Bots

Started as internal research of data visualization methods, Bots became an exploration of evolution in a digital environment. The film features dozens of imaginary robots, each with its own behavior and character, interacting with each other and their environment.

Led by a playful spirit of experimentation, we aspired to find underlying principles of growth and development existing in this world.

SHOT_040_v01.SHOT_040.mov_snapshot_00.00-1

Background

Spectacularly powerful, our computing technologies have only been around for several decades, while natural evolution has been happening for millions of years. 

Through many dead-end turns and few lucky events, human intelligence appeared, still unattainable for the AI technologies.

Creating enhancements to nature's best inventions, humans can act more directionally and thus speed up natural progress.

However, biological systems are still too complex even for our understanding—studying the variety of organisms existing in nature for centuries, we don't yet fully comprehend our own bodies.

Natural structures are even more complicated for our attempts to reproduce them artificially because of plenty of influencing factors to be considered.

Instead of ill-fated competition with the invisible forces of nature, scientists choose cooperation with them.

Without an exhaustive theoretical understanding needed for building systems from scratch, researchers can still apply some of the well-studied natural principles to their work.

Bots_Lines_01_00000_hor-1

One of such rules and frameworks is evolution. Experimental at its core, this process is made possible in digital environments by the development of powerful processors and simulation software.

With the ability to generate and test hundreds of models without creating actual physical prototypes, the fittest options can be found more rapidly than before, accelerating the progress in many fields of knowledge.

SHOT_110_v03.SHOT_110.mov_snapshot_00.01

With the ability to generate and test hundreds of models without creating actual physical prototypes, the fittest options can be found more rapidly than before, accelerating the progress in many fields of knowledge.

The disciples of these approaches use mechanisms of natural selection, mutation, and reproduction found in nature to detect new algorithms and configurations to apply to their respective objects of study.

In computer science, evolutionary strategies are embodied in genetic algorithms. Generally, the process starts from the generation of candidates for evaluation, each of them getting a random set of parameters.

Then, the options are evaluated based on their performance in a certain task or on a certain set of parameters called fitness function. The best variants are once again randomly tweaked, and the 'mutated' versions are evaluated again.

Such a mechanism has many applications: automated design of hardware, network optimization, prediction of molecule structure. It can be combined with other methods inspired by biology, for instance, neural networks.

This fusion is used for the training of driverless cars: neural networks with random sets of parameters learn to detect obstacles and are periodically evaluated during the process of studying.

The best of them are used to create progenies, and the worst are excluded from the population.

SHOT_170_v03.SHOT_170.mov_snapshot_00.11

Evolutionary robotics deals with both the 'minds' and 'bodies' of robots. Their intelligence can be trained either for an individual machine, using a mechanism similar to one in driverless cars, or for a group that can be trained to perform a difficult task as a whole.

For the physical features, the evolutionary method is used to predict the most effective constructions.

The approach was applied by NASA in the development of antennas for spacecrafts: the structure created by random generation outperformed every design created by humans.

However, in more complex optimizations, simulations can't accurately predict every aspect of the material world.

'Reality gap', as called in robotics, can lead to the creation of a robot that walks perfectly in a digital simulation but falls in real life.

Our research

Our experiment started from the idea to mix industrial design and biological principles of movement.

Having small jumping robots as the result, we decided to generate more of them and see how we can change their features to alter their behavior.

Bots_part_two_DP_v03.cam01.mov_snapshot_00.05
Bots_part_two_DS_v04.cam03.mov_snapshot_00.02

Over the course of work, we created a wide variety of bots: big and small, fast and slow, simple and complex.

Eventually, this research turned into a story about evolution—we see bots evolving and beginning to make the mark on the world they live in.

Bots_part_two_DP_v05.cam01.mov_snapshot_00.02

Working together, they give structure and meaning to their environment.

Mechanical and intricate, or random and chaotic, patterns drawn by the bots remind of strict mathematical diagrams and art, further accentuate the connection between nature, science, and innate creativity explored in the project.

Bots_Studio_DP_v09_09

Credits

Creative Direction:
Maxim Zhestkov, Igor Sordokhonov, Helge Kiehl

Design, Animation:
Dmitriy Ponomarev, Denis Semenov, Sergey Shurupov

Art Direction, Edit:
Dmitriy Ponomarev

Graphic Design:
Xenia Turubanova

Music:
Toxe

Sound:
Artyom Markaryan

Text:
Anna Gulyaeva

Year:
2021