Autonomous Learning Intelligent Vehicles Engineering: ALIVE 1.0
Jihene Rezgui, Émile Gagné, Guillaume Blain · 2020
The increasing number of vehicles on roads brings more risks associated with vehicular travel. Nevertheless, with the massive attraction towards self-driving vehicles and the use of artificial intelligence, a trained physical Autonomous Vehicle (AV) is now a major part of transports future. This paper discusses the limitations of the related research based on autonomous vehicles; particularly those who are not taking into account the real-world physics. It also proposes an Autonomous Learning Intelligent Vehicles Engineering, called ALIVE to let each vehicle have additional information about its surroundings in order to get an extended perception of its environment. Moreover, ALIVE car sensors will gather in real-time the required data concerning the vehicles environment which are fused into a learning algorithm predicting the vehicle's response. We tested our algorithm through different mazes to evaluate its efficiency to avoid obstacles and its capacity to adapt to any type of terrain. This has been done to make ALIVE versatile, open source, low-cost and work in any environment. Preliminary results demonstrate the effectiveness of ALIVE in terms of obstacle avoidance and delay minimization. Besides, we hope that our project can be used by other researchers to test their artificial intelligence in the real world instead of keeping it in a simulation.