The team and the author grow fat cells, transformed into neurons, on a plate with electrodes (MEA). The neurons trigger the robot to move in space with their electrical signals, similar to a brain on a chip. Through a sensitive and complex sensing and programming system, the signal from the neurons is converted from analog to digital. It is then amplified and used to find the most appropriate shape for the bot and to move it.

To determine the shape of the Biobot, the AI program uses the output signals from the neurons and compares them with possible leg movements in a library of movements from different arthropods. From these juxtapositions, the AI deduces the appropriate amount of joints and limbs and suggests the most optimal skeletal constitution for movement. The results are reflected in the shapes of the unpredictable zoomorphic bodies, offered by the AI according to the possibilities of movement, and chosen by the artist according to his interest in the handicapped movement. The algorithmic search for the shape of the bot according to the stimulated biological activity is revealed as an uncontrolled evolutionary process by iterations of signals from the brain organelle. The possibilities of a hybrid bio-cybernetic aliveness with its own aesthetics, intelligence, and movement open up before us.

Concept, development: Zoran Srdić Janežič
Bioengineering: Kristijan Tkalec, BioTehna Lab; Prof. Dr. Helena H. Chowdhury, Laboratory of Endocrinology – Molecular Cellular Physiology, Faculty of Medicine, UL
Expert assistance: Martina Perše, PhD, Medical Experimental Centre, Faculty of Medicine, UL
Programming: Benjamin Fele
Biosensor electronics, programming: Erik Krkač
Electronics, PCB design: Gregor Krpič
Measurements of signals from neurons: Jakob Grčman
3D design: Cveto Kuneševič
Production: Zavod Kersnikova – Galerija Kapelica | konS ≡ Platforma za sodobno raziskovalno umetnost
Curator: Jurij Krpan

In several years of artistic research, a team of experts together with the artist explore, experiment, and develop neural tissue to process input signals from it  and control the robot.

Biotechnologists and the author grow fat cells, differentiate them into neuronal cells, and develop them on a Multielectrode Array (MEA) into a simple hybrid organelle that, like a brain-on-a-chip, can control the movement of the robot in space. Through a sensitive and complex sensor system, the signal from the neurons is converted from analog to digital, amplified, and then used by the AI to find the most appropriate shape for the bot and its movement.
To determine the shape of the Biobot, the AI program uses the output signals from the neurons and compares them with possible leg movements in a library of movements from different arthropods. From these juxtapositions, the AI deduces the appropriate amount of joints and limbs and suggests the most optimal skeletal constitution for movement. The results are reflected in the shapes of the unpredictable zoomorphic bodies offered by the algorithm according to the possibilities of movement, and chosen by the artist according to his interest in the handicapped movement.
The algorithmic search for the shape of the bot according to the stimulated biological activity is revealed as an uncontrolled evolutionary process by iterations of signals from the brain organelle. The project is revealing the possibilities of a hybrid bio-cybernetic aliveness with its own aesthetics, intelligence, and movement.

Hardware
1. Custom-designed incubator is allowing us to exhibit and observe cell cultures in real time.
Observation is direct, through a built-in microscope and through an inbuilt computer which is connected to other data acquisition and visualization systems in the next stage for data processing.
1. The incubator was part of the S+T+ARTS residence, made in collaboration with lab equipment manufacturer Kambič. The basic distinction from other incubators is also in nine Peltier plates which allow a precise temperature range needed for growing cell cultures. The CO2 atmosphere is provided.
2. Data acquisition system for reading neural signals is made of Open Ephys hardware, software, and protocols which allowed us to build a low-cost method to acquire signals from the neurons, grown on the in vitro multi-array technology MEA. Basic protocols and blueprints for building required hardware represent a cost-effective and multifunctional precise amplifying system for in vitro electrophysiological investigations with multi-electrode arrays* which also review connections between the custom PCB board and the INTAN RHD200-EVAL board, showing the electroporation, the acquisition, and the computer transfer paths – that is a connection to Open Ephys acquisition board and the computer.
3. Robot is custom-made by the author from SLA printed parts, assembled to the Dynamixel X-Series servomotors which are a series of firm ROBOTIS. The XL330-M288-T Servomotor is equipped with contactless magnetic encoders allowing 360° Rotations with up to 61 RPM. It is capable of torque up to 0.52N.m with a voltage of 5V.

Software
When designing robots with many degrees of freedom, selecting the appropriate morphology and kinematics can be challenging. To address this issue, the Biobot project leverages RoboGrammar, a two-stage approach that solves these problems in simulation. In the first stage, an evolutionary optimization algorithm searches for a robot configuration that can move within a given environment. This involves generating various robot configurations from a defined grammar and selecting the most promising ones for evaluation in the second stage, based on a trained value function. Here, grammar specifies which robot body parts can be combined together. In the second stage, Model Predictive Control (MPC) is used to identify the sequence of moves that maximize the robot’s position change. This evolutionary algorithm is employed to select the most promising sequence of moves within a given time horizon. We enhance this two-stage pipeline by incorporating neuron signals obtained from OpenEphys. Although we continue to use MPC to find the sequence of moves, action potential frequencies regulate the torques in the joints. This implies that smaller or larger frequencies correspond to a smaller or greater ability to move a specific joint, respectively. This can influence the robot’s ability to move and the morphology of the robot that we seek in the first stage of the algorithm. We determine joint positions for each time step of the robot using the Bullet physics simulation engine and transfer the identified positions directly to the real-world robot. Our implementation is executed in Python, utilizing libraries such as Numpy, and PyTorch, as well as code from the RoboGrammar repository.

*Leonardo D. Garma,Laura Matino,Giovanni Melle,Fabio Moia,Francesco De Angelis,Francesca
Santoro, Michele Dipalo: Cost-effective and multifunctional acquisition system for in vitro
electrophysiological investigations with multi-electrode arrays. Published in: Plos One, March 25,
2019.

The biobot is a multi-year development project to create a cybernetic life form:

that is a combination of biological, organic, data processing, and machine construction. From the start, the work has been divided into several disciplines: tissue engineering, where we work with neural cells, to electronics, mechanical engineering, and computer programming. For example, to meet the conditions dictated by the biological part of the cybernetic body, we had to create a suitable living environment. This environment had to sustain a certain temperature range, humidity, and carbon dioxide content for the growth of neurons – so we developed a special incubator with the industrial sector. To this environment, we added a communication channel to connect with the cells, where electrical impulses from them had to be decoded.
In the BioTech lab, Kersnikova colleagues and I differentiated fat cells into neuronal cells, superimposed them on a multielectrode array, and measured the electrical signal from them. Using a complex electronic system and amplifiers, we used the signals as input for artificial intelligence. In the first stages of the project, I was looking for mimetic forms from nature to build a Biobot, but then we took as a new starting point the idea that an emergent form of Biobot could be built through an evolutionary algorithm. Artificial intelligence used signals from neurons and compared them to the different number of legs and links in arthropods to build the shape of the biobot, but you could also say the other way around, that signals from neuronal cells co-created certain shapes of the biobot with the help of artificial intelligence. We are talking about merging artificial intelligence with neuronal cell intelligence already at the level of conceptualizing the biobot robot itself.
Different movements generate different forms of the arthropoid robot. All my creative work has always been related to the movement of mechanisms, and I have always been attracted to complex dysfunctional or non-normative movements, whose aim is not to perform a task as well and productively as possible, but to give a certain disinterested poetics to their unusual gait, so I ultimately chose the more irregular which in society we would label them as handicapped forms.