Multi-Strategy Improved Particle Swarm Optimization for China UAV Path Planning

In the field of autonomous flight for China UAV, path planning in complex three-dimensional environments remains a critical and challenging research focus. The core objective is to identify a safe and cost-effective trajectory that satisfies all flight constraints and avoids obstacles. Particle Swarm Optimization (PSO) has been widely adopted in China UAV path planning due to its simple convergence mechanism. However, when applied to high-dimensional, multi-extremum complex terrains, the standard PSO algorithm often suffers from population diversity loss and a tendency to become trapped in local optima. To address these issues, we propose a multi-strategy improved particle swarm optimization (MSPSO) algorithm tailored specifically for China UAV applications. Our algorithm integrates Tent map initialization, a nonlinear decreasing inertia weight strategy, and a Gaussian mutation mechanism to enhance population diversity, balance global exploration with local exploitation, and provide a robust escape mechanism from local optima. Through comprehensive benchmarking on CEC2005 test functions and extensive simulations in complex mountainous environments, we demonstrate that the MSPSO algorithm significantly outperforms standard PSO and other metaheuristic algorithms in terms of convergence accuracy, robustness, and the quality of planned flight paths for China UAV.

1. Introduction

The rapid advancement of China UAV technology has led to an increasing demand for autonomous flight capabilities in complex and dynamic environments. Path planning, as a core component of autonomous flight, determines the feasibility, safety, and efficiency of China UAV missions. The fundamental challenge lies in generating an optimal or near-optimal trajectory from a starting point to a target destination while avoiding obstacles, minimizing fuel consumption, and adhering to strict flight dynamics constraints. This problem becomes particularly complex in three-dimensional terrains characterized by multiple peaks, valleys, and threat zones.

Particle Swarm Optimization (PSO), inspired by the social behavior of bird flocking, has been extensively applied in China UAV path planning due to its simplicity, ease of implementation, and relatively low computational cost. In the standard PSO framework, each potential solution is represented as a particle that navigates the search space by learning from its own historical best position and the global best position discovered by the entire swarm. Despite its popularity, the standard PSO algorithm exhibits several critical limitations when applied to complex three-dimensional path planning for China UAV. The random initialization of particles often leads to uneven coverage of the search space, reducing the probability of discovering feasible initial paths. The fixed or linearly decreasing inertia weight fails to dynamically balance global exploration and local exploitation throughout the iterative process. Most importantly, the algorithm lacks an effective mechanism to escape from local optima, resulting in premature convergence and suboptimal path quality.

To overcome these limitations, researchers have proposed various improvements to the PSO algorithm. Some studies have introduced Voronoi diagram-based geometric partitioning to simplify the search space, but such methods require complex pre-processing and lack flexibility in unstructured terrains. Logistic chaotic mapping has been employed to improve population initialization; however, the Logistic map follows a Chebyshev distribution with non-uniform characteristics, leading to limited traversal uniformity. Artificial potential field methods have been integrated into PSO to enhance obstacle avoidance, yet they often suffer from local oscillations in dense obstacle environments and lack effective escape mechanisms from deep local optima.

Motivated by these challenges, we propose a Multi-Strategy Improved Particle Swarm Optimization (MSPSO) algorithm specifically designed for China UAV path planning. Our contributions are threefold: first, we replace the Logistic map with the Tent map for population initialization, leveraging its strict uniform distribution properties to eliminate initialization blind spots and ensure comprehensive coverage of the search space. Second, we design a nonlinear decreasing inertia weight strategy that dynamically adjusts the exploration-exploitation balance throughout the optimization process, enabling broad global search in early iterations and fine-grained local refinement in later stages. Third, we incorporate a Gaussian mutation mechanism that provides a forced escape capability when the population stagnates, effectively preventing premature convergence to local optima. Through extensive experimental validation, we demonstrate that our MSPSO algorithm significantly improves convergence accuracy, robustness, and the quality of planned flight paths for China UAV in complex three-dimensional environments.

