Intelligent connected vehicles (ICVs) can achieve efficient information exchange and collaboration with nearby vehicles with the implementation of deep integration of contemporary interactions and network infrastructures, thereby enhancing energy efficiency, comfort, and traffic efficiency during the driving process. However, the open communication environment is prone to serious cybersecurity threats, where attackers could exploit vulnerabilities to maliciously manipulate traffic information in the vehicle perception layer, severely jeopardising road safety. Additionally, urban road networks exhibit complex and variable geometric features, often comprising hybrid road forms such as straight and curved segments, and such discontinuous geometric features often induce jerky acceleration and deceleration. To tackle these challenges, we propose a straight-curved mixed road lattice hydrodynamic model simultaneously considering both jerk dynamics and malevolent cyber-attacks. First, according to the H-infinity norm as the basis for the transition function requirement and the Routh-Hurwitz criteria, the sufficient conditions for the stability of the new model are determined; moreover, when the requirements for stability are not fulfilled, while integrating the proportion of different geometric roads in hybrid scenarios, a geometry-based secure robust switcher is produced that takes into account the distinction between the estimation optimal and current flux. Results confirmed by Bode diagram examination that the proposed control approach works to minimize traffic oscillation induced by malevolent cyber-attacks and traffic jerk. Finally, we conducted a series of computational experiments by employing periodic boundary constraints. Results reveal that the cyber-attack intensity factor, the traffic jerk coefficient and the control gain term exert fundamentally distinct impacts on the stability of traffic flow: the latter are positive, while the remaining are negative. These findings align with the theoretical inquiry stated above.
Zhai et al. (2026) studied this question.