Introduction
Aircraft skin inspection is a critical process to ensure flight safety. The aircraft skin, being a primary structural component, undergoes repeated stress cycles during flights, leading to potential damage from environmental factors such as corrosion, lightning strikes, and hail. Over time, such damage can compromise aerodynamic performance and even cause structural failures, posing significant safety risks. Traditional inspection methods rely on manual visual checks or handheld non-destructive testing equipment, which are inefficient, prone to human error, and often fail to guarantee full coverage.
To address these challenges, autonomous mobile robots equipped with inspection devices have emerged as a promising solution. However, due to the complex geometry of aircraft, a single type of robot—whether unmanned aerial vehicles (UAVs), unmanned ground vehicles (UGVs), or wall-climbing robots—struggles to achieve complete coverage. UAVs offer flexibility in accessing elevated regions like the wings and tail but suffer from limited payload capacity, short battery life, and instability in confined spaces. UGVs, while more stable and capable of carrying multiple sensors, cannot reach high-altitude areas. Wall-climbing robots, though theoretically capable of full coverage, remain impractical due to inefficiency and the need for direct contact with the surface.
This paper introduces a novel cooperative coverage path planning (AG-CCPP) framework that leverages the complementary strengths of UAVs and UGVs to achieve efficient and comprehensive aircraft skin inspection. The proposed method integrates task allocation and path planning into a unified optimization model, ensuring minimal inspection time while adhering to operational constraints such as workspace limitations and battery endurance.
Problem Formulation
The AG-CCPP problem involves generating a set of viewpoints that fully cover the aircraft skin, assigning these viewpoints to UAVs and UGVs based on their capabilities, and planning collision-free paths for each robot. The key constraints include:
- Workspace Constraints: UAVs and UGVs have distinct operational height limits. UAVs cannot inspect low-altitude regions beneath the fuselage, while UGVs are restricted from accessing elevated areas.
- Endurance Constraints: Each robot has a limited operational time due to battery capacity.
- Motion Constraints: UAVs and UGVs move at different speeds, affecting path planning efficiency.
The optimization objective is to minimize the total inspection time, defined as the maximum time taken by either robot to complete its assigned path.
Methodology
- Viewpoint Classification and Reachability Matrix
The first step involves classifying viewpoints based on robot accessibility:
- Class 1: Viewpoints only reachable by UAVs (e.g., upper fuselage, wings).
- Class 2: Viewpoints only reachable by UGVs (e.g., lower fuselage).
- Class 3: Viewpoints accessible by both robots (e.g., mid-altitude regions).
A reachability matrix is constructed to encode which robot can access each viewpoint. This matrix guides the subsequent task allocation process.
- Collision-Free Time Cost Matrix
Since direct paths between viewpoints may be obstructed by the aircraft structure, collision-free paths are computed using sampling-based motion planning algorithms. A time cost matrix is then generated, storing the travel time between every pair of viewpoints for each robot. This matrix reduces redundant computations during path optimization.
- Multi-Elite Dual-Chromosome Genetic Algorithm
To solve the joint task allocation and path planning problem, a genetic algorithm (GA) is enhanced with the following features:
a) Dual-Chromosome Encoding
- Viewpoint Chromosome: Represents the sequence of viewpoints.
- Allocation Chromosome: Determines which robot is assigned each viewpoint.
This dual representation allows simultaneous optimization of both task allocation and path planning.
b) Greedy Allocation Strategy
Viewpoints accessible by both robots are assigned based on:
- Endurance Constraints: Prioritizing the robot with remaining battery life.
- Path Efficiency: Assigning viewpoints to the robot that can reach them faster.
c) Multi-Elite Population Strategy
To improve convergence speed and solution quality, the GA maintains three sub-populations:
- Elite Population: Retains the best-performing solutions.
- Elite + Greedy Mutation: Applies greedy mutations to elite solutions.
- Elite + Random Mutation: Introduces diversity through random mutations.
This approach reduces inefficient crossover operations and accelerates optimization.
Experimental Results
Simulation Setup
The proposed method was tested on a high-fidelity Boeing 737-300 model with 380 viewpoints. Key parameters included:
- UAV Speed: 0.5 m/s, endurance of 1800 seconds.
- UGV Speed: 0.3 m/s, endurance of 7200 seconds.
- GA Parameters: Population size = 300, iterations = 100, elite count = 2.
Performance Comparison
AG-CCPP was benchmarked against two state-of-the-art methods:
- Planner (Greedy GA): A multi-robot path planner that balances viewpoint distribution but ignores robot heterogeneity.
- Dec-PPCPP: A decentralized predator-prey inspired method that uses local greedy optimization.
Results showed that AG-CCPP:
- Achieved the shortest total inspection time (707.54s for UAV, 708.00s for UGV).
- Outperformed Planner and Dec-PPCPP in both coverage time and computational efficiency.
- Demonstrated balanced task allocation, with UAVs handling more viewpoints due to their higher mobility.
Conclusion
This paper presented AG-CCPP, a cooperative coverage path planning framework for UAV-UGV teams in aircraft skin inspection. By integrating task allocation and path planning into a unified optimization model and leveraging a multi-elite genetic algorithm with greedy heuristics, the method ensures efficient and complete coverage while respecting robot constraints.
Future work will address multi-robot collision avoidance, path smoothing, and handling dynamic uncertainties in real-world environments.
DOI: 10.19734/j.issn.1001-3695.2024.09.0336
Was this helpful?
0 / 0