Here is pre-print of my article.

Abstract: This article presents Biomimicry algorithms, their applications, and possible future perspectives with reasonings. It also discusses combining one or more Biomimicry processes to solve varied, complex problems, which can be mimicked using hybrid process mechanisms derived from several Biomimicry process frameworks. Do we still have unsolved problems? Why the deep study of Biomimicry? Human Brain learning, based on which sophisticated AI algorithms are based, comes in Biomimicry as well. AI is a superset involving not just Bio-Intelligence mechanisms. Why more intelligence in machines? How biological beings such as birds and insects predict bad weather systems, while humans can’t! Is it the brain intelligence or learning, or do they have some sensors? Or is it instincts? If so, how does this intelligence in birds/insects work? We have studied how birds flock in the sky to find food; what about the other intelligent abilities of these beautiful beings? These studies can hold the potential to answer questions about whether AI can ever compete with natural biological intelligence of various kinds! Or is it complex neural learning? How do we theorize this to make us freeze on the decision to endlessly pursue the growth of intelligence? What about some white box rather than black box-based AI? So, what, apart from theorizing, is the very fundamental question of where the race to AI leads? There are applications that Biomimicry can solve, for example, possessing intelligent machines with multiple intelligent behaviors, each of a human brain intelligence, honey bee “wangle dance,” and the intelligence of a bird swarm to guide multiple swarms of self-organizing robots flying in deep space to bring in the latest live videos from an active part of the Universe! The aim of developing newer intelligence in machines is not to do things humans like to do but to do things humans can’t do, such as the challenging tasks of mining precious minerals back from garbage piles. An object-orient framework for biomimicry is needed; for example, in climate disasters, machines operating in a semi-supervised way with firefighters can make intelligent decisions based on classes derived from Biomimicry specializations to most needed applications based on run-time instantiation. Hence, a need for a planned, structured object-oriented development for generalization, specializations, and implementations of these algorithms with the help of ecologists is required. The aim is never to take away human jobs but to assist in tasks human can’t!

Biomimicry, a field that draws inspiration from sophisticated natural/biological processes, holds immense potential for developing artificial systems. The algorithms in Biomimicry replicate the natural biological process and mimic its workings to solve complex real-life problems, such as climate calamities and forest fires, along with humans in a semi-supervised manner. Once a natural/biological process is learned, these intelligent algorithms and learning phenomena can be used as Artificially Intelligent Algorithms to solve problems in fields of different complexities. When designed with the correct parameters and generalizations, these systems can achieve their designated aims, showcasing the power of biomimicry algorithms. Biomimicry can also be considered as a term serving the purpose of mimicking a single agent or multi-agent behaviours that natural processes utilize. These natural phenomena of intelligence can exist in part or whole in a living organism or biological processes and can be analyzed for their unique characteristics. Examples of such processes in Biomimicry can be applied to challenging engineering, technical, decision-making, and economic problems to mention a few.

Various Biomimicry-based algorithms have been in use for a long, even before this term was used prominently as it is today. Some critical biological processes, which have been mimicked as artificially intelligent systems, have been briefly explained here. Examples of possible applications of these Biomimicry algorithms are also illustrated here.

1. Swarm Intelligence: Swarm Intelligence (Trelea, 2003; Clere and Kennedy, 2002) is structured use of deep mechanisms followed by the individuals in the swam in decentralized ways to attain their goals. Here, a flock of swarms and their processes of searching for food, shelter, migration, mating, and intelligence are studied, analyzed, and processed for applications in real-world problems. The problems they solve can vary from less complexity and advances to high complexity as the methodology is understood in detail and matures. The swarms can be swarms of birds of various kinds, fish of different kinds, whales, and so on. Each of these is studied as an independent intelligence mechanism; I clubbed it into one for simplicity. Each has a unique characteristic to study, analyze, and understand; for example, whale intelligence differs significantly from the school of fish intelligence. Though both of these may have some similarities and many differences.

