News

Light-Based Computers Used to Find Solutions to Complex Logistics Equations

November 06, 2016 by Robin Mitchell

Researchers at Stanford have developed a computer that can solve specific problems using magnetic fields and lasers.

Light-based computers have been dreamed about for decades as an alternative method to conventional computing. Researchers at Stanford have developed a computer that can solve specific problems using magnetic fields and lasers.

Specialised vs Generic

Modern conventional computers use electrical current controlled by billions of transistors integrated onto silicon wafers.

One major factor as to why modern computers are so popular is due to their “genericness”. In other words, modern computers are not designed for any specific problem and thus are only constrained by their programming. This means that the microchip which powers your washing machine cycle could also be programmed to run an operating system, play a game, or even control your automatic IoT cat feeder. But this is not without its drawbacks as computers controlled systems may not be as efficient as they could be for each situation that they are employed in.

One common example is the two main CPU architectures that exist, von Neumann and the Harvard. The von Neumann architecture relies on large memory models with program code and variables being stored in the same linear memory address. This type of memory arrangement is very useful in programming models that rely on large memory requirements (such as games and simulation software) or the ability to run code that can be dynamically placed into memory (for example, when you launch an application your OS loads that program into RAM and then executes).

The Harvard architecture has its variables and program code separated into different memory addresses so that a variable and executable instruction could have the same memory address but be stored in separate memory locations. This type of configuration is ideal for use in embedded systems where the program code rarely needs changing and there are few variables needed.

Even then, these two architectures still both provide generic development environments which may still not be fully optimized for their task. This is why there are devices that are designed for specific signal processing (such as ASIC devices and FPGAs). This specialization, while more expensive, can perform a specific task much faster than generic devices. This is the it direction that light-based computers may take.

The Travelling Salesman Problem

Just like quantum computers, there is a misconception that all computers will be replaced by quantum and light-based devices in the future. In reality, these devices may be in devices as a co-processor but they will be used to perform very specific operations such as encryption or pathfinding. This is exactly what the light-based computer, designed by a research team at Stanford University, does.

Before we look at how the researchers designed the device, we need to look at the problem that the computer actually solves.

In computer science, there exists a problem known as the “Travelling Salesman Problem”. Imagine you are a salesman and you need to visit all the cities in the region so as to maximize your profit. It does not matter which order you visit each city but there are two aspects to the journey that you need to comply with:

  1. You must end where you started from
  2. Each city can only be visited once
  3. You must pay for the journeys with your own money, so your travel distance needs to be minimized

 

Travelling salesman problem showing nodes and paths. Image courtesy of Nojhan (own work) [CC BY-SA 2.5-2.0-1.0]

 

Simply put, the solution will be a path where the cost is minimized to visit each city while ending back where you started. This problem can be found in many real-world contexts including delivery companies, reducing Wi-Fi interference, and even protein folding.

Conventional computational machines can solve for up to 85,950 cities—but there is a problem with sizes bigger than this. The number of all possible paths for n nodes (cities) is equal to an (n – 1) factorial which is small for a few nodes. However, when n becomes much larger, the number of possibilities becomes so large that it would take the entire age of the universe to solve for the most efficient one.

Clearly, a generic computational device cannot handle such a problem.

Light-Based Computing

Researchers at Stanford University have created a light-based computer that uses pulses to solve the travelling salesman problem. The computer that has been designed is called an Ising machine, which essentially acts as a reprogrammable magnetic network.

Each “artificial magnet” in the network represents a node in the traveling salesman problem and can have one of two orientations of magnetism: high-energy or low-energy direction. As the system as a whole tends to the lowest energy state, the problem is solved and the state of the magnetic nodes indicate the solution.

 

Researchers, Peter McMahon and Alireza Marandi. Image courtesy of L.A. Cicero via Stanford University.

 

In the light-based device, a laser called a degenerate optical parametric oscillator that uses pulses of light to represent nodes in the problem. The laser also has spin which represents the magnetic state of each node. The initial issue with the system was coupling pulses together (so that the problem could be solved) which was achieved using an expensive method of controllable optical delay.

But the new system uses a digital circuit to emulate the optical connections among the pulses which are used to program the problem. The laser system, in conjunction with the digital circuit, can still solve problems and has been used to solve 100-variable problems with any arbitrarily set of connections between variables.

One major advantage of the current design is that nearly all the parts for the computer are off-the-shelf components. This is crucial for commercial interest. One fact that cannot be disputed is the historical relationship between innovation and commercial drive where inventions and discoveries that can be turned into products tend to be much more quickly funded and developed. 

The Future Of Light?

In some cases, electricity propagates faster through a wire than light does through an optic fiber cable. But the number of times that an electrical signal can be switched on and off is much lower as compared to light—e.g. a telephone wire can transmit 3000 calls whereas fiber-optic cables can handle 31,000 calls.

Light is also virtually immune to electromagnetic interference whereas metal wires pick it up and consequently weaken the original signal. Light also has quantum properties which is beginning to be exploited in computing systems including a recent development with trapping silicon atoms in diamond to store quantum information.

 

Fiber-optic cables are quickly replacing metal cables in telecommunication. Image courtesy of blizzy78 [CC BY-SA 2.0]

 

Considering the efficiency of light-based systems and their bandwidth capabilities, it comes as no surprise that scientists are becoming increasingly interested in exploiting this fundamental particle to shed light on future technology.