2. China UAV Path Planning Model

The path planning problem for China UAV involves generating a feasible trajectory from a specified start point to a target destination while satisfying various performance constraints and minimizing a comprehensive cost function. In this section, we formally define the performance constraints and the mission cost function that constitute the optimization model.

2.1 UAV Performance Constraints

Due to terrain masking effects, flying at lower altitudes significantly reduces the probability of China UAV being detected by radar systems. Therefore, the flight altitude must be maintained within a safe range to avoid obstacles while minimizing detection risk. Let the i-th waypoint be denoted as (x_i, y_i, z_i), where z_i represents the current flight altitude of China UAV. The terrain masking constraint J_a is defined as:

$$J_a = \begin{cases} |z_i – (z_{max} + z_{min})/2| & z_{min} \leq z_i \leq z_{max} \\ \infty & \text{otherwise} \end{cases}$$

where z_{min} and z_{max} denote the minimum and maximum allowable flight altitudes, respectively. This constraint penalizes deviations from the optimal cruising altitude while strictly enforcing the altitude boundaries.

The turning angle constraint ensures that the flight path respects the maneuverability limits of China UAV. Let the angle formed by consecutive waypoints i-1, i, and i+1 be denoted as θ_i. The maximum allowable turning angle is θ_{max}. The turning angle θ_i is calculated as:

$$\theta_i = \arccos\left(\frac{(x_{i+1} – x_i, y_{i+1} – y_i)(x_i – x_{i-1}, y_i – y_{i-1})^T}{|(x_{i+1} – x_i, y_{i+1} – y_i)| \cdot |(x_i – x_{i-1}, y_i – y_{i-1})|}\right)$$

where (x_{i+1} – x_i, y_{i+1} – y_i) and (x_i – x_{i-1}, y_i – y_{i-1}) represent the direction vectors from waypoint i to i+1 and from waypoint i-1 to i, respectively. Let a_i be an indicator variable defined as:

$$a_i = \begin{cases} 0 & \theta_i \leq \theta_{max} \\ 1 & \theta_i > \theta_{max} \end{cases}$$

The turning angle constraint J_r is then expressed as:

$$J_r = \begin{cases} 0 & \sum_{i=1}^{N_\omega} a_i = 0 \\ \frac{1}{N_\omega} \sum_{i=1}^{N_\omega} a_i & \sum_{i=1}^{N_\omega} a_i > 0 \end{cases}$$

In the vertical plane, the pitch angle φ_i at each waypoint must remain within the allowable range between φ_{min} and φ_{max}. The pitch angle φ_i is defined as:

$$\phi_i = \arctan\left(\frac{z_i – z_{i-1}}{||(x_{i+1} – x_i, y_{i+1} – y_i)||}\right)$$

Let b_i be an indicator variable defined as:

$$b_i = \begin{cases} 0 & \phi_{min} < \phi_i < \phi_{max} \\ 1 & \text{otherwise} \end{cases}$$

The pitch angle constraint J_s is then expressed as:

$$J_s = \begin{cases} 0 & \sum_{i=1}^{N_\omega+1} b_i = 0 \\ \frac{1}{N_\omega+1} \sum_{i=1}^{N_\omega+1} b_i & \sum_{i=1}^{N_\omega+1} b_i > 0 \end{cases}$$

Table 1: China UAV Performance Constraint Parameters
Parameter Symbol Typical Value Range Description
Minimum altitude z_{min} 20 – 50 m Minimum safe flight altitude above terrain
Maximum altitude z_{max} 200 – 500 m Maximum allowable flight altitude
Maximum turning angle θ_{max} 45° – 60° Maximum turning angle between consecutive segments
Minimum pitch angle φ_{min} -30° – -15° Minimum allowable pitch angle (dive)
Maximum pitch angle φ_{max} 15° – 30° Maximum allowable pitch angle (climb)
Number of waypoints N_ω 10 – 50 Number of intermediate waypoints excluding start and end

2.2 Mission Cost Function

The mission cost function for China UAV path planning integrates multiple objectives, including path length, threat exposure, terrain masking, turning angle, and pitch angle compliance. The path length cost J_L is calculated as the sum of Euclidean distances between consecutive waypoints:

$$J_L = \sum_{i=1}^{N_\omega-1} \sqrt{(x_{i+1} – x_i)^2 + (y_{i+1} – y_i)^2 + (z_{i+1} – z_i)^2}$$

Shorter path lengths generally lead to reduced fuel consumption and lower exposure to threats, making this a primary optimization objective for China UAV missions.

