Optimization Strategy for Car Following and Lane Changing Models of CAV in Mixed Traffic Environments

A mixed traffic environment is an environment where different types of agents, for instance, Connected Autonomous Vehicles, Human Driven Vehicles, and pedestrians in the same traffic space. In reality, such a mixed traffic environment is the most common for Connected Autonomous Vehicles, so it is practical to study the trade-off between the safety and efficiency of Connected Autonomous Vehicles. The paper proposes an optimization strategy for car-following and lane-changing models of Connected Autonomous Vehicles in mixed-traffic environments. In this study, real-time data (e.g., acceleration, position, signal status, etc.) from CARLA's inbuilt sensors are utilised to dynamically adapt the vehicle's decision-making logic. Compared to existing offline optimisation methods, it can better adapt to the uncertainty in real road environments. In order to check the validity, we use Carla to set up a simulation environment and evaluate the behavior of autonomous vehicles. Furthermore, we collect data through multiple sensors, such as acceleration sensors, to accurately measure vehicle status. Ultimately, we gather the data from the sensors and analyze it by mathematical methods. Through this experiment, we find out that the lane change strategy avoids unnecessary lane changes and shows strong adaptability.