💻 computer science

CICDWOA: A Collective Cognitive Sharing Whale Optimization Algorithm with Cauchy Inverse Cumulative Distribution for 2D/3D Path Planning and Engineering Design Problems

This paper proposes CICDWOA, an enhanced Whale Optimization Algorithm incorporating collective cognitive sharing, Cauchy inverse cumulative distribution, and hybrid mutation strategies to achieve superior convergence, accuracy, and robustness in solving 2D/3D path planning and complex engineering design problems.

Original authors: Junhao Wei, Yanxiao Li, Seyedali Mirjalili, Dexing Yao, Yifu Zhao, Haochen Li, Xudong Ye, Zikun Li, Qingyang Xu, Baili Lu, Ngai Cheong, Dengcheng Yang, Sio-Kei Im, Yapeng Wang, Xu Yang

Published 2026-03-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Junhao Wei, Yanxiao Li, Seyedali Mirjalili, Dexing Yao, Yifu Zhao, Haochen Li, Xudong Ye, Zikun Li, Qingyang Xu, Baili Lu, Ngai Cheong, Dengcheng Yang, Sio-Kei Im, Yapeng Wang, Xu Yang

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to find the absolute best spot to set up a campsite in a massive, foggy, and mountainous forest. You want the view to be perfect, the ground to be flat, and you need to avoid the bears (obstacles). But you can't see the whole forest at once, and you only have a limited amount of time.

This is essentially what optimization is: finding the best solution to a complex problem.

For years, scientists have used a computer program called the Whale Optimization Algorithm (WOA) to solve these problems. It's inspired by how humpback whales hunt. The whales swim around, spot a fish (the solution), and then swim in a shrinking circle or a spiral to catch it.

The Problem:
The original "Whale" program is good, but it has a few flaws:

It gets stuck easily: Sometimes the whales get trapped in a small pond (a local optimum) thinking it's the whole ocean, missing the giant fish nearby.
It's a bit random: The whales sometimes swim in circles without a clear plan, wasting time.
They don't talk enough: In the original version, each whale mostly acts on its own or just follows the single "leader" whale, ignoring what the rest of the group is doing.

The Solution: CICDWOA
The authors of this paper created a super-charged version called CICDWOA. Think of it as upgrading a standard fishing boat into a high-tech, AI-driven fishing fleet. Here is how they improved it, using simple analogies:

1. The "Perfect Start" (Good Nodes Set)

Old Way: Imagine dropping 30 people into a forest randomly. By chance, they might all land in the same small clearing, leaving the rest of the forest unexplored.
New Way (CICDWOA): Before the search even begins, the algorithm uses a special mathematical trick (Good Nodes Set) to place the explorers evenly across the entire forest. No one is crowded together; everyone covers a different part of the map immediately. This ensures they don't miss any hidden valleys right from the start.

2. The "Group Chat" (Collective Cognitive Sharing)

Old Way: In the original algorithm, a whale might just randomly pick one other whale to follow. If that random whale is lost, the first whale gets lost too.
New Way (CICDWOA): Now, every whale has a "Group Chat." Instead of just following one leader, they look at the average position of the whole group and the best position found so far.
- Analogy: Imagine you are lost in a mall. Instead of blindly following one stranger, you look at where the crowd is generally moving and where the person with the map is. You combine that info to make a smarter decision. This stops the group from getting scattered or stuck in a dead end.

3. The "Spiral Dance" (Adaptive Exponential Spiral)

Old Way: The whales used to swim in a very predictable, straight-line spiral toward the target. It was efficient but rigid.
New Way (CICDWOA): The whales now dance in a more complex, wavy spiral. They can speed up, slow down, and change the tightness of their spiral based on how close they are to the goal. It's like a dancer who knows when to do a wide, sweeping move to check the room, and when to do a tight, precise spin to grab the object.

4. The "Magic Jump" (Cauchy Inverse Cumulative Distribution)

Old Way: If the whales got stuck in a small valley (a local trap), they had a hard time jumping out because their movements were too small.
New Way (CICDWOA): The algorithm introduces a "Magic Jump" mechanism. Most of the time, the whales take small, careful steps. But occasionally, the math allows them to take a giant, random leap across the forest.
- Analogy: Imagine you are walking up a hill and hit a dead end. Usually, you'd just turn around. But with this new method, you might suddenly teleport to the other side of the mountain to see if there's a better path there. This helps them escape traps they would otherwise be stuck in forever.

