1. The beginning of the research (around 2015)
Can we construct machines that exhibit the flexibility of living organisms? Or can we engineer micro-nanomachines that mimic microorganisms? These are questions that have captivated my mind for quite some time. To realize such machines, I posit that the creation of artificial “machine cells” and the subsequent construction of robots through the self-organized assembly of these machine cells is paramount. Robots constructed in this manner are hereby referred to as “bio-robots.”
Bio-robots, akin to living organisms, arise from the aggregation of artificial cells (modules) and the differentiation of these modules into a unified morphology. This ensemble of modules then operates as a cohesive unit, executing the desired behavior. Bio-robots possess the remarkable ability to self-repair by replacing malfunctioning modules and can even self-replicate to produce identical replicas.
What are the fundamental functionalities required for the machine cells that constitute a robot with such capabilities? And what governing principles should orchestrate the interactions of these machine cells to give rise to bio-robots?
In my laboratory, we are initially developing an emergence model for bio-robots. Our ultimate goal is to establish an emergence simulator for bio-robots.
Figure 1 Image snapshot of Bio-Robot
2. Current Research
(1) Proposal of a novel model generating more lifelike movements (2018)
I proposed a Turing pattern model (Young model) augmented with life-game properties, henceforth referred to as the Ishida model. Extremely simple transition rules enabled the reproduction of cells exhibiting movement and division.
Figure 2 A self-replicating model that combines the Turing pattern model and Conway’s life game.
[1] Ishida, Takeshi, Possibility of controlling self-organized patterns with totalistic cellular automata consisting of both rules like game of life and rules producing Turing patterns, Micromachines, 9, 339, (2018), https://doi.org/10.3390/mi9070339
(2) Development of a Model that Generates Turing Patterns Solely through Numerical Exchange with Adjacent Cells (2020)
In the Ishida model, determining the state of a single cell requires tallying the states of all cells within a certain radius of surrounding cells. Implementing such a model in an actual modular robot would necessitate intricate information transfer. To address this challenge, we devised a model that generates Turing patterns solely through numerical exchange with neighboring cells only.
Fugure 3 Schematic of Ishida’s numerical value aging model. A 2-D hexagonal grid consists of two-state cells (black and white). a) Modification process of tokens generated within the black cells. The token values are increased up to X times within each cell, whereas some tokens remain unmodified in their respective cells. The frequency distribution of token numbers ultimately determines the subsequent state of each cell. Variable X represents the maximum value associated with a token and is a positive integer. b) Calculation example of a generated Turing pattern.
[2] Takeshi Ishida, Emergence of Turing Patterns in a Simple Cellular Automata-Like Model via Exchange of Integer Values between Adjacent Cells, Discrete Dynamics in Nature and Society, Volume 2020, Article ID 2308074, 12 pages (2020.1)
https://doi.org/10.1155/2020/2308074
(3) Hierarchical Morphogenetic Model Construction Based on the Ishida Model (2021-)
Utilizing the Ishida model and parameter adjustments, we can induce the emergence of a stationary circular pattern. To further explore morphogenetic complexity, we constructed a hierarchical model in which the Ishida model is iteratively applied to the circular pattern. By layering three to four model iterations and fine-tuning parameters within each layer, a diverse array of forms can emerge. (Paper in progress)
Figure 4 Example of calculation results of morphogenetic model
(4) Development of a Control Model for a Modular Swarm Robot: An Integration of Accumulation-Action and Self-Replication Models (2024)
Leveraging the model presented in (2), I constructed a simulation model of a modular swarm robot. I successfully simulated a group of modules that aggregated and behaved as a cohesive unit solely through local interactions among neighboring modules. This model exhibits the remarkable ability to navigate towards a light source or traverse a gap in a wall without the need for instructions from external or distant modules, a leader, or a coordinate system.
Figure 5 Example of simulation of swarm modules passing through a gap in a wall toward a light source
[3] Takeshi Ishida, An algorithm applied the Turing pattern model to control active swarm robots using only information from neighboring modules, arXiv, http://arxiv.org/abs/2405.17868
3. Future Directions
(1) Research Phase I: 2024-2027
Goal 1: Method Development and Simulation of Robot Self-Organization (2024-2025)
Goal 2: Validation by Prototyping Self-Assembling Robots (2025-2026)
Goal 3: Development and Release of Bio-Robot Design and Control Simulator (2026-2027)(Aim to establish a start-up)
(2) Research Phase II (Artificial Cell Design): 2028-
Development of Simulator for Self-Replicating Artificial Cell Module
(Cellular emergence model has already been established.)