Abstract
We propose a new foraging model based on the framework of random search with cognition and memory. We consider foraging to be a cognitive process of exploration to find sparse and revisitable targets in a finite landscape and use an exploration-return mechanism to mimic the spatial memory and returning behaviors in foraging. We find that optimal Lévy-flight search strategies, characterized by a power-law step-size distribution $P(l) \propto l^{-\mu}$, can emerge in this setting. The value of $\mu \in (1, 3)$ that maximizes the search efficiency can shift substantially depending on two modeling parameters in the foraging process: the number of steps and the intensity of returning, which can represent a number of general internal or external hidden constraints in foraging. Our model provides a new understanding of the origin of Lévy-flight foraging as well as the substantial difference of the $\mu$-values observed in empirical data, and therefore opens a new avenue for analyzing trajectory and movement patterns of animals.
Introduction: Beyond the "Blind" Searcher
For decades, the study of animal movement has been dominated by the concept of Lévy flights—a specific type of random walk characterized by a power-law distribution of step sizes, $P(l) \propto l^{-\mu}$. In these patterns, many short steps are punctuated by occasional, very long "leaps." Theoretical models have long suggested that a Lévy-flight exponent of $\mu \approx 2$ represents a universal optimum for searching in environments where resources are sparse and revisitable.
However, empirical observations tell a more complex story. From the foraging patterns of albatrosses to the movements of jellyfish and tuna, researchers have recorded $\mu$ values ranging from 1.1 to nearly 3.0. This discrepancy suggests that the "magic number" of 2 is not a biological constant. This paper addresses this gap by introducing a model that incorporates cognition and memory, moving beyond the assumption of "blind" searchers to account for how animals remember and return to known locations.
The Exploration-Return Mechanism
The core of the proposed model is a stochastic $N$-step foraging process in a two-dimensional landscape containing revisitable targets. Unlike traditional random walk models, this framework introduces an exploration-return mechanism consisting of two distinct phases:
- Exploration: The forager moves according to a truncated Lévy flight to discover new resource patches.
- Return: The forager utilizes spatial memory to return to a previously visited target, moving in a straight line.
A critical component of this model is the decision-making process. The probability of an animal choosing to return to a known "hub" increases as it travels further without finding new resources. This is governed by a parameter $\beta$, representing the "intensity of returning." This parameter can be viewed as a proxy for internal constraints such as energy depletion, the need to feed offspring, or environmental stress.
Results: Why the Optimal Exponent Shifts
By defining search efficiency ($\eta$) as the ratio of distinct targets found to the total distance traveled, the study reveals that the optimal strategy is highly sensitive to the search context.
The Influence of Search Duration ($N$)
The research finds that the optimal exponent, $\mu_{opt}$, is not fixed but is an increasing function of the total number of steps $N$. For short-term foraging bouts, "ballistic" motion (where $\mu$ approaches 1, indicating very long leaps) is most efficient. As the duration of the search task increases, the optimal strategy shifts toward Brownian motion ($\mu \to 3$). The classic theoretical optimum of $\mu \approx 2$ was found to occur at approximately $N \approx 1.44 \times 10^5$ steps, suggesting that for many biological lifespans, true Lévy flights ($1 < \mu < 2$) are indeed superior to Brownian motion.
The Impact of Returning Intensity ($\beta$)
The intensity of the drive to return to known locations significantly alters the optimal search pattern. As $\beta$ increases—meaning the animal is more likely to return to base—the optimal $\mu$ shifts to lower values. Essentially, if an animal is a "home-body" that frequently returns to a central hub, it must compensate by taking much longer, more aggressive leaps during its exploration phase to maintain efficiency. This provides a theoretical explanation for why species with high site fidelity might exhibit lower $\mu$ values.
Conclusion: A New Framework for Movement Ecology
This model provides a theoretical bridge between idealized mathematical models and the messy reality of empirical data. By accounting for memory and the necessity of returning to known sites, it explains why different species exhibit such diverse Lévy-flight exponents. A jellyfish with a $\mu$ of 1.2 and a tuna with a $\mu$ of 2.4 may both be foraging optimally, given their specific internal constraints and the durations of their search tasks.
While the model currently assumes a simplified memory structure and a static landscape, it opens new avenues for movement ecology. Future research may build on this by incorporating more sophisticated memory structures—such as prioritizing targets by quality—or exploring how collective foraging and resource depletion influence these optimal strategies.
Summary updated