The stability for CACC system with time delays and reconstitution information of vehicles for compensating delays based on Bi-LSTM

Abstract

The vehicle platoon using the cooperative adaptive cruise control (CACC) transmits information between vehicles via communication networks to increase the control performance. However, time delays are inevitable during the network transmission of information, which influence the stability of the CACC vehicle system. This paper proposes a method for compensating information affected by time delays based on a Bi-LSTM model. First, the third-order dynamics of the CACC vehicle systems are established, and the control strategies are proposed with the leading, preceding and following vehicles. The conditions of local stability and string stability for the CACC vehicle systems without time delays are derived based on the Routh-Hurwitz stability criterion and the frequency domain methods, which reveal the relationship between the model parameters and the controller parameters. For the CACC vehicle systems with time delays, the maximum time delays that ensure the local stability and string stability are achieved using the similar methods accordingly. However, the stability of the CACC vehicle systems is destroyed, when the time delay exceeds the maximum value. To deal with the impact of time delays, the bidirectional long short term memory (Bi-LSTM) model is adopted to predict and reconstitute the information affected by time delays. Furthermore, the relevant parameters are set and the real vehicle data is used for calculation and simulation. The simulation results confirm the local and string stability can be ensured, and further show the boundary of the maximum time delay may reach 0.45s for the CACC vehicle systems in this paper. In order to highlight superiority of Bi-LSTM, by comparing LSTM and KF with BiLSTM, the simulation results show Bi-LSTM has the highest correlation coefficient and the smallest root mean square error, which verify that Bi-LSTM reconstructing information affected by time delays is more effective than KF and LSTM.