In recent years, automated vehicles have attracted much attention all over the world. This paper focuses on the freeway-exiting position decision problem of automated vehicles (AVs). Specifically, the paper addresses the determination of the lane-changing initiation location in the process of exiting the freeway. The location of the freeway-exiting decision point has a significant impact on the safety and efficiency of automated vehicles. If the lane-changing location is too close to the off-ramp, the AV may not succeed in exiting and may even collide with other vehicles. If the decision point is too far from the off-ramp, the AV will enter into the slower lane too early, increasing the travel time. However, the freeway-exiting lane-changing position problem of AVs has not been investigated thoroughly in the existing literature. This paper proposes a freeway-exiting position decision model to find the optimal freeway-exiting decision position to balance the efficiency and safety in the freeway-existing process. Field data is collected to validate the proposed model, and simulations are also conducted to analyze the variations of the exiting success probability (ESP) and the optimal exiting decision (OED) position under various traffic conditions. The results show that the proposed model can predict the value of ESP with high performance (MAPE is less than 13%) and help an automated vehicle to generate an appropriate freeway-exiting decision point to ensure a high ESP without sacrificing efficiency. An AV can increase its ESP by decreasing or increasing its speed to meet more safe lane-changing gaps on the target lane, and the speed-decreasing method has a more significant effect than the speed-increasing method. The speed difference between the two adjacent lanes greatly influences ESP and the OED point, and maintaining the speed difference in an appropriate range can increase ESP.
In recent years, autonomous driving has become a hot topic and attracts worldwide attention (Zhou et al., 2020/02/01, Luo et al., 2019/11/01, Yu et al., 2019). Automated vehicles (AVs) are considered to have a huge potential to enhance traffic safety (Ma et al., 2017), ease traffic congestion (Sun, Zheng and Sun, 2020), improve traffic flow stability (Zhou and Ahn, 2019), and reduce traffic pollution (Li and Li, 2019). AVs have now been intensively tested on real-world roadway networks, such as Waymo (Nourinejad, Bahrami and Roorda, 2018), (Mobileye et al., 2020), and nuTonomy (Mattioli, 2018, Cui et al., 2018), and the industry and research community believe that AVs may develop rapidly in the future decades (Berger and Rumpe, 2014, Shladover, 2018).
Conclusions and future work
This paper focuses on how an AV generates an appropriate freeway-exiting decision that can ensure the AV exit the freeway successfully with less travel time. An Exiting Success Probability (ESP) model for AVs is first proposed. Based on the ESP model, a new model to determine the optimal exiting decision (OED) point is developed. The proposed ESP model is validated using field data collected in Chengdu, China. Numerical simulations further analyze the characteristics of the proposed models. The following main conclusions are drawn in the paper.
(1) According to the validation results based on the field data, the MAPE of the proposed model is not more than 13%. The results indicate that the model can predict the freeway-exiting success probability with acceptable accuracy and can be applied to the AV to generate an appropriate freeway-exiting decision point in the freeway-exiting process.