The threat cost J_t quantifies the risk of China UAV being detected or engaged by radar systems and other threat sources. Let the center of the m-th threat source be located at (x_m^r, y_m^r, z_m^r), with danger level σ and coverage region C_k. The distance from China UAV to the threat center is D_i = \sqrt{(x_i – x_m^r)^2 + (y_i – y_m^r)^2 + (z_i – z_m^r)^2}. The threat value T_i is defined as:

$$T_i = \begin{cases} \sigma(D_i + D) & (x_i, y_i, z_i) \in C_k \\ 0 & \text{otherwise} \end{cases}$$

where D represents the physical dimensions of China UAV. The total threat cost is then:

$$J_t = \sum_{i=1}^{N_\omega} T_i$$

The comprehensive fitness function for China UAV path planning integrates all constraint and objective components through a weighted sum formulation:

$$J = k_1 J_L + k_2 J_t + k_3 J_a + k_4 J_r + k_5 J_s$$

where k_1, k_2, k_3, k_4, and k_5 are weighting coefficients that balance the relative importance of path length, threat cost, terrain masking constraint, turning angle constraint, and pitch angle constraint, respectively. By appropriately tuning these coefficients, the mission planner can prioritize different aspects of the flight mission according to operational requirements.

Table 2: Weighting Coefficients for China UAV Mission Cost Function
Coefficient Default Value Priority Effect
k_1 0.4 Path length minimization (fuel efficiency)
k_2 0.3 Threat avoidance (safety)
k_3 0.1 Terrain masking constraint compliance
k_4 0.1 Turning angle constraint compliance
k_5 0.1 Pitch angle constraint compliance

3. Multi-Strategy Improved Particle Swarm Optimization Algorithm

3.1 Standard PSO Algorithm

The standard PSO algorithm is a population-based metaheuristic inspired by the social behavior of organisms such as bird flocks and fish schools. In this framework, each candidate solution is represented as a particle with position vector x and velocity vector v. Throughout the optimization process, particles navigate the search space by updating their positions based on their own historical best position (pbest) and the global best position discovered by the entire swarm (gbest). The velocity and position update equations are:

$$v_i(t+1) = w v_i(t) + c_1 r_1 (pbest_i – x_i(t)) + c_2 r_2 (gbest – x_i(t))$$

$$x_i(t+1) = x_i(t) + v_i(t+1)$$

where v_i(t) and x_i(t) represent the velocity and position of particle i at time t, w is the inertia weight controlling the balance between global exploration and local exploitation, c_1 and c_2 are cognitive and social learning factors, and r_1 and r_2 are random numbers uniformly distributed in the interval [0, 1]. Despite its wide application in China UAV path planning, the standard PSO algorithm suffers from several inherent limitations. The random initialization strategy often results in uneven population distribution, reducing the algorithm’s ability to comprehensively explore the search space. The fixed inertia weight fails to adapt to the different requirements of exploration and exploitation at different stages of the optimization process. Most critically, the algorithm lacks an effective mechanism to escape from local optima, leading to premature convergence and suboptimal solution quality.

