Markov prediction and hybrid genetic-slime mold algorithm with multi-strategy optimized longitudinal control for intelligent vehicles

Abstract

With the rapid development of autonomous driving technology, longitudinal control of intelligent vehicles has become a critical research focus. This study addresses proportional–integral–derivative (PID) parameter tuning challenges by proposing a fuzzy PID controller enhanced with Markov prediction and a hybrid genetic-slime mold algorithm with multi-strategy (MHGSMA). First, a vehicle longitudinal dynamics model is established, with a Markov chain model used to predict future speed changes, providing forward-looking data for control. The fuzzy PID mechanism dynamically adjusts parameters via fuzzy logic. Next, MHGSMA synergizes global genetic search with local slime mold optimization, augmented by good point set, dynamic reverse learning, and pinhole imaging strategies. Experimental results demonstrate that the MHGSMA-optimized fuzzy PID controller reduces the mean tracking error by 13.9% (4.0225E-03 vs. 4.6717E-03) and variance from 6.9445E-01 to 6.9376E-01 compared to conventional fuzzy PID control. MHGSMA also outperforms six benchmark algorithms on high-dimensional functions, achieving three-order-of-magnitude lower fitness (2.50E+07 vs. genetic algorithm (GA) 5.77E+09) in the F1 Bent Cigar test. In conclusion, this paper provides a novel and efficient control method for longitudinal control in intelligent vehicles, offering important applications in improving vehicle control accuracy and stability.