Improved bald eagle search algorithm with dimension learning-based hunting for autonomous vehicle including vision dynamics

The lateral deviations on account of the penetrations of road curvature and the parameters uncertainties are the main issues against the autonomous vehicles (AVs) controller to provide effective performance including less error, fast response, and small overshoot. In this regard, this paper suggests a new improvement for the bald eagle search (BES) algorithm by dimension learning-based hunting (DLH). The proposed DLH strategy enhances the exploration behavior of the original BES to tackle different issues such as slow convergence, trapping in local optima, and the loss of the diversity in early stage. The proposed strategy improves the learning of each eagle from its neighbors rather than the knowledge of all individuals in the population. The improved BES (I-BES) is dedicated to overcome the tuning issue of the model predictive control (MPC) for AVs including vision dynamics. Besides, frequency domain bounds are formulated based on Hermite-Biehler theorem to handle the parameters uncertainty issue due to the variation of the AV speed and the road curvature during the tuning of the MPC gains. The optimization process is performed to ameliorate the damping performance of the AV response particularly by decreasing the system steady-state error, settling time, and maximum overshoot. A newly developed multi-objective formula is utilized to achieve the diminishing of the performance criteria simultaneously. The proposed I-BES and BES are tested against various standard benchmark optimization test functions and different statistical tests to confirm that the I-BES can perform better than BES. Moreover, the proposed I-BES is confirmed with the default BES, neural network algorithm (NNA), and genetic algorithm (GA) in cases of AV controllers. Furthermore, the response of the inspired robust MPC based on the I-BES algorithm can tackle the lateral deviation due to the variations of the road curvature with small overshoot and short settling time less than 0.1745% and 0.05 s respectively better than the fuzzy logic controller. Various scenarios are carried out to confirm the effectiveness of the suggested technique against the road curvature variation and the AV parameters uncertainty. The results emphasize the superiority of the inspired robust MPC based on the I-BES algorithm to provide the best-damped response and stabilize the AV system against the parameters uncertainty and the road curvature variation compared with the other techniques.

Automated summary

This paper proposes an improved bald eagle search (BES) algorithm to address the lateral deviations in autonomous vehicles (AVs) controllers. By introducing a dimension learning-based hunting strategy, the exploration behavior of the original algorithm is enhanced and the system's response performance is improved through an optimization process.