5. The "Smart Teamwork" (Hybrid Mutation)

Old Way: Once the whales got close to the fish, they would just keep swimming in the same tight circle, sometimes missing the final catch.
New Way (CICDWOA): Near the end, the whales use a mix of two strategies (Gaussian and Cauchy). They do a mix of tiny, precise adjustments (like a surgeon) and occasional big jumps (like a trapeze artist). This ensures they don't just get close to the answer, but find the exact perfect answer.

Did it work?

The authors tested this new "Super Whale" on three types of challenges:

Math Puzzles: It solved 23 different complex math problems faster and more accurately than almost any other algorithm.
Robot Path Planning: They asked a robot to navigate a maze with obstacles. The Super Whale found the shortest, smoothest path without crashing, even in very crowded mazes where other robots got stuck.
Real-World Engineering: They used it to design things like aircraft wings, car brakes, and gas pipelines. In almost every case, the Super Whale found a design that was lighter, cheaper, and stronger than what other methods produced.

The Bottom Line

The CICDWOA is like taking a smart, solitary whale and turning it into a highly coordinated, communicative, and adventurous super-team. By making sure they start in good spots, talk to each other, dance in smarter patterns, and occasionally take giant leaps, they can solve the world's hardest engineering and planning problems much better than before.

1. Problem Statement

The Whale Optimization Algorithm (WOA) is a popular metaheuristic algorithm inspired by the bubble-net hunting behavior of humpback whales. While effective, the standard WOA suffers from several critical limitations:

Premature Convergence: It tends to get trapped in local optima, especially in high-dimensional and multi-modal problems.
Weak Search Diversity: The population often loses diversity in later iterations, leading to stagnation.
Imbalanced Exploration/Exploitation: The standard linear decay of the convergence factor $a$ creates a rigid transition between global exploration and local exploitation, failing to adapt dynamically to complex search landscapes.
Initialization Issues: Traditional pseudo-random initialization often results in uneven population distribution, causing clustering and blind search regions.

The paper aims to address these deficiencies by proposing a robust, enhanced variant capable of solving complex continuous optimization problems, path planning tasks, and constrained engineering design problems.

2. Methodology: CICDWOA

The proposed Collective Cognitive Sharing Whale Optimization Algorithm with Cauchy Inverse Cumulative Distribution (CICDWOA) introduces five core mechanisms to reconstruct the standard WOA:

A. Good Nodes Set (GNS) Initialization

Instead of pseudo-random initialization, CICDWOA uses the Good Nodes Set method (based on number theory).

Mechanism: Generates a dimension-independent sequence where projections on any subspace exhibit uniformity.
Benefit: Ensures a highly uniform distribution of initial individuals across the search space, reducing clustering effects and providing a high-quality starting point for global exploration.

B. Collective Cognitive Sharing (CCS) Mechanism

This mechanism modifies the "Search for Prey" phase to replace random individual selection with collective intelligence.

Mechanism: Individuals update their positions based on a weighted combination of the Global Best (Leader) and the Population Mean Position.
Formula: $X(t+1) = \frac{1}{2}(X(t) + X_{mean}(t)) + G(t) \cdot |\alpha \cdot X(t) - X^*(t)|$
Benefit: Enhances information utilization and group collaboration, preventing the population from scattering aimlessly and improving convergence directionality while maintaining diversity.

C. Adaptive Exponential Spiral (AES) Mechanism

Replaces the linear encircling mechanism of standard WOA.

Mechanism: Introduces a nonlinear spiral trajectory controlled by an exponential term and cosine function. The step size is dynamically modulated by iteration count.
Benefit: Breaks the linear limitation of the original WOA, allowing individuals to approach the optimal solution with higher flexibility and periodic deviations, which aids in escaping local optima.

D. Enhanced Spiral Updating with Cauchy Inverse Cumulative Distribution (CICD)

Improves the "Spiral Updating" phase to strengthen global exploration.

