ACPSO: a novel PSO variant for 3D UAV path planning in multi-threat environments

Particle Swarm Optimization (PSO) has been widely applied in the fields of numerical optimization and UAV path planning, but it suffers from inherent limitations such as a tendency toward premature convergence, weak parameter adaptability, and insufficient population diversity. To address these issues, this paper proposes a novel PSO variant algorithm (ACPSO) that integrates the Actor-Critic framework with genetic strategies. First, a dynamic collaborative parameter optimization mechanism based on Actor-Critic is designed to break through the constraints of traditional fixed or single-parameter adjustment strategies. The Actor network outputs the optimal combination of core PSO parameters ( w , \(c_1\) , and \(c_2\) ) according to the real-time population state, while the Critic network evaluates the effectiveness of parameter strategies via a long-term cumulative reward mechanism, thereby achieving a dynamic balance between global exploration and local exploitation. Second, two-point crossover and adaptive mutation operations are incorporated into the iteration process to enhance population diversity, which effectively helps the algorithm escape local optima. Finally, for the UAV path planning task in multi-threat scenarios, a two-stage path smoothing strategy combining three-point neighborhood moving average filtering and cubic spline interpolation is proposed to ensure the smoothness and obstacle avoidance safety of the planned path. To verify the competitiveness of the proposed algorithm, comparative experiments were conducted on the CEC 2017 test suite, against the canonical PSO and seven other state-of-the-art PSO variants, as well as two championship algorithms. In addition, based on the innovatively designed two-stage path smoothing strategy, UAV path planning experiments were carried out under four scenarios with varying difficulty levels to verify the path planning capability of the ACPSO algorithm. The results demonstrate that the proposed algorithm demonstrates excellent performance in both numerical optimization and UAV path planning tasks. The source-code link for the ACPSO algorithm is: https://github.com/msz-deep/msz.git .