image credit : freepik.com
The Subject of discussion: Robot Evolution and its various stages
What is Robot Evolution?
From the Egyptian water clock found at the tomb of Amenhotep I to Perseverance rover successfully landing on Mars, from Talos being the first humanoid robot from the Greek era to Boston Dynamic’s Atlas performing backflips and well-choreographed dance routines, and from Leonardo da Vinci’s mechanical men to Boston Dynamic’s Spot carrying out coordinated tasks with utmost precision, robotics has come a long way.
Facts of Robot Evolution
A robot is typically a machine designed to carry out routine or complex tasks through programming by scientists and engineers.
Primitive robots started with simple time-telling and performing routine tasks only, with limited or no human intervention. Human beings’ curiosity has been evident in the fantasy worlds of science fiction created over the years. These ideas broadcasted in reel life have highly inspired robot evolution in real life.
Robot Evolution: Evolutionary Robotics
How is Evolutionary Robotics defined?
Consider how human beings have evolved from the earliest known species of the Hominid family, displaying significant anatomical changes- Homo Erectus. There are precise shreds of evidence, majorly in the aspects of bipedalism, brain expansion and culture, which has brought Homo Sapiens to where they are today and these changes are influenced by natural selection.
Similarly, Evolutionary Robotics is an embodied approach to achieving artificial intelligence through selection, variation and heredity principles in robot design. This is a subset of Robot Evolution that leads to the innovation of more robust and adaptive robots. This leads to the optimization of a robotic system or one of its subsystems against a behavioral goal by exploiting its design features using a holistic approach.
Centred on a new form of experimentalism, evolutionary robotics can also be used as an approach of revolutionary approach for studies. Using robotics as an activator can help solve problems that are difficult to analyze, if not impossible, by computer models and practical experiments.
How does Evolutionary Robotics occur?
An experiment in evolutionary robotics considers randomly generated samples of robot designs pertaining to individual characteristics. The worst-performing variants are eliminated and replaced with better template mutations and/or combinations. This evolutionary algorithm continues until the completion of a specified period of time or exceeds any target output parameter.
For engineering devices that must work in situations in which humans intuition is not in the best of its use, evolutionary robotics techniques can be exploited well. It is also possible to use sophisticated simulated robots as scientific instruments to produce new biological and cognitive science theories and evaluate old hypotheses involving tests that have proved challenging or impractical.
When did Evolutionary Robotics begin as an experiment?
The experimentation on evolutionary robotics began on robot control systems in 20th century Europe under Dario Floreano and Francesco Mondada at EPFL. They started experimenting with the control system of the Khepera Robot. This was followed by the evolutionary robotics experiment on the Gantry robot’s control system at the University of Sussex. There weren’t many significant hints of evolution on the hardware body of these robots, though.
MIT Media Lab conducted the first simulations of their evolved robot designs in the late 20th century itself. However, they never turned into practical machines. The first actualization of an evolved robot design has been due since the turn of the 21st century.
Can robots really evolve?
Natural evolution is an open-ended process and achieving success with witnessing such a regime in an artificial machine still remains a challenge. The said process is defined with constant morphological and behavioural advancements, and that requires a human-like brain for the exhibition. The non-existence of a human-like brain limits these artificial machines in the intuitive decision-making process. Even if it is continuously thought of developing a brain as smart as a human, it is first necessary to understand what principles are involved in this and how to embed them into the artificially produced brain.
But the question lies in why indeed, is it necessary to create an open-ended evolutionary process for robots. To this, scientists from various fields have a wide range of viewpoints. Biologists believe that this feature will bring more accuracy into the models studying the evolution of human beings by maintaining the complexity and criticalness of the features in the original findings. Simultaneously, engineers believe that these features will enable better and efficient solutioning of the complex engineering design problems.
What can we expect from Robot Evolution?
Here, it is discussed with what expectations are the scientists approaching towards experimenting on Robot Evolution.
- Coupling of mind and body to reach equilibrium– There is a balanced complexity between the mind and body, and such a compromise between the two could significantly simplify future robotics. Evolutionary Algorithms are stable models of natural selection that can thus cope with morphological changes and have the capacity to generate structures allowing them to develop both controllers and morphologies.
- Emergence of intelligence– Designing systems with intelligent behaviour influenced the initial development of genetic algorithms. Intelligent action is a composite capacity to forecast one’s world in the light of any purpose, paired with a transformation of each prediction into an adequate reaction. This makes the artificial system more adaptive towards the environment.
- Social behaviour evolution– Social behaviour evolution asks whether typically necessary conditions and ability to interact inter-species and intra-species exist or not.
- Proof of existence– Robot evolution being an unbiased search process, can be used to recognize better solutions that a human being would neglect or never think of.
Tools and Methods of Design
- Morpho-functional machine design– It requires individual development of morphology, apparatus, and the control system design in a traditional manner. This approach is simple in the sense that scientists with expertise mainly in each of the fields can work towards developing them parallelly. Although, exploiting the synergies between these parts become difficult. In contrast, robot evolution depends on assessments of the entire robot in contact with its environment.
- Representation of solutions– Robot evolution requires a representation of the solutions to be examined to be described and representation should be consistent with and provide solution of interest, random generation, mutation, and cross over. In particular, finding suitable representations in evolutionary algorithms, in general, is one of the most critical problems. Many direct and indirect encodings have been proposed for exhibiting such modularity.
It is not feasible to optimize and repair robot features prior to launch since the operating requirements are not well understood in advance and/or change over time. The conventional solution will not work in these circumstances, and different kinds of selection-reproduction schemes are required.
Concerning the evolutionary algorithm, such frameworks have a variety of remarkable properties. Firstly, there are two focused evolutionary goals: utility and viability. Another notable feature is that the relationship of the robot evolution is to involve the evolutionary algorithm for the further upgradation of robot evolutions.
Will robots be able to reproduce?
To answer this question, we first must understand what we already know. There have been theories developed about self-replicating autonomous machines, although, to our good, they have not been realized yet wholly. Who is to say that such machines will work only towards human’s benefits? The day they advance to a level where their brain can control their bodily functions, they may decide not to be assistants to humankind anymore and develop their own indigenous species instead of as fully functional and self-reliant creatures.
The theory of self-replicating machines states that a self-replicating machine autonomous in nature can use raw materials in the environment to replicate itself autonomously, thereby demonstrating self-replication in a manner similar to that found in nature. This has been theoretically demonstrated in the Von Neumann probe. Von Neumann also worked on what he called the universal constructor, a self-replicating system that he formalized in an automatic cellular environment that could evolve.
A variety of proposal such as mining of moon and formation of lunar factories, and solar powered satellites example of such technologies.
In 2005, Cornell University researchers created a machine that can build copies of itself. Although this advancement is evidently marked towards evolutionary robotics, it doesn’t really do anything better other than self-replicating. This surely doesn’t fulfil the purpose of robot evolution and is just a proof of concept.
Are human beings self-replicating robots?
The researchers find out that because the offspring are not identical clones, human beings regenerate but do not actually self-replicate. And the ability to reproduce depends on the context in many situations. Rabbits, for example, are decent woodland replicators, bad desert replicators and abysmal deep space replicators.
Although, there exists a separate group of argumentative theories that state otherwise. Hence it’s more of a debatable subject. Some believe that A, C, T, G are the equivalent of the binary’s 0 and 1 of the code of our genes. They encrypt some data when you encode letter alphabet computers with a binary value that encodes human genes with those four molecules.
They also believe that pain and emotion are also coded by evolution in the same manner that you start printing when you click a button on the keyboard. Hence to them, human beings are self-replicating robots only with the capability of having trained better.
I believe in the formerly discussed concept personally. The analogy generally comes from the fact that you might assume that our mind is like software and that hardware is like our bodies. That works well on a symbolic level and maybe under some primitive paradigm of conductivism, but to translate that, one needs to simplify a lot on a realistic level.
Both software and language are ways of transmitting data, but a punch in the face is also a way of transmitting the information. On that basis, the software is a language. Proper human language is a thing that does not suit well under the definition of computing. It is never done, it is never here, it is never direct, it works anyway, it has no glitches or faults, and it is shared to some degree. It need not be coherent.
- Aircraft Fuel Tank System and its 5+ Important Subsystems
- Aircraft Fuel Tank – 10+ Important Features and Design Characteristics
- Aircraft Fuel Pump – All concepts you need to know and 5 Important Classifications
- Robot Sensors – All concepts you need to know and 5+ Important FAQs
- What is Pneumatic Gripper? | 5+ Important Applications and Advantages