What can be done when evolution meets algorithms?

Everyone knows about the evolution of plants and animals.

But you know

Engineers can also take advantage of this natural process

To invent something?

Animals and plants in a difficult and complex environment

Need to evolve to adapt

As the saying goes, survival of the fittest

Image source: Photo Network

Biologists have long studied how evolution works, and mathematicians and computer scientists have teamed up with biologists to help engineers invent things by creating computer programs that can improve designs - these are called evolutionary optimization algorithms , and they can be used to design faster planes, stronger bridges, and even better games.

What exactly is evolution?

Evolution is the word we use to describe how plants and animals change over long periods of time.

For example, a child will resemble both of his or her parents. Perhaps the child has the same hair color as the mother and the same height as the father. The transmission of this similarity between generations is called inheritance.

There can also be subtle differences between a child and his or her parents, perhaps a bigger nose or better eyesight. These differences are called mutations.

Image source: Photo Network

In nature, tiny mutations can mean the difference between life and death for plants and animals. For example, if two zebras are running away from a lion that's looking for lunch, the fastest will be the one that escapes and survives. The surviving zebra will be able to give birth to offspring that will likely inherit their parents' running speed. The fast baby zebras will also be more likely to survive and have babies of their own, so over time the zebra population will become fast runners. This process is natural selection.

Species evolve to survive in their environment, a combination of genetics, mutations and natural selection.

Organisms evolve in amazing ways, doing incredible things—from moths that can change color to hide from predators to lizards that can drink water through their skin.

So, can we use these ideas from nature to create inventions?

Evolutionary Optimization Algorithms

Regarding this problem, several computer scientists have proposed the idea of ​​evolutionary optimization algorithms.

Between 1950 and 1960, evolutionary optimization algorithms were used for all sorts of things, like designing airplanes, building levels in video games, and even creating art.

Now, let's imagine we are trying to design a bridge:

First, we need to set a goal. For example, we want the bridge to support as much weight as possible. Once we have a goal, we can compare two bridge designs - whichever one can support the most weight is better.

**The next step is to create some starting bridge designs. **This can be done randomly with a computer, or we can spend some time designing them ourselves. By running simulations on the computer, we can predict how much weight each bridge design can handle. Using this information, we can rank the designs according to their quality.

Once we know which designs can bear the most weight, we can select them as “parents” — similar to what happens in nature, remember? Only the fastest zebras get to have babies.

Of course, bridges can't actually have "children," but scientists can write computer programs that can combine two good designs to produce a new "child" design that shares features of both "parents."

**Continuously generate new designs using inheritance and mutation, then simulate these new designs, determine which designs are best, and repeat the process. **All of this is done automatically by the computer. Over a long period of time, perhaps weeks or months, better and better bridge designs are found.

Advantages and disadvantages of evolutionary optimization algorithms

Evolutionary optimization algorithms are very useful when we need to design something new and don't know where to start.

However, there is a randomness in evolutionary optimization algorithms, which causes some engineers to distrust evolutionary optimization algorithms . The randomness also means that designers need to run evolutionary optimization algorithms many times to ensure that they get the best design. Running evolutionary optimization algorithms over and over again can take a long time, even longer than it would take an experienced engineer to sit down and design something himself.

Image source: Photo Network

**In some cases, evolutionary optimization algorithms struggle. **For example, there is often more than one objective to consider. Adding more objectives means that the evolutionary optimization algorithm will take longer to find a solution.

It takes hours for an evolutionary optimization algorithm to "evolve" a good design, but it only takes seconds for an engineer to realize that the simulation is wrong, and when that happens, the scientist and engineer need to fix the error. So while evolutionary optimization algorithms are useful tools, they will never replace human designers.

References:

[1] Mitchell M, Taylor CE 1999 Evolutionary computation: An overview

[2] Hornby, J., Globus, D., and Linden, J. 2006. Automatic Antenna Design with Evolutionary Algorithms. Reston, VA: American Institute of Aeronautics and Astronautics.

[3] J. Vincentlake, S. Walton, B. Evans, 2021 It’s the journey, not the destination: Building genetic algorithms that practitioners can trust

[4] Wang Shuaifa, Zheng Jinhua, Hu Jianjie, et al. Multi-objective evolutionary method for adaptive preference radius partitioning. Journal of Software, 2017, 28(10): 2704-2721)

[5] Qiu Feiyue, Wu Yushi, Qiu Qicang, Wang Liping. High-dimensional objective evolutionary algorithm based on bipolar preference dominance. Journal of Software, 2013, 24(3): 476-489)

[6] Gong Dunwei, Liu Yiping, Sun Xiaoyan, et al. High-dimensional multi-objective parallel evolutionary optimization method based on goal decomposition. Acta Automatica Sinica, 2015, 41(8): 1438-1451

END

Author: Not the quarterback who wears No. 9

Qingdao University of Science and Technology, Robotics and Intelligent Manufacturing Technology

Audit expert: Chen Mingwei, School of Mechanical and Electrical Engineering, Qingdao University of Science and Technology

Editor: Guru