Mechanism: Applies a Cauchy Inverse Cumulative Distribution to perturb the leader's position. The Cauchy distribution's heavy-tailed property allows for occasional large jumps (long-distance exploration) while maintaining small perturbations for local refinement.
Inertia Weight: Introduces a Sigmoid-based adaptive inertia weight ( $\omega$ ) that starts low (favoring exploration) and increases rapidly in later stages (favoring exploitation).
Benefit: Significantly enhances the probability of escaping local optima and balances the transition between exploration and exploitation.

E. Hybrid Gaussian-Cauchy Mutation (based on Differential Evolution)

Applied in the mid-to-late stages to prevent stagnation.

Mechanism: Uses a Differential Evolution (DE) strategy to generate intermediate solutions, followed by a mutation operator that combines Gaussian (for fine-grained local search) and Cauchy (for coarse-grained global jumps) distributions.
Benefit: Provides multi-scale perturbation, ensuring the algorithm can refine solutions locally while retaining the ability to jump out of deep local optima.

F. Nonlinear Adaptive Convergence Factor

Replaces the linear decay of factor $a$ with a nonlinear Sigmoid-based function.

Benefit: Allows for a more natural and adaptive transition between search phases, accelerating convergence in the middle stages while maintaining robustness in the early and late stages.

3. Key Contributions

Algorithmic Innovation: Proposes CICDWOA, a novel WOA variant integrating GNS initialization, CCS, AES, CICD, and Hybrid Mutation.
Theoretical Framework: Establishes a "Collective Cognitive" paradigm where population mean information guides individual updates, bridging the gap between individual learning and group consensus.
Comprehensive Evaluation: Validates the algorithm across four distinct domains:
- 23 Standard Benchmark Functions (Unimodal, Multimodal, Fixed-dimension).
- 2D Robot Path Planning (Obstacle avoidance).
- 3D UAV Path Planning (Dynamic constraints and height optimization).
- 10 Constrained Engineering Design Problems (e.g., Speed Reducer, I-beam, Gas Compressor).
Open Source: The code and benchmark data are made publicly available to ensure reproducibility.

4. Experimental Results

The performance of CICDWOA was compared against classical WOA, other WOA variants (MWOA, IWOA, etc.), and state-of-the-art metaheuristics (GWO, AROA, IPSO, etc.).

Benchmark Functions:
- Achieved the best average Friedman rank of 1.6790 among all tested algorithms.
- Demonstrated significantly faster convergence and higher accuracy on 23 benchmark functions.
- Statistical tests (Wilcoxon rank-sum) confirmed the superiority of CICDWOA is statistically significant.
Path Planning (2D & 3D):
- 2D Robot: In complex environments with high obstacle density, CICDWOA achieved the shortest average path lengths (e.g., 28.6140 in Map 4 vs. 29.5550 for GWO) with the lowest standard deviation, proving superior stability.
- 3D UAV: Achieved the best objective value (76.95) with the lowest standard deviation (1.83), successfully navigating obstacles while satisfying dynamic constraints.
Engineering Design:
- Solved 10 constrained engineering problems (e.g., Sawmill Operation, Reactor Network, Industrial Refrigeration).
- Precision: In many cases (e.g., Multiple Disk Clutch Brake, Three-Bar Truss), CICDWOA reached the theoretical global optimum with zero standard deviation.
- Robustness: In highly nonlinear problems like the Reactor Network and Industrial Refrigeration System, CICDWOA outperformed competitors by orders of magnitude (e.g., achieving an objective value of 0.00001 vs. 12.19 for the next best in Reactor Network).

5. Significance and Conclusion

Generalizability: CICDWOA proves to be a versatile framework effective for both unconstrained mathematical functions and highly constrained, real-world engineering problems.
Balance of Exploration/Exploitation: The combination of Collective Cognitive Sharing and Cauchy-based perturbations successfully resolves the "No Free Lunch" trade-off, maintaining high diversity early on while ensuring precise convergence later.
Practical Impact: The algorithm provides a reliable tool for complex industrial applications such as autonomous navigation, structural design, and energy system optimization, where getting stuck in local optima can lead to significant economic or safety losses.
Trade-off: The paper acknowledges a slight increase in computational time (15-25% longer than standard WOA) due to the added mechanisms, but argues this is negligible compared to the substantial gains in solution quality and stability.

In summary, CICDWOA represents a significant advancement in metaheuristic optimization, offering a robust, high-precision solution for complex, multi-modal, and constrained engineering challenges.