Firefly-Whale Optimization Algorithm: Navigating Robot Paths with Efficiency

By Dr Silpaja Chandrasekar, PhDReviewed by Susha Cheriyedath, M.Sc.Jan 10 2024

In a paper published in the journal Biomimetics, researchers introduced a novel approach, the firefly-whale optimization algorithm (FWOA), to navigate complex mobile robot environments efficiently. By combining the WOA and firefly algorithm (FA) into a hybrid model with multi-population and opposite-based learning, the FWOA demonstrates superior performance in mobile robot path planning (MRPP).

Through extensive benchmark testing against ten classical metaheuristic algorithms, the FWOA proved its mettle by showcasing remarkable convergence speed and exploration capabilities, establishing itself as a robust competitor among advanced metaheuristic methods.

Related Work

Previous studies have showcased a diverse array of path-planning strategies for mobile robots across various industries, drawing the attention of scholars globally. Eyuboglu, Marashian, Majer, and others introduced innovative approaches catering to different facets of autonomous mobile robot navigation, addressing challenges in static, dynamic, and unstructured environments.

Path planning, a pivotal aspect in numerous applications, aims to chart safe, obstacle-free routes from start to endpoint, considering factors like planning time, path smoothness, and overall convenience. The rise of swarm-based algorithms in MRPP has revolutionized solution accuracy, departing from conventional methods.

FWOA Enhancements Overview

The proposed FWOA section details a comprehensive hybrid algorithm derived from the classical WOA. It introduces three key enhancements: multi-populations, FA integration, and food perception implementation. This hybridization mirrors the division and cooperation mechanisms observed in social creatures. It divides the initial whale population into two distinct groups: the search population (SP) and the hunt population (HP). The SP focuses on exploration, swiftly navigating the search space to identify potential optimal regions. In contrast, the HP specializes in exploitation, honing in on these optimal areas identified by the SP.

During the search phase, SP whales navigate towards potential solutions based on their leader's position, whereas HP whales exploit the identified optimal regions, akin to the classic WOA. A subsequent phase, Encircling Prey, combines SP and HP into a single combined population (CP) to enhance computational efficiency while maximizing exploration and exploitation.

Introducing the concept of food perception augments the algorithm's sophistication, mirroring how spiders utilize sensing. This inclusion enhances exploration by guiding the population toward potential optimal solutions. A comparison mechanism maintains randomness while avoiding local optima, facilitating the swift discovery of the optimal value.

Integrating the FA supplements the algorithm's capabilities by simulating the flickering behavior of fireflies. FA adds a mechanism to optimize movement towards optimal solutions. Furthermore, incorporating opposition-based learning, which Tizhoosh proposed, enhances the algorithm's efficiency. It computes opposite solutions during the search process, improving solution estimation and converging towards the optimal solution.

The culmination of these enhancements in the FWOA framework offers a sophisticated and multifaceted approach to mobile robot path planning. By integrating multi-populations, FA, food perception, and opposition-based learning, the algorithm aims to bolster exploration, exploitation, and convergence capabilities, providing a comprehensive solution for optimizing mobile robot path planning.

Algorithm Performance Analysis

Researchers are diving into comprehensive algorithm testing across 23 widely utilized benchmark functions. These functions encompass uni-modal, multi-modal, and fixed-dimension multi-modal types, each emphasizing various performance aspects. The experimental setup involves 1000 iterations and 100 search agents, executed independently 30 times on each benchmark function.

Researchers conducted a comparative analysis against classic algorithms such as particle swarm optimization (PSO), salp swarm algorithm (SSA), WOA, grey wolf optimizer (GWO), and seeker optimization algorithm (STOA). The exploration and exploitation analyses shed light on FWOA's competitive performance. For uni-modal functions emphasizing benchmark exploitation, FWOA showcases superior results due to its adeptness in pinpointing optimal function values, attributed to its implemented food perception mechanism.

Moving on to multi-modal functions, which pose stringent demands on algorithm performance due to numerous local optima and escalating complexity with dimensionality, FWOA continues to exhibit strong competitiveness, particularly in fixed-dimension multi-modal functions. It outperforms other algorithms in most cases, showcasing its ability to find smaller best values and minimal maximum values compared to alternative algorithms like GWO and PSO, highlighting FWOA's significant research potential.

The standard deviation analysis underscores FWOA's robustness and stability, evident through its consistently more minor standard deviation across most cases. This stability owes itself to FWOA's multi-population mechanism, facilitating a balance between development and detection while ensuring stability and performance.

Finally, the convergence analysis through convergence graphs indicates FWOA's tendency to converge swiftly, primarily attributed to its search population mechanism discussed earlier. These comprehensive analyses collectively validate FWOA's efficacy in solving a diverse range of benchmark functions compared to established metaheuristic algorithms. The subsequent sections will further explore the algorithm's performance in practical problem-solving scenarios, validating its effectiveness through comparisons with well-known algorithms.

Conclusion

In summary, this study validates FWOA's strength in solving MRPP, comparing it against diverse intelligent algorithms. FWOA addresses traditional algorithm limitations, showcasing swift convergence, robust exploration, and vital optimization across 23 benchmark functions. Results demonstrate its competitiveness against metaheuristic algorithms like WOA and its effectiveness in practical problem-solving. FWOA's advancements emphasize its substantial advantages in tackling MRPP, hinting at promising applications in complex scenarios.