The aim of the naturally existing swarm is typically to find food and to migrate at the right times to the right places. Many of these Swarm Intelligence Systems work in a way that, to humans, looks like solving complex combinatorial optimization problems. And they are really good at it. Hence, the proposed use of these algorithms is to intelligently solve many complex combinatorial optimization problems. Each individual participant in the system [any Swarm System or any model for Metaheuristic Algorithms] they are typically called an agent. These agents in Swarm Intelligence Systems can be a group of birds or fishes, though no concrete boundaries have been drawn till now on it. For example, the swarm of birds of kinds in the Southern Hemisphere may perform intelligent behaviors differently than those in the Northern Hemisphere. There may be some changes in intelligent mechanisms; we need to study these changes as while on deep sea exploration to clean sea beds of rotten shipwrecks by multi-agent autonomous robots, these robots may need to adapt to different techniques derived from, say, the base classes of the Southern Hemisphere fishes/birds or Norther Hemisphere fishes/birds at run time, i.e., using dynamic polymorphism based on conditions faced by these robots in deep freezing sea (non-human friendly conditions). The class to instantiate on run-time would depend on the characteristics we need in, say, these deep space explorations, what behaviours are required, and what challenges are faced in there by these multi-agent brains. And what Biomimicry class can help solve this problem. Hence, the proposed framework in the article to define abstractions and generic mechanisms of a variety of Biomimicry techniques can not only help in providing the users/engineers/modelers with a plethora of customized options to use but also make this huge biological science a concrete and indexed field of Biomimicry. However, the example of this southern hemisphere is just an illustration; ecologists can for sure provide more details of variants in intelligent behaviors. Once such an object-orient framework for biomimicry is ready, any new robotic system needing independent work, say in climatic/fire disasters, can make decision-making along with humans as a toolkit for humans, based on Biomimicry implementation most suited to a need again based on run-time polymorphism. One example of swam intelligence is Particle Swarm Algorithm (Trelea, 2003; Clere and Kennedy, 2002), most of the algorithm works on the following key criterions:

1. Each particle (or agent) in a swarm has a velocity.

2. Each particle has a position.

3. The best particle parameters are notes

Various variants of these algorithms are studied, tested, and applied in real-time applications. Although this is still an object-oriented abstraction, specialization and instantiation-based implementation is required due to the presence of plethora of variants of PSO in the research repository only.

Consider a flock of swarms — which can be birds of different kinds, whales, or fish. These can be studied as natural entities to understand the intelligent processes followed by the swarm to find food, mate, and migrate at the correct times, as the case may be. The generic structure of the study, for example, for common birds, follows the phenomenon that these groups of entities match the position and location of the best bird in locating food and guiding directions. The birds in the flock adjust the velocity and displacement according to the local best and global best of some of the best birds in the flock. This is a considerably intelligent behavior that birds possess and has proven capability to solve many combinatorically complex problems.

The swarms are decentralized, and hence Multi-Agent Systems. One real application of this decentralized system can be seen in Space Robotics; however, robots can perform in centralized mechanisms as well. So, one must appreciate the use of Swam intelligence in tasks that are really decentralized, in behavior, such as a virtual artificial being communicating to attain some goals, maybe a self-developed used mineral mining from piles of garbage performed by a swarm of mini-Robots.

2. Brain

The human brain has fascinated scientists for a long time. Researchers and philosophers have wanted to replicate its workings for a long time until, in the 1950s, the field was dedicated to this research, and the term Artificial Intelligence (AI) was coined to mimic the workings of intelligence in the human brain. Later, AI was expanded to include all intelligent behaviours, even intelligence exhibited by Biomimicry-based and other algorithms. Most advanced algorithms in AI are based on the concept of Neural Networks, which is the very foundation of how the human brain works.

