CC BY
8INTERNATIONAL SCIENTIFIC AND TECHNICAL CONFERENCE "DIGITAL TECHNOLOGIES: PROBLEMS AND SOLUTIONS OF PRACTICAL IMPLEMENTATION IN THE SPHERES" APRIL 27-28, 2023
SMART ROBOT TRANSPORTATION SYSTEMS Mukhamedieva Dildora1, Mirzayeva Nilufar2
1Tashkent University of Information Technologies named after Muhammad al-Khwarizmi,
Professor of the Department "Software of Information Technology", 2 TashkentUniversity of Information Technologies named after Muhammad al-Khwarizmi, Assistant of the Department "Software of Information Technology" https://doi.org/10.5281/zenodo.7856176
Abstract. We are already witnessing the implementation of innovative technologies and concepts, with unmanned vehicles being one of the most fascinating, promising, and widespread technologies. Furthermore, the development of this technology has the potential to revolutionize other areas, including training vehicles to communicate with each other, recognize objects in their environment, and use the internet, among other things that drivers can do. By doing so, unmanned vehicles could eliminate human errors such as driving under the influence, speeding, reckless driving, cutting off other drivers, violating traffic laws, and other dangerous behaviors.
Keywords: car communication, intelligent vehicle; road conditions; traffic; driving assistance, expert systems.
INTRODUCTION
By implementing effective traffic regulation, various issues can be resolved such as reducing wait times at intersections, minimizing fuel consumption and wear and tear of vehicles, decreasing harmful emissions, enhancing the environmental conditions of cities, and promoting the mental well-being of urban residents. Nevertheless, intelligent intersection control is expensive, intricate, and time-consuming. A new method of optimizing traffic flow is needed, which involves informing drivers in real-time about the status of traffic lights and allowing them to adjust their driving accordingly.
The availability of smartphones, navigators, and GPS-equipped devices allows a significant percentage of drivers to use the "Mobile Driver Assistant" system, which is discussed in chapter three. Vehicles equipped with this system can be perceived as autonomous agents in the road network, and city traffic can be regarded as a multi-agent environment. By acquiring knowledge of each agent's location, direction of movement, and speed, a new type of traffic control can be implemented.
As mentioned earlier, the development of artificial intelligence in various fields of human activity will require uncovering universal decision-making mechanisms aimed at automatically narrowing down search results based on a specific initial situation. This solution to the problem is expected to take many years to come.
A complex task may require expertise from related areas, prompting a specialist to seek advice from a relevant expert or a specialized automated "expert" system. The knowledge possessed by specialists can be categorized into formalized and poorly formalized. Formalized knowledge is presented in books and manuals in the form of laws, formulas, models, and algorithms. It is typically associated with exact sciences like mathematics, physics, chemistry, and astronomy. Descriptive sciences, on the other hand, usually operate with poorly formalized knowledge. Examples of such sciences include zoology, botany, ecology, sociology, pedagogy, medicine, and others.
INTERNATIONAL SCIENTIFIC AND TECHNICAL CONFERENCE "DIGITAL TECHNOLOGIES: PROBLEMS AND SOLUTIONS OF PRACTICAL IMPLEMENTATION IN THE SPHERES" APRIL 27-28, 2023
This passage describes expert systems that use probabilistic estimates and weight coefficients to provide solutions to problems. These systems have the ability to explain their solutions in a language that is easy to understand by the end user. The expert system operates in two modes - knowledge acquisition and problem solving/consultations. In the knowledge acquisition mode, a knowledge base is formed, while in the problem-solving mode, communication with the expert system is carried out by the end user. The system uses fuzzy mathematics algorithms to process knowledge with confidence coefficients, which may be adjusted during trial operation.
The learning process of an expert system can be automatic using a learning algorithm or with the help of a cognitive engineer acting as a teacher. The passage also discusses the two most important aspects of an intelligent transport traffic management system, which are ATMS (Advanced Traffic Management Systems) and ATIS (Advanced Traffic Information Systems). Each active element in the transport system has a behavior model determined by its function.
The ITS (Intelligent Transportation System) comprises of both active and passive elements, which include transport infrastructure facilities with instruments for signal measuring, transmitting, relaying and receiving; tools for remote monitoring and measurement; elements of the transport complex's information and telecommunication infrastructure; communication and remote monitoring tools equipped in vehicles and cargo; and remotely controlled devices, components and assemblies for indication and actuation.
The ITS technological complex incorporates several technical systems and tools, mostly forming a feedback channel with human operators and controlled technical components of the transport system.
These systems and tools include coordinate-time and meteorological security systems, communication and data transmission networks and lines, remote monitoring tools, information collection, processing and accumulation systems, automated control systems, display and communication tools, and other software and hardware technical means. The ITS's functionality and consumer characteristics of information services offered by ITS depend on several main parameters.
Currently, only a small percentage of the predictive capabilities of data obtained through remote monitoring has been fully mastered, and their economic value is not fully realized by consumers. Intelligent Transportation Systems (ITS) integrate data, video, and voice over IP networks to create comprehensive vehicle management and control systems.
These systems include communication systems and technological electronics control, and they are utilized to monitor and manage the flow of vehicles, reduce congestion on highways, and provide alternative routes during traffic congestion. Ultimately, the goal of these systems is to