The BP-VMC method leverages the learning and training capabilities of BP neural networks to achieve adaptive tuning of key parameters in Virtual Model Control (VMC), thereby enabling adaptive control of robotic walking motion. However, BP neural networks are prone to converging to local optima during training, making it difficult to attain the global optimal solution. To address this issue, this paper proposes a Particle Swarm Optimization-enhanced BP-VMC method (PSO-BP-VMC). By employing the PSO algorithm to pre-optimize the weight matrix of the BP neural network, an optimal initial weight matrix under given conditions is obtained, guiding the BP network to converge toward the global optimum. Simulation experiments demonstrate that the improved PSO-BP-VMC method significantly enhances tracking accuracy in both position and velocity, leading to markedly improved control performance.