serves for level 1 autonomous driving which can be further scaled up for higher levels.
Autonomous Car Driving Using Deep Learning
21.05.2021
2240197 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Deep Learning for Autonomous Driving
Springer Verlag | 2023
|Deep Learning Algorithm for Autonomous Driving Using GoogLeNet
British Library Conference Proceedings | 2017
|Deep Federated Learning for Autonomous Driving
IEEE | 2022
|Autonomous Driving with Deep Reinforcement Learning
SLUB | 2023
|