3.2 Tent Map Initialization Strategy

To address the issue of uneven population distribution in standard PSO, we introduce the Tent map initialization strategy. The Tent map exhibits excellent ergodicity and aperiodicity, enabling particles to be distributed more uniformly throughout the three-dimensional flight space for China UAV operations. This uniform distribution significantly improves the probability of discovering feasible initial paths in complex threat environments. The Tent map is defined as:

$$x_{t+1} = f(x_t) = \begin{cases} x_t / \alpha & x_t \in [0, \alpha) \\ (1 – x_t) / (1 – \alpha) & x_t \in [\alpha, 1] \end{cases}$$

where α ∈ (0, 1) is the chaos coefficient that determines the randomness and ergodicity of the generated sequence. Through extensive experimental analysis, we have found that α ≈ 0.7 provides near-ideal population distribution characteristics for China UAV path planning problems. The Tent map produces a strictly uniform distribution of particles across the search space, eliminating the initialization blind spots that commonly occur with Logistic mapping and pseudo-random initialization. This uniform initial coverage ensures that the algorithm starts with a diverse set of candidate solutions, enhancing its ability to explore the entire flight corridor and identify promising regions for subsequent refinement.

3.3 Nonlinear Decreasing Inertia Weight Strategy

China UAV path planning requires a delicate balance between global exploration for identifying promising flight corridors and local exploitation for refining trajectory smoothness and optimality. The fixed inertia weight in standard PSO cannot satisfy this dynamic requirement throughout the optimization process. We therefore propose a nonlinear decreasing inertia weight strategy that adaptively adjusts the exploration-exploitation balance. The inertia weight is calculated as:

$$w(t) = w_{max} – (w_{max} – w_{min}) (t / t_{max})^k$$

where w_{max} and w_{min} are the maximum and minimum inertia weight values, t is the current iteration number, t_{max} is the maximum number of iterations, and k is a control factor that determines the decay rate of the inertia weight. In the early stages of optimization, the relatively large inertia weight promotes broad global exploration, enabling particles to rapidly identify promising regions of the search space. As the optimization progresses, the inertia weight decreases nonlinearly, shifting the focus toward local exploitation for fine-grained adjustment of waypoint positions. For complex terrains with multiple peaks and dense obstacles, we recommend setting k in the range [1.2, 1.5] to maintain prolonged global exploration and prevent premature convergence. This nonlinear decay profile provides a smooth transition from exploration to exploitation, significantly improving the convergence characteristics of the algorithm for China UAV path planning problems.

3.4 Gaussian Mutation Strategy

In complex three-dimensional environments with multiple peaks and dense obstacles, the particle swarm is prone to premature convergence and stagnation at local optima. To overcome this challenge, we incorporate an adaptive Gaussian mutation strategy into the PSO framework. When the probability condition τ > 1 – (t/t_{max})^{1/2} is satisfied, a Gaussian perturbation is applied to the particle position:

$$x_i^{new} = x_i + N(0, \sigma)$$

where x_i is the current position of particle i, and N(0, σ) is Gaussian noise with mean zero and standard deviation σ. The standard deviation σ controls the intensity of the perturbation. If σ is too large, the algorithm degenerates into random search; if σ is too small, the perturbation is insufficient to help particles escape from deep local optima. Through systematic parameter tuning for China UAV path planning scenarios, we have determined that σ should be set adaptively based on the search space dimensions and the current state of convergence. The mutation probability increases as the algorithm progresses, providing a higher likelihood of escape when the population is more likely to have converged. This adaptive mechanism ensures that the Gaussian mutation is applied judiciously, maintaining a balance between exploration and exploitation without disrupting the convergence process.

3.5 Algorithm Framework and Parameter Configuration

The MSPSO algorithm integrates the three improvement strategies into a cohesive optimization framework specifically designed for China UAV path planning. The algorithm begins by initializing the particle population using the Tent map within the three-dimensional flight space. Each particle represents a complete flight path, with its dimensional values corresponding to the coordinates of key waypoints. During the iterative optimization process, the nonlinear decreasing inertia weight is updated according to Equation, followed by velocity and position updates. If the updated waypoint coordinates exceed the predefined map boundaries, they are clamped to the boundary values and the corresponding velocity components are set to zero, ensuring that all generated paths remain within the feasible flight corridor. To prevent premature convergence, the Gaussian mutation mechanism is activated for particles meeting the probabilistic condition. After each iteration, the fitness of each particle is evaluated using the comprehensive cost function, and the individual and global best positions are updated accordingly. The algorithm terminates when the maximum number of iterations is reached or when the convergence criterion is satisfied.

