Advanced driving algorithms, control strategies, and their optimization in self-driving vehicles in various scenarios are hotspots in current research; 4WS-4WD (four-wheel steering and four-wheel drive) is another hotspot in the study of new concept models; and the nonlinear dynamic characteristics of self-driving vehicles (AVs) are prominent in the fast cornering mode, which leads to a significant reduction in the accuracy and stability of trajectory tracking. Based on these research backgrounds, this paper proposes a control strategy optimization idea based on the 4WS4WD vehicle and its optimization model. The main content includes the establishment of a 3D vehicle model that takes into account vehicle load transfer and position change, and the establishment of a hierarchical control strategy based on the optimized NMPC and 4WS4WD models. The controller consists of two parts: an upper tracking controller based on the new vehicle model and NMPC, and a lower decoupled controller. The tracking control effect of the algorithmic control strategy based on the model and controller is validated in the high-speed serpentine motion mode and double-shift linear motion mode on the joint simulation platform of Car Sim and Simulink.

In dynamic environments, convolutional neural networks (CNNs) often produce image feature maps with significant redundancy due to external factors such as moving objects and occlusions. These feature maps are inadequate as precise image descriptors for similarity measurement, hindering loop closure detection. Addressing this issue, this paper proposes feature compression of convolutional neural network output. The approach is detailed as follows: (1) employing ResNet152 as the backbone feature-extraction network, a Siamese neural network is constructed to enhance the efficiency of feature extraction; (2) utilizing KL transformation to extract principal components from the backbone network’s output, thereby eliminating redundant information; (3) employing the compressed features as input for NetVLAD to construct a spatially informed feature descriptor for similarity measurement. Experimental results demonstrate that, on the New College dataset, the proposed improved method exhibits an approximately 9.98% enhancement in average accuracy compared to the original network. On the City Center dataset, there is an improvement of approximately 2.64%, with an overall increase of about 23.51% in time performance. These findings indicate that the enhanced ResNet152 performs better than the original network in environments with more moving objects and occlusions.