Other species brains such as monkey brain, rat brains, birds brains can be studied in the same way to understand the mechanism of learning. Although the basis of learning is neurons but what is the difference in intelligence of a rabbit brain and a human brain? Even though both are based on concept of learning! Studying different brains for varying intelligence can open wings to theoretical differences in the intelligence of species. Yes, studying a less complex brain, such as a rabbit brain, won’t solve problems we can solve by learning the complex human brain learning algorithms. So, why do we need to study these less complex brain structures? Well, to understand the working of other species and how they differ from human learning capabilities. You may disagree here; but how else can we differentiate the theoretical model followed by a human brain and a rabbit brain? How long can we pursue AI to gain human-like intelligence? We need to conceptualize this! Why do we need to understand the difference as one day, we want to see the AI, which is a black box, turn into a white box, if not a transparent box, on the first go. This is why children’s brains are also studied as a different study for quick learning brain model. Another reason why we need to study different brains as a separate entity separated from the study of the human brain?

Another context to illustrate is when extreme weather is forecasted, birds behave in different ways, while humans are unaware of it, maybe some sensing capabilities. Is this in the brains of birds or some sensing powers? Even insects can predict extreme weather and earthquakes, as they hide in hives when such a thing is about to happen. How?

Is it instincts? If so, how does intelligence work? Is it all neural learning in the brain? Can this answer questions about whether AI can ever compete with biological intelligence?

How biological beings such as birds and insects predict bad weather systems while humans can’t! Is it brain intelligence, or do they have some sensors? If so, how does this intelligence in birds and insects work? We have studied how birds flock in the sky to find food; what about other intelligent abilities of these beautiful beings?

These studies can hold the potential to answer questions about whether AI can ever compete with real biological intelligence of various kinds! Or is it complex neural learning? How do we theorize this to make us freeze on the decision to endlessly pursue the growth of intelligence, some white box rather than just black box-based AI?

These are just some differences in the intelligence levels of birds and other species, which can predict certain odd things better than a highly complex and extremely intelligent human brain. These can, hence, be included in the broad context of Biomimicry as an independent subject to theorize the studies on biological and natural intelligence mechanisms and how much of it can be mimicked with the applied science of Biomimicry.

These are sets of Biomimicry algorithms that utilize the working of the human brain to solve problems, just the way the human brain finds solutions to the problems. Extensive work has been done in this area. A recent new approach of considering brain functioning as a brain energy optimizer to derive Hebbs’s law-like equation has been derived from Luczak et al. (2022). This has been proven to mimic the prediction behaviors of living brains. And it is proved with this reduced to Hebb’s law. Hebb’s law was often overlooked. However, we have solved many of the Machine Learning problems, there is a need to solve generic problems, and hence, it is required to understand the working of the brain as a biological process rather than just a perceptron learning model, a target error minimizer and in a way a function modeling of the human brain. In this work, brain energy optimization has been performed, and the way energy is consumed in neural models of various living organisms varies. Hence, we can see this as, again, a collection of variations of algorithms. Further, this is just an initial step and more complex algorithms for predictions shall be coming soon. Hence our proposed framework of generalizing and structuring the Biomimicry Algorithms shall pay dividends in the future apart from ease of understanding in decision making of choice of algorithm to choice of bias and parameters.

3. Biological Reproduction

The application of this theory in AI is called Genetic Algorithms, and it is based on the way biological genetics in living beings work, the way offspring are produced, and Darwin’s Theory of survival of the fittest, which involves choosing the fittest gene to advance in biological succession. In Biomimicry of the concept of biological reproduction, a population of competing candidates is taken, from which candidates are selected for the artificial mating process; this is performed by experimental crossover algorithms. Once crossover is performed, the mutation is induced to explore the solution space for newer feasible possibilities. In the end, the fitness of each candidate offspring is computed, and henceforth, a new population is created based on some specifications. Mostly, all living organisms follow the same process of selection, crossover, and mutation, and once this process is implemented as an artificial genetic algorithm. The genetic algorithm has a proven record of wide applications. There are problems that regular algorithms can’t solve that efficiently, which can be solved in efficient time by genetic algorithms; yes, there is a clause of accuracy that can be set with time and parameters of execution of these algorithms. Again, there are various kinds of crossover reproduction processes varying over the living beings on the planet; hence, an object-oriented representation of existing code via refactoring is required.