Table 3: MSPSO Algorithm Parameter Configuration for China UAV Path Planning
Parameter Symbol Value Rationale
Population size N 50 Balance between diversity and computational cost
Maximum iterations t_{max} 500 Sufficient for convergence in complex terrains
Cognitive learning factor c_1 2.0 Standard value for individual experience weighting
Social learning factor c_2 2.0 Standard value for swarm experience weighting
Maximum inertia weight w_{max} 0.9 Promotes global exploration in early iterations
Minimum inertia weight w_{min} 0.3 Enables local exploitation in late iterations
Decay control factor k 1.3 Balanced exploration-exploitation transition
Tent map chaos coefficient α 0.7 Optimal uniformity for population initialization
Gaussian mutation std σ 0.1·(x_{max}-x_{min}) Adaptive perturbation intensity

The computational complexity of the MSPSO algorithm is O(N·D·t_{max}), where N is the population size, D is the dimensionality of the search space (3·N_ω for three-dimensional path planning), and t_{max} is the maximum number of iterations. The additional computational overhead introduced by the Tent map initialization, nonlinear inertia weight calculation, and Gaussian mutation is minimal compared to the standard PSO algorithm, making MSPSO suitable for offline mission planning scenarios for China UAV where solution quality is prioritized over real-time computation.

4. Experimental Validation and Analysis

4.1 CEC2005 Benchmark Test Functions

To rigorously evaluate the performance of the MSPSO algorithm, we conducted comprehensive tests on six representative functions from the CEC2005 benchmark suite. These functions include three unimodal functions (F1, F2, F3) for assessing convergence speed and exploitation capability, and three multimodal functions (F9, F10, F11) with numerous local optima for evaluating global search ability and local optima escape capability. We compared the MSPSO algorithm against four baseline algorithms: standard PSO, Moth-Flame Optimization (MFO), Sine Cosine Algorithm (SCA), and Genetic Algorithm (GA). Each algorithm was run 30 independent times on each test function, and the average solution accuracy was recorded.

Table 4: CEC2005 Test Functions Used for Algorithm Evaluation
Function Type Range Global Minimum Characteristics
F1: Sphere Unimodal [-100, 100]^D 0 Continuous, convex, single global optimum
F2: Schwefel 2.22 Unimodal [-10, 10]^D 0 Continuous, non-separable, single optimum
F3: Schwefel 1.2 Unimodal [-100, 100]^D 0 Continuous, non-separable, rotated
F9: Rastrigin Multimodal [-5.12, 5.12]^D 0 Highly multimodal, many local optima
F10: Ackley Multimodal [-32, 32]^D 0 Multimodal with deep local minima
F11: Griewank Multimodal [-600, 600]^D 0 Multimodal with widespread local minima
Table 5: Average Solution Accuracy Comparison on CEC2005 Functions (30 runs)
Function MSPSO PSO MFO SCA GA
F1 0.00 2.41 3.02×10^3 1.11×10^1 2.21×10^4
F2 1.44×10^{-30} 4.10 3.47×10^1 2.04×10^{-2} 5.54×10^1
F3 0.00 2.02×10^2 1.88×10^4 8.84×10^3 5.36×10^4
F9 0.00 1.58×10^2 1.56×10^2 4.48×10^1 2.73×10^2
F10 4.44×10^{-16} 2.61 1.53×10^1 1.18×10^1 1.98×10^1
F11 0.00 1.06×10^{-1} 1.59×10^1 1.03 1.96×10^2

The experimental results presented in Table 5 clearly demonstrate the superior performance of the MSPSO algorithm across all test functions. For the unimodal functions F1, F2, and F3, the MSPSO algorithm achieves solution accuracies of 0.00 or near machine precision, significantly outperforming all comparison algorithms. This exceptional performance is attributed to the combined effect of the Tent map initialization, which provides a uniformly distributed starting population, and the nonlinear decreasing inertia weight, which enables fine-grained local refinement in later iterations. The standard PSO algorithm, by contrast, exhibits convergence stagnation and achieves relatively poor accuracy due to its fixed parameter configuration and lack of effective local optima escape mechanisms.

