TL;DR
This paper introduces the Multi-strategy Improved Harris Hawk Optimization (MIHHO) algorithm, designed to overcome the limitations of the standard Harris Hawk Optimization (HHO) in robot path planning. By incorporating novel strategies for adaptive weighting, dimension learning, and position updates, MIHHO significantly enhances convergence speed, accuracy, and the ability to avoid local optima, leading to more efficient and robust path planning for mobile robots.
What Problem Does This Paper Address?
Autonomous navigation for mobile robots critically relies on effective path planning – finding an optimal, collision-free route from a starting point to a destination. While existing algorithms like Dijkstra and A-Star have their merits, they often struggle with issues such as slow convergence, low solution accuracy, and a tendency to get stuck in local optima, especially in complex or dynamic environments. The Harris Hawk Optimization (HHO) algorithm, a popular metaheuristic, also faces these challenges when applied to path planning. This paper aims to address these limitations by proposing an enhanced version of HHO, MIHHO, specifically tailored to improve path planning performance.
Methods
The authors propose a Multi-strategy Improved Harris Hawk Optimization (MIHHO) algorithm by integrating three key enhancements into the original HHO framework:
- Double Adaptive Weight Strategy: This strategy dynamically adjusts inertia weights to improve both the global search capability during exploration and the local search capability during exploitation, aiming to boost convergence accuracy and speed.
- Dimension Learning-Based Hunting (DLH) Search Strategy: This approach introduces dimensional learning to help balance global and local search, maintain population diversity, and thus enhance the algorithm's overall optimization ability, particularly in high-dimensional problems.
- Position Update Strategy Based on Dung Beetle Optimizer (DBO): This novel position update mechanism is incorporated to reduce the likelihood of the algorithm getting trapped in local optima, drawing inspiration from the navigation and rolling behaviors of dung beetles.
These strategies are integrated into the HHO framework, modifying its exploration and exploitation phases to achieve superior performance.
Datasets
The paper evaluates the proposed MIHHO algorithm using two types of benchmarks:
- Classical Benchmark Functions: Twelve standard test functions (F1-F12) are used to assess the algorithm's optimization capabilities, including unimodal, multimodal, and fixed-dimension multimodal functions.
- CEC2022 Test Functions: A more complex set of 12 test functions (F13-F24) are employed to further validate the algorithm's performance and robustness on challenging optimization problems.
- Simulated Robot Environments: For path planning evaluation, the authors create simulated environments using a raster method. These include simple and complex scenarios with varying obstacle densities and distributions, at two different map sizes (20x20 and 40x40).
Key Insights and Ideas
- Synergistic Improvements: The combination of multiple distinct strategies (adaptive weights, dimension learning, and DBO-inspired updates) leads to a synergistic effect, significantly outperforming individual improvements or the base HHO algorithm.
- Balancing Exploration and Exploitation: The adaptive weighting and DLH strategies are crucial for effectively balancing the exploration of new search spaces and the exploitation of promising regions, a common challenge in optimization algorithms.
- Robustness in Complex Environments: MIHHO demonstrates a remarkable ability to maintain performance and find efficient paths even in cluttered and complex environments, suggesting its practical utility for real-world robotics.
- Avoiding Local Optima is Key: The DBO-inspired position update strategy appears particularly effective in preventing the algorithm from getting stuck in suboptimal solutions, a critical factor for successful path planning.
Key Results
- Classical Functions: MIHHO consistently ranked first in performance across 12 classical benchmark functions, showing significant improvements in optimization ability, convergence speed, and stability. It achieved optimal values for several functions and demonstrated superior stability with much lower standard deviations compared to other algorithms.
- CEC2022 Functions: MIHHO showed strong performance on the CEC2022 test suite, often achieving the lowest mean and standard deviation, indicating high accuracy and stability.
- Path Planning: In simple environments, MIHHO reduced average path lengths by up to 14.11% compared to other algorithms. In complex environments, it achieved an average path length reduction of up to 14.45% compared to HHO, demonstrating its effectiveness in challenging scenarios. MIHHO also showed faster convergence and more stable path planning results.
Limitations and Future Work
- 2D Focus: The current study is limited to 2D motion scenes. The authors suggest future research should extend to 3D environments to broaden the algorithm's applicability.
- Simulation-Based: The experiments are primarily simulation-based. Future work should involve real-world robot deployments to validate the algorithm's practical application and performance in actual operating conditions.
Summary updated