4. Plant Intelligence

Mendez and Marcum (2019) proposed a Biomimicry mechanism based on the processes of growth and reproduction in plants. They proposed a step-by-step analysis of the intelligent process of growth of a sapling from seed to a self-sufficient plant, defined as an intelligent process. Further, even the process and time of intelligent reproduction and flowering of plants are discussed as Biomimicry processes. These are illustrated well with the help of the most easily studied plants, which are beans. Once these are formalized in an intelligent algorithmic mechanism, applied, and evaluated on existing problems, and maybe these can be analyzed for use in new challenges. One application of Plant Intelligence that we propose here is robots diving in deep space on their own just like plants, using this mechanism. Another reason is to know what kind of intelligence plants have. What is there in the plants, trees, and nature? Can they learn, do they feel too? Does the neural-like structure exist there as well? Don’t we want to theorize the foundations of living beings’ intelligence? Plants, too, are living beings — and they produce oxygen; isn’t it tempting enough to know them more, know their plant intelligence more?

5. Insect Intelligence

Insects of various kinds perform their operations of finding food, mates, and water in various ways. There are, again, many such variations in studies in literature. All these variations can be instantiated as a specialized class in the Biomimicry of insect intelligence. Let us discuss one famous of these, which is Bees, the way bees find, flower nectar, make honey travel around, mate, and even pass on the information to other intelligent bees in space. These algorithms come under the broad category of Bee Intelligence Algorithms (Teodorović, 2009, Rajeswari et al, 2017, Chong et al, 2006).

The algorithms inspired by the study of the way bees search and collect food have intrigued many researchers and scientists, not just in Ecology but also in AI and Applied and Theoretical Sciences where interest grew in the way, bees are able to intelligently find the nectar and it is stored well in the beehive as if the bees are guided by some kind of mechanism like GPS, a location identifier, for both the flowers with the right nectar as well as the tracking of the path to go back to the hive. The main principle in mechanism followed by the bees were well understood by scientists from all these diverse fields and it was found that the primary strategy bees followed was to find the right target and deposit it in the hive, once the target was found the status of the bees were changed to informed bees form uninformed bees. The informed bees participate in the Waggle dances in just above the areas of food, in these Waggle dances, it is expected that information is shared between the informed and uninformed bees. Further, other information that is shared in these Waggle dances is about long-distance and short-distance targets. Also, once the food finishes in a region, the bees are again termed as uninformed till a new food region is found and a new learning mechanism, as explained above, is constructed. Some popular algorithms based on the way bees operate are as follows (Teodorović, 2009, Rajeswari et al, 2017, Chong et al, 2006)

a. Bee Colony Optimization

b. Bee System: The proposed system claimed to perform better than Genetic Algorithms. The system appended Genetic Algorithms with more operations based on the way Bees optimize their search for the right nectar. This was included in a concentrated crossover operation, which I will describe briefly.

c. Beehive Algorithms

d. MBO Algorithm, based on the marriage process of bee Queen.

e. Virtual Bee Algorithm

These are applied in various fields and used to solve various problems not limited to Mathematical Optimization, Computer Science, Computer Networks, Medicine, Neural Networks, and Fuzzy Reasoning, to mention a few. Further, the above are just some of the variations of Bee Intelligence Algorithms; hence, a need for a planned, structured object-oriented development for generalization, specializations, and implementations of these algorithms with the help of ecologists is required. This can be utilized in an Artificially intelligent Biomimicry algorithm that picks and chooses the best out of all these variations, and some can include most of the properties, not just of the waggle dance, but also of informed search to making more bees, which is reproduction.