For the multimodal functions F9, F10, and F11, the advantages of the MSPSO algorithm are even more pronounced. These functions contain numerous local optima that typically trap standard optimization algorithms. The MSPSO algorithm achieves the global optimum for F9 and F11 with zero error, and approaches machine precision for F10 with an error of 4.44×10^{-16}. This remarkable performance is attributed to the Gaussian mutation strategy, which provides an effective escape mechanism when the population stagnates at local optima. The comparison algorithms, including PSO, MFO, SCA, and GA, all exhibit significant performance degradation on these multimodal functions, with solution accuracies several orders of magnitude worse than MSPSO. These results validate the effectiveness of our multi-strategy improvement approach for solving complex optimization problems relevant to China UAV path planning.

4.2 Path Planning Simulation in Complex 3D Terrain

To validate the practical applicability of the MSPSO algorithm for China UAV path planning, we conducted extensive simulations in a complex three-dimensional mountainous environment. The terrain model includes multiple conical obstacles representing mountain peaks and cylindrical obstacles representing no-fly zones. The simulation domain spans 250 meters in the x-direction and 200 meters in the y-direction, with variable terrain elevation. The start point is located at coordinates (10, 10) and the target destination at (230, 180). China UAV is required to navigate from the start to the target while avoiding all obstacles, maintaining flight altitude within safe bounds, and minimizing the comprehensive cost function defined in Equation.

We compared the path planning performance of the MSPSO algorithm against enhanced versions of GA, SCA, and MFO algorithms, as well as the standard PSO algorithm. All algorithms were configured with their optimal parameter settings for fair comparison. The MSPSO algorithm combined with cubic spline interpolation was used to generate smooth, flyable trajectories that satisfy the maneuverability constraints of China UAV.

Table 6: Parameter Settings for Comparison Algorithms in Path Planning Simulation
Algorithm Parameter Value
MSPSO c_1, c_2, w 2.0, 2.0, [0.3, 0.9]
PSO c_1, c_2, w 2.0, 2.0, [0.3, 0.9]
MFO c [0, 2]
SCA a 2.0
GA Cross, mutation 0.2, 0.1

The simulation results reveal significant differences in the quality of paths generated by the different algorithms. The GA-generated path exhibits notable迂回 (detours) in the horizontal plane and directly climbs over high mountain obstacles in three-dimensional space, resulting in substantial redundant climbing costs and the longest total path length. This behavior severely compromises the flight efficiency and safety of China UAV operations. The SCA and MFO algorithms, while performing better than GA, produce paths that navigate through narrow terrain saddles with insufficient obstacle clearance margins. This limitation stems from their restricted local exploitation capability, resulting in paths that closely follow terrain obstacle boundaries without adequate safety redundancy. Furthermore, the trajectory smoothness is suboptimal, with abrupt directional changes that may violate the turning angle constraints of China UAV.

In contrast, the MSPSO algorithm demonstrates exceptional adaptability to complex environments and superior global optimization performance. The planned path achieves an optimal balance between obstacle avoidance safety and path length efficiency. By strictly avoiding cylindrical obstacles and towering mountain peaks, the MSPSO-generated path selects the most reasonable angle to traverse through the valley region between mountains. This strategic path selection effectively avoids unnecessary altitude changes and extensive detours, minimizing the comprehensive three-dimensional path cost. The curvature variation along the MSPSO-generated path is the most gradual among all comparison algorithms, exhibiting superior smoothness that better satisfies the minimum turning radius and dynamic constraints of China UAV.

4.3 Convergence Analysis

We conducted a detailed convergence analysis of all algorithms on both the CEC2005 benchmark functions and the path planning problem. The convergence curves demonstrate that the MSPSO algorithm achieves the fastest convergence rate among all comparison algorithms. In the early iterations, the synergy between the Tent map-generated uniform population and the relatively large inertia weight enables particles to rapidly locate promising regions of the search space. This initial advantage is particularly important for China UAV path planning, where early identification of feasible flight corridors significantly reduces the overall computational cost.

As the optimization progresses, the nonlinear decreasing inertia weight smoothly transitions the search behavior from global exploration to local exploitation. In the later stages, the Gaussian mutation mechanism provides a forced escape capability that prevents the population from stagnating at local optima. This three-pronged strategy ensures that the MSPSO algorithm maintains a healthy balance between exploration and exploitation throughout the entire optimization process, resulting in superior convergence accuracy and robustness.

The comparison algorithms exhibit various limitations in their convergence behavior. The standard PSO algorithm shows rapid initial convergence but quickly stagnates at suboptimal solutions due to its lack of diversity maintenance and escape mechanisms. The MFO and SCA algorithms display oscillatory convergence patterns, reflecting their inherent instability in complex search spaces. The GA algorithm converges slowly due to its reliance on mutation and crossover operators that lack directional guidance from the population’s collective experience. These findings underscore the effectiveness of our multi-strategy improvement approach for addressing the convergence challenges in China UAV path planning.

5. Conclusion

In this research, we have proposed a Multi-Strategy Improved Particle Swarm Optimization (MSPSO) algorithm specifically designed for China UAV path planning in complex three-dimensional environments. The MSPSO algorithm integrates three complementary improvement strategies to overcome the fundamental limitations of standard PSO. The Tent map initialization strategy ensures uniform population distribution across the search space, eliminating initialization blind spots and enhancing the probability of discovering feasible initial paths. The nonlinear decreasing inertia weight strategy dynamically balances global exploration and local exploitation throughout the optimization process, enabling broad search coverage in early iterations and fine-grained refinement in later stages. The Gaussian mutation strategy provides a robust mechanism for escaping local optima when the population stagnates, preventing premature convergence and ensuring high-quality final solutions.

Comprehensive experimental validation on CEC2005 benchmark functions demonstrates that the MSPSO algorithm achieves superior convergence accuracy and robustness compared to standard PSO, MFO, SCA, and GA algorithms. On unimodal functions, the MSPSO algorithm achieves solution accuracies approaching machine precision, reflecting its exceptional exploitation capability. On multimodal functions with numerous local optima, the MSPSO algorithm achieves global optimal solutions where comparison algorithms consistently fail, validating the effectiveness of the Gaussian mutation mechanism for local optima escape.

Simulation results in complex three-dimensional mountainous environments further confirm the practical applicability of the MSPSO algorithm for China UAV path planning. The MSPSO algorithm, combined with cubic spline interpolation, generates smooth, flyable trajectories that effectively avoid obstacles while minimizing path length and satisfying all maneuverability constraints. Compared to paths generated by GA, SCA, MFO, and standard PSO algorithms, the MSPSO-generated paths exhibit superior obstacle clearance, reduced path length, and enhanced smoothness, demonstrating the algorithm’s ability to find optimal solutions in complex real-world scenarios.

While the MSPSO algorithm has shown remarkable performance in static offline path planning, several avenues for future research remain open. The current framework assumes a static environment with known obstacle locations and terrain characteristics. Extending the algorithm to handle dynamic environments with moving obstacles and changing threat conditions would enhance its applicability for real-time China UAV operations. Additionally, adapting the MSPSO algorithm for multi-UAV cooperative path planning scenarios, where multiple China UAV must coordinate their trajectories to achieve mission objectives while avoiding inter-vehicle conflicts, represents a promising research direction. Finally, integrating the MSPSO algorithm with deep reinforcement learning frameworks could potentially achieve both the global optimization capability of metaheuristics and the real-time responsiveness of learned policies, further advancing the state of the art in autonomous flight for China UAV.

Scroll to Top