An improved genetic algorithm-particle swarm optimization (GA-PSO) was proposed to address the limitations of conventional algorithms in trajectory planning for 7-DOF hybrid-joint manipulators, including slow convergence, low accuracy, and the challenge of balancing smoothness with obstacle avoidance. A kinematic model was constructed based on a modified Denavit-Hartenberg (D-H) parameter method, followed by an analysis of the Cartesian workspace. A multi-objective fitness function incorporating jerk, energy consumption, and collision penalty was formulated. The algorithm was enhanced by initializing the population via improved Latin hypercube sampling, along with an exponentially decreasing inertia weight and adaptive crossover-mutation probabilities. After determining the optimal parameters through orthogonal experiments, comparative simulations were conducted against the sparrow search algorithm (SSA) and the whale optimization algorithm (WOA) under both noise-free and 0-20% Gaussian noise conditions. Results show that in noise-free environments, the improved GA-PSO achieved a median comprehensive fitness of 300, only one-fourth that of SSA and one-seventh that of WOA, while generating smooth joint trajectories without abrupt changes. Under 20% noise intensity, its fitness variation rate remained at 4.85%, demonstrating significantly superior robustness. The proposed algorithm provides an effective solution for high-precision trajectory planning of multi-degree-of-freedom manipulators.