Development of lane-changing and overtaking function based on artificial intelligence algorithm

Abstract: With the continuous development of automobile electrification and intelligence, the driving scenarios of automobiles are becoming more and more diverse and complex, which has prompted the continuous innovation of automobiles from assisted driving to intelligent driving. With the introduction of artificial intelligence, the intelligent driving function of automobiles is becoming more and more practical, and is gradually developing towards the stage of freeing the driver's hands and replacing the human brain with the advanced driving assistance system to respond to complex driving scenarios in real time; the high-level complex scene intelligent driving function is based on the realization of the assisted driving function, and is developed in the direction of automatic driving operation in complex vehicle scenes based on the actual driving experience of the driver and combined with artificial intelligence algorithms. The development of the lane change and overtaking function based on artificial intelligence algorithms is introduced, that is, through the intelligent selection of lane change conditions, the vehicle automatically completes the lane change and overtaking process in the best way.

0 Introduction

In daily driving, vehicles often encounter a slow car ahead of them. According to driver habits, if the adjacent lane is in good condition and permitted by law, they usually change lanes to overtake. At present, the vehicle-mounted assisted intelligent driving system has included an automatic lane change function. Its detection of adjacent lanes is based on the prediction of the driving state of the adjacent lane vehicle (hereinafter referred to as the "adjacent vehicle") at the beginning of the lane change action. However, in many real situations, this function cannot accurately predict the reaction of the adjacent vehicle driver. For example,

For example, when a vehicle in this lane (hereinafter referred to as "this vehicle") has a tendency to change lanes (turns on the turn signal to prepare for changing lanes), the driver of the target vehicle may accelerate to block it or slow down to avoid it. These operations will affect the state conditions for this vehicle to change lanes.

This paper is based on the state jump of the basic lane-changing function'2), and uses the artificial intelligence algorithm to derive the reaction trend of the adjacent vehicle to predict its action, thereby optimizing the lane-changing calculation results of the vehicle and obtaining the optimal solution for the lane-changing speed change of the vehicle.

1 Functional principle design

When the driver considers whether to change lanes and overtake because the vehicle ahead (hereinafter referred to as the "front vehicle") in the same lane is driving slowly (assuming that changing lanes and overtaking does not increase the risk), the driver's common sense is that changing lanes can ensure that the vehicle can drive at the current speed, while continuing to drive in the lane requires slowing down. Therefore, the distance benefit from the front vehicle is the only criterion for determining whether to change lanes. Therefore, the calculation of whether the vehicle changes lanes and overtakes, and the optimal solution of the lane change point is to solve the distance benefit function, and the loss of distance benefit of following the vehicle relative to changing lanes is treated as one of the parameters.

Similarly, the calculation result of the distance benefit function is also used in the behavior logic of the adjacent vehicle based on artificial intelligence judgment. The difference in distance benefit between the target adjacent vehicle taking aggressive blocking action or conservative avoidance action in response to the lane-changing tendency of the vehicle is used as a parameter in the distance benefit function of the adjacent vehicle.

1.1 Distance-based Revenue Function

The evaluation indicators of the vehicle's distance-based benefit function include: distance benefits of different strategies, distance benefits corresponding to fuel consumption, lane change penalty coefficient, and safe distance between two vehicles. The distance benefit function cannot only consider the benefits of two different strategies. For example, if the target vehicle accelerates to block, its fuel consumption will be converted into the distance value obtained; in order to reach the theoretical optimal lane change point within a unit time, the vehicle needs to change speed. If it accelerates, its fuel consumption will also correspond to the distance value obtained; and this value is used as the distance benefit function of the vehicle.