6. Animal Intelligence

These are Biomimicry algorithms that work on the way herds of sheep, elephants, or even tigers find food, exist, mate, and attain essential agendas to be attained. The variations in algorithms are again too diverse and frequent, and a framework and structural classification can really help in solving real problems in Biomimicry. Similar logic for making an object-oriented development of these algorithms, abstraction, generalizations, and specializations to be multi-inherited in deep space dynamically based on challenges faced by multi-agent robots to attain newly discovered aims, a place where humans of Earth may never be able to go!

7. Organ Intelligence

Biomimicry algorithms can also include the study of intelligent organs. For example, the heart of a pig is an intelligent organ that takes some input and produces some output and can hence be utilized. We are mentioning this as this article covers all currently possible Biomimicry algorithms.

So, what, apart from theorizing (if theorizing is a possibility), is the very fundamental question of where the race to AI leads? Where?

There are applications that Biomimicry can solve, for example, possessing intelligent machines with multiple intelligent behaviors, each of a human brain intelligence, honey bee “wangle dance,” and the intelligence of a bird swarm to guide multiple swarms of self-organizing robots flying in far off deep space to bring in the latest live videos from an active part of the Universe! Note these are places humans can never dive into!

The aim of intelligence in machines is not to do things humans like to do but to do things humans can’t do, such as the tough tasks of mining minerals back from garbage piles. Biomimicry Algorithms can assist machines in gaining adequate intelligence, which can aim for autonomous machines for intended purposes, especially independent far-off space explorations (here humans have limited capabilities) where machines need to possess multiple parallel intelligence characterizations are required. Why space explorations with Biomimicry Intelligence or even AI? Imagine a live TV channel just like National Geographic replica for space, showing multiple star sunsets in a deep universe. These machines may be able to catch all these intelligently and travel to places in the universe that are interesting and conduct independent research there. Why do we even need such a thing for space? Well, why do we run after rabbits? Why do we photograph tigers? This is mankind’s natural instinct to see the natural creations and natural beauties, care for them, and visualize the blessings of nature!

These biological processes in Biomimicry can be subclassed, generalized, multi-inherited, and conceptualized with an abstract framework. The abstract framework is needed for each of these generic Biomimicry classes; this has several benefits, one being generalization, specialization, multiple instantiations, and multiple inheritance for gaining various characteristics with ease of implementation. Proper subclasses be defined to instantiate the intelligent processes and behaviors in those particular classes. These categorizations can enable understanding and inheriting multiple behaviors from multiple intelligent biological processes, creating Hybrid Biomimicry customized for use in problems such as dealing with autonomous forest fires, which can be used as a modern toolkit by human firefighters, a toolkit they deserve to run on their own, a collection of intelligent robotic drones to manage fires by human firefighters; there are constraints here. Another example — let us suppose there is an abstract behavior of an intelligent model of the human brain and an abstract behavior of a swarm of fish attaining hybrid Biomimicry intelligence. An application inherits some instantiations and specific specializations of these abstract behaviors to create an artificially intelligent swarm of fish with human brains to pursue freezing-deep ocean missions to save coral reef species in extinction or being bombarded by age old crashed airplanes and shipwrecks, those that are undoable by humans. This shall play a key role in robotic working in multi agent systems.

The research in Biomimicry has expanded its horizons in width and depth and hence deserves to be an independent subject on its own. Just like AI was once a part of computer science, AI now stands on its own as a full-fledged subject that cannot be covered in one course anymore. One must note that no new field can be sparked for long growth till there is some reason for it to grow. Hence, sooner Biomimicry may need its own course to be studied, not just for applications in existing problems but also to find solutions to problems that mankind has to face in coming times; an example of that can be a robot handling its own mechanism of setting units of explorations Mars, for sure for more than just watching sunsets on Mars. Humans can’t go too deep space explorations but can sure go to the Moon one day; who knows? But these intelligence can help if humans make a decision to go this way, the decision is always in the hands of humans, not AI. Humans run AI, AI should not run humans.

