An overview of modern software ecosystems enabling quantum computing is presented. The differences between ecosystems of quantum computing (EQC) and ecosystems of classical computing are highlighted. Provides a brief description of the most advanced quantum computing ecosystems (Google, IBM, Rigetti, Azure, Amazon, Intel, Strangeworks, IonQ). The current state of EQC, the possibilities of their use for various applications and development prospects are analyzed. Perspective areas of application of quantum computing in transport are considered.
1 Advanced countries, including Russia,
have adopted programs for the development of
technologies united by the general term
"Quantum computing" [
As noted in [
As shown in [
The term "Ecosystem" was introduced by
the English geobotanist Arthur Tensley in [
Over time, analogs of ecosystems have
appeared, denoting systems to ensure
functioning in modern society, such as
"Business Ecosystems". In particular, in [
Among software ecosystems, software
development ecosystems stand out. In
particular, [
The software development ecosystem should include the following main components: - programming languages; - tools for creating program code; - tools for building diagrams that display the structure of the program code (class diagrams); - software debugging tools; - sets of libraries and components for the implementation of algorithms, exchange and presentation of data.
For example, the Microsoft software
development ecosystem [
- programming languages C ++, C # and others;
- a set of development tools for MS Visual Studio, including, in particular, debugging tools, a class designer, a data structure designer; - numerous SDKs and libraries; - cloud environment for software development and deployment;
- Windows operating systems of various versions.
With the emergence and development of quantum computing, one can distinguish quantum computing ecosystems (QCE), a distinctive feature of which is the use of a quantum computing module and / or its emulator. These ecosystems provide access to quantum computing and serve to:
- manage existing physical quantum computing devices,
- evaluate the efficiency of the implementation of quantum algorithms on future devices, - study the concepts of quantum computing, - check quantum algorithms and their implementations,
- teach quantum computing.
Quantum computing consists of the following main steps: 1) data preparation, 2) preparing a quantum computing scheme, 3) loading a quantum computing scheme and raw data into a quantum processor, 4) performing computation in a quantum processor,
5) receiving data from a quantum processor, 6) interpretation of data.
On a conventional (classical) processor, steps 1), 2) and 6) are performed, steps 3) and 5) are associated with the transfer of data between the quantum and classical processors, and step 4) is performed on the quantum processor. When debugging software at stage 4), a quantum processor simulator is used.
Let's consider the most developed, at the present time, QCE.
Google is developing various tools built
into the company's extensive hardware and
software infrastructure [
Cirq works with quantum computing schemes. Quantum programs in Cirq are represented by "Scheme" and "Schedule", where "Scheme" represents the quantum circuit, and "Schedule" represents the quantum circuit with information about the sequence of actions. Programs can run on local simulators.
TensorFlow Quantum is a special library that allows Cirq circuits to be used as TensorFlow tensors, in addition, it contains specialized layers (tf.keras.layers) for recurrent neural networks. This library is an example of a mixed quantum-classical approach.
Open Fermion is a specialized library of algorithms for modeling chemical processes.
QCE Google provides access to its quantum 50-qubit computer and related simulators.
QCE developed by IBM is presented in
[
The QCE includes the IBM Quantum Composer tools for constructing quantum computing circuits and the IBM Quantum Lab development environment. The user is provided with the following main modules: - module for composing quantum programs at the level of circuits and pulses with optimization and taking into account the physical characteristics of a particular physical quantum computer (Terra);
- simulator of quantum computing and simulator of noise errors (Aer);
- subsystem of noise reduction in quantum circuits (Ignis); - library of quantum algorithms (Aqua).
QCE provides access to a line of proprietary quantum computers. At the time of this writing, the most powerful were a 27-qubit processor with a long characteristic lifetime of quantum states and a 67-qubit computer with a shorter characteristic lifetime of quantum states.
For debugging programs and researching algorithms in this QCE, access to various simulators is provided, with a capacity of up to 1000 qubits.
The Rigetti QCE is described in [
QCE Rigetti contains all the necessary modules to perform quantum computing.
Quantum operating system. Access to the quantum operating system is provided through network APIs. At this level, basic services are implemented, such as: - user authentication, service authorization, - control of the computation scheme and its transfer to the quantum processor, - memory management, -control of simultaneously-running processes.
This API is accessed using the Rigetti SDK software.
Quil instruction language [
ForestSDK programming tools include the pyQuil Python library and the Quilc optimizing compiler, which can be configured to create programs on non-Rigetti quantum processors.
A 31-qubit quantum computer is used to execute programs. Various noise simulation simulators are used for debugging.
Microsoft is developing a set of tools and
technologies [
This QCE does not have its own quantum computer. Access to Honewell and Quantum Circuits quantum computers is provided.
QCE, developed by Amazon, bears the
commercial name Amazon Braket [
Intel reports that work is underway to
create an QCE [
This ecosystem collects various quantum
services for access through a single entry point
[
QCE IonQ provides access to its own
quantum computer [
development and application of
The material [
It is assumed that with the development of QCE, software shells will develop that hide from the user the details of the implementation of quantum algorithms. The development of QCE will thus make it possible to solve applied problems for the solution of which the computational power of classical supercomputers is not enough.
With the development of quantum
computers and QCE, quantum computing
algorithms will also develop in applications for
optimization problems and for analyzing big
data. The importance of big data analysis in
transport is shown in [
Let us consider a possible approach to
solving the problem of optimizing transport
logistics using a quantum computer. The
problem of optimization of
MozheKantorovich logistics in the trivial formulation
of a linear programming problem can be
represented as follows [
n ∑ j xi, j = ai ,
m ∑i xi, j = a j , ∑ ∑ Ci, j xi, j → min,
i j where: m, n – the number of points of consumption and production of a homogeneous product;
ai – production volume at the i-th production point;
bj – the volume of production at the j-th consumption point;
Сij – costs of transporting a unit of goods between nodes i and j.
This problem, as has been repeatedly shown, is NP-hard, which means that there is no time-polynomial algorithm for solving this problem.
As shown in [
a=1 where z – an n-bit string and Ca (z) = 1 if z satisfies constraint a and 0 otherwise.
Considering the representation of n bits by
n qubits, in [
First things first, we need to confirm that
big data methods are really that important in
modern transport area. Initially the sign of big
data was a set of 3 Vs: volume, velocity and
variety [
Even the basic analysis of today’s state of transport is enough to say that it works with big data:
Volume. How it was stated before, transport now is under a huge digitalization, despite the fact that there is a lot to achieve yet. Thanks to that, lots of new sources of data were discovered. More sources mean more data. Today every transportation machine is equipped with tons of sensors, which data needs analysis. And this is only a few examples of data sources: data can be taken from logistical systems, road sensors, CCTV etc.
Variety. Sources themselves became more heterogeneous. Some stream bit data, some – text data. There are sources with video and audio data. It all depends on porpoise of source and possibilities of application its data.
Velocity. Modern data transmitted through streams. To provide relevant processing and analysis of data, systems must have enough power. This concerns transport too, where computation done in time can prevent delays or even accidents.
Some classifications add different signs of
big data, like veracity viability. One worth
mentioning is value [
Big data processing opens a door to many
new services and upgradeability of old ones.
These are only few of them [
- Monitoring of infrastructure. With bigger data amount, infrastructural analysis becomes more accurate, resulting in more effective management of infrastructure itself.
- Increasing in mobility of services. Today users need to have everything in their hand – on their smartphone. Transport services are no exception. Processing of big data grant a possibility of providing services in a useroriented way, also giving company additional data, which can be useful in future.
- Understanding client’s needs. One of the most popular sources of big data – social networks. Analyzing information from there can give a lot of useful business-details. First and foremost, this concerns client’s demands, which is important on transport as on any other area.
- Human flow visualization. Visualization is sometimes added to big data signs, which stresses its importance. Indeed, this is very powerful tool, that can solve a lot of tasks of transport management, also being simple and understandable to average person and not only IT-experts.
- Traffic control. This item speaks for itself. More sources and data results in more accurate prediction and more correct traffic control.
- Machine state diagnostics. This item is especially relevant on the railroad, because technical checkup and repairs are one of the most important parts of transportation on train, and they are not achievable without big data.
It is worth to mention Smart Transport
initiative in a Smart City idea. The are many
definitions of Smart City, but for the sake of
simplicity let’s focus on that is a city, where
all of main components of infrastructure are
most intellectual, connected and efficient
because of IT usage [
The initiative itself implies
intellectualization of transport in the city,
including community transport and roads.
These are main aspects of Smart Transport
[
- Smart Roads (changing road markings, accident detection, smart lights, sensors equipment).
- Smart traffic lights (data analysis to regulate traffic relatively to traffic density).
- Smart cars (processing data from various sources to provide autopilot and driver safety).
- Smart road signs (analysis of data about road state, sign changing).
In the introduction to this article, the problem of power of modern IT-systems is concerned. One of the solutions to this is quantum computers. Next, we give a short specification of quantum computing and its capabilities of processing big data. 7.
The main difference between quantum and classical computing is a presence of two physical phenomena, observed only in microworld: quantum superposition and entanglement.
Quantum computers operate not in bits, like classical PCs, but in qubits. The difference between them is in the amount of their possible states. Classical bit interprets the presence or absence of electrical current and can be either 0 or 1. Qubit interprets one of the properties of a particle (for example electron’s spin or photon’s polarization) and can not only be ∣0⟩ or ∣1⟩, but in a state of superposition of these values (d0∣0⟩+ d1∣1⟩), where d02 and d12 – possibilities of qubit transition to that state.
Quantum superposition phenomena is hard to describe, because in classical physics there are no suitable experimental examples. To put it simply, superposition is a state, where qubit is 0 and 1 at the same time. Unfortunately, this phenomenon cannot be seen or felt by a human. Every attempt to measure it results in qubit collapsing in one of the states with certain possibility. However, qubit in that state still can be worked with.
Its important to know, that qubit states are written in brackets because they are actually vectors. For example, vector ∣0⟩ can be written like 1 and ∣1⟩ can be viewed as 0 .
0 1
Second distinctive feature of quantum world – entanglement phenomena. Two or more qubits can transit into one superposition for all of them. Measurement of one of these qubits will instantly collapse superposition not only for it, but for every other entangled qubit. All of them will transit into one prescribed state. For example, if one will measure first qubit in state (d0∣00⟩+ d1∣11⟩) and get 0, then it certain, that second qubit is 0 too.
Entanglement phenomena is used in computations, but do not give any speedups. It mostly usable for the quantum security. From the other side, superposition can give that necessary push in computer power for big data processing.
For example, we can consider two bits and
two qubits. Two bits can potentially turn into
one of four states (00, 01, 10, 11). But in this
moment in time two bits can represent only
one of these states. Two qubits can also transit
into one of these states plus their
superposition. This means two qubits can
represent in this moment in time all of these
sates at once [
This brings us to an ideal model of quantum computing’s superiority over classical computing, represented on Figure 1.
We need to imagine some function and sets of income data for it. The task is to get every outcome value for every income. In classical computing we need to run function for every income value. In quantum computing, we can set all income values into superposition and run function only for this state. By this method we can compute function only once, instead of n.
Of course, this model is very abstract and
everything is not so simple on practice. There
are many pitfalls, preventing this ideal course
of things. To show this, we can look at already
implemented on a real quantum computers
model – Grover’s algorithm of quantum search
[
Detailed algorithm description is a separate topic. This is a glimpse on principal of its work. There is scheme of algorithm on Figure 2.
The circuit consists of quantum gates – orthogonal matrices, which apply to qubit states with tensor product. Also, circuit has oracle F – function, representing black box. It is also a orthogonal matrix.
This circuit is designed for two qubits. There is a function f(x0,x1) and 4 possible qubits states (∣00⟩, ∣01⟩, ∣10⟩, ∣11⟩). Function returns 0 or 1, while returning 1 only for one of possible states. We need to find this state.
On circuit entrance two qubits being put in state |00⟩ and one additional qubit in |1⟩ state. Let’s observe two upper qubits. This pair goes through Hadamard gate and transits into superposition with equal possibilities for each state. After oracle, one of the states changes its operator to ‘-’. This is right state. To get the answer, we need to increase its possibility and decrease possibility of other states. For this, superposition goes through A gate – a gate that flips possibilities over average value. At the end, needed state has possibility of 1, and any other has 0.
Unfortunately, Grover’s algorithm won’t be able to work so perfectly with 3 or more qubits. In this situation, after possibilities flip, the possibilities of wrong answers won’t go into 0. This problem resolves by repeating the circuit for out outcome data. Every time the wrong possibilities will decrease.
Grover himself computed a value of
needed repeats of algorithm for the most
believable answer. It is О(√ ) times [
Despite the dramatic decrease in time of computing, for the average person tis speedup may seem not so impressive. However, it shows, that quantum technologies already can be used in real tasks as more efficient way.
It needs to be remembered, that quantum
computing is in its dawn and many things are
simply not yet discovered. Quantum
computing’s potential is well shown by Shor’s
algorithm [
It is worth to mention a quantum machine
learning area. ML is vastly used in big data
analysis and its speedup would dramatically
affect this domain. On picture 3 there are few
ML methods, which potentially can be
accelerated by quantum computing [
As you can see some methods receive exponential speedup. However, there is still a lot work to do to get an actual value. 8.
An important difference between QCE and ecosystems of classical computing is the large role played by the scheme of quantum elements in quantum computing.
Various emulators of quantum computers with support for emulation of different types of noise are used for debugging.
QCE contains the same basic modules as classical computing ecosystems.
Specific features of ECU: - Quantum computing scheme development environment.
- A set of components for accessing a quantum computer.
- Emulator of quantum computer. - Quantum computer.
The QCE types by mode of development: 1. On advanced classical computing ecosystems basing such as Microsoft and Google.
2. Completely newly developed systems such as IBM and Rigetti.
3. Some QCE collect a variety of services: software libraries, emulators and quantum computers, for example, Strangeworks.
In terms of application, Google QCE is distinguished, containing a developed toolkit of the machine learning system TensorFlow Quantum.
Google, Azure Quantum, IBM and Rigetti QCE can be used for training purposes, providing a convenient user interface and advanced documentation with examples.
It is proposed to consider as promising
tasks for further research on the possibilities of
using quantum computing ecosystems:
- applied use of QAOA [