The most complex constructions have been created by a team of engineers and scientists at the National Physical Laboratory (NPL) in the US.
The team, led by the former head of the National Robotics Institute (NRI), Dr Robert Bixby, built the robots from scratch using the latest computer science and robotics technology.
The robot was able to work autonomously, without human intervention, and was capable of moving at speeds of up to 100 kilometres per hour.
“It’s a lot of work, but it’s done with a very simple design,” Bix by way of explanation told Next Big Futures.
“The goal was to build a robot that learns to learn.”
Building a robot from scratch The team started by creating the robot with a single, 3D model, which allowed them to create a 3D template that could be easily built out of plastic.
The model was then converted into a 3d printer that could create a new 3D version of the model from scratch.
“In this process we used an array of sensors and actuators to create what is known as a ‘dynamically generated robot’, which is a robot designed to learn from its environment,” Brix explained.
“Once we had the 3D design we built an artificial intelligence algorithm to learn the design from its experiences.”
After the algorithm learned, the robot then built its own environment.
The artificial intelligence algorithms worked in tandem with the designers to create an artificial “learning environment”, a network of robots that mimicked the way that human brains work.
After learning how to do this, the team could then turn to their next step: building the robot’s neural network, which the scientists call the “neural network”.
The team used a method called reinforcement learning to build the neural network.
Reinforcement learning involves a computer learning how the environment is connected to its environment by means of feedback loops.
The neural network is then able to respond to the environment in a way that is not necessarily the same as the environment it was built from.
The researchers then built a new neural network on top of the existing one, and used it to build their robot’s entire artificial intelligence system.
After building the neural model, the researchers used a software program called LIGO to detect and classify objects in the environment, including the object’s location, speed, direction and position.
“This is a huge breakthrough because it is able to infer the meaning of an object from its surroundings,” BIX said.
“We also know that the objects we have seen in the past were not real objects, but were a combination of digital images captured in a real environment.”
The researchers have also used their model to create more complex and sophisticated artificial intelligence systems, such as the “super-computer” of the NRI.
The system is able for example to recognize a number of different shapes and sizes.
The future is bright, but the team are working hard to find out how to apply this technique to their robots, before it is ready for commercial use.
Bix is also hoping that the technology will eventually be applied to a new type of computer chip called the “universal supercomputer”.
The researchers believe that it is possible to build computers that are able to run on just about anything and could have a wide range of applications.
“With this kind of technology we can be creating a new era of computing,” Bax told Next Future.
“By combining the capabilities of supercomputers with our own computational capabilities, we can create the computing systems that are needed for the future of the human race.”