If this term, Biomimicry, is used independently of AI, can have far-reaching impacts as a group of robots can be considered as a group of bees with an aim to accomplish. This is no longer limited to one brain-mimicking model. There can be many brain modes, as the birds in flocks are not just entities that maintain the speed and velocity and compare with a memory of local best and global best. No, now, these can be considered as multi-brains in a flock to serve some aim. Each brain possessed an advanced deep learning kind of model. A fusion! A change? Needed or just an experiment to conduct? Think, if you want it!

AI was presented with problems, researchers and AI experts solved them, good accuracies were attained, and models were developed, learned, and applied to unknown data with considerable accuracy. So, why do we need to study Biomimicry in more detail? Why do we need to research more? The motivation is — the new challenges we are facing and the growth of machines for problems not to interfere in human life but in problems such as mining rare metals from huge piles of garbage, a new kind of mining, and maintaining forest fires, to mention a few. Had the same approach solved these new problems, we wouldn’t have required more need to study more. Analyzing intelligent behaviors in nature can help us in solving many upcoming problems, including Intelligent Multi-Agent Problems, such as a group of drones cooperating to control the spread of fires of kinds not limited to forest fires.

The use of structuring in design implementation would not just help in applying these variations in real-time applications, as illustrated all over in this article. It would also yield ease of use, choice of algorithms, and the possibility of extending these algorithms. However, the main use of the development of the Biomimicry algorithm framework is the multiple inheritance of a variety of Biomimicry algorithms. This shall solve customized problems. This in no way means that all the research and code written in the Biomimicry algorithm are waste; no, it just needs to be re-factored and framed well to fit in a universally defined framework of structure and subclasses. Conversion of structured programming to object-orient programming is not a tough task for an experienced developer, if we see the benefits outweigh mild work.

However, deciding on a Biomimicry algorithm framework is not in the hands of an AI researcher only; it requires people from diverse fields, which include ecologists, environmentalists, biologists, zoologists, computer scientists, and AI researchers. Hence, the exact conceptualization of the Biomimicry algorithm framework is left as an open task to be implemented after universal consensus is reached on freezing the framework details, just like CDA HL7 has been formally frozen as a universal medical documentation language.

The scope is extensive, and future work involves conceptualizing frameworks and making structures and programs fit in this Object-Oriented Model so that a dynamic biomimicry algorithm can be constructed as and when an application comes in. Further, studying each intelligence capability in its own uniqueness has many reasons, not just for applications but to know the theoretical differences and the possibility of any human-like intelligence in machines. Given birds and insects reflect behavious which are more than neural learning! Can this intelligence ever be embedded in machines? If No, why not replicate what we can to deal with problems at hand and work on developing semi-supervised models to assist in intelligent tasks, such as firefighters working with robots and machines to deal with fire in towns or forests? Can human-like intelligence ever be achieved? We need to freeze these questions for an endless pursuit of the intelligence of machines; hence, theoretical representation and study of biomimicry are required.

References

[1] Luczak et al. Neurons learn by predicting future activity. Nature Machine Intelligence (4). 2022.

[2] Trelea, The Particle Swarm Optimization Algorithm: convergence analysis and parameter selection. Information Processing Letters (85). pp 317–325. 2003.

[3] Clere and Kennedy, The Particle Swarm- Explosion, Stability, and Convergence in a Multidimentional Complex Space. IEEE Transactions on Evolutionary Computations (6). pp. 58–72. 2002.

[4] Mendez and Marcum, Deep Learning with Biomimicry. Journal of Design and Science. 2019.

[5] Teodorović, Bee Colony Optimization (BCO), C.P. Lim et al. (Eds.): Innovations in Swarm Intelligence. pp. 39–60. 2009.

[6] Rajeswari et al, Directed Bee Colony Optimization Algorithm to Solve the Nurse Rostering Problem, Computational Intelligence and Neuroscience. 2017. https://doi.org/10.1155/2017/6563498.

[7] Chong et al, A Bee Colony Optimization Algorithm To Job Shop Scheduling. Proceedings of the 2006 Winter Simulation Conference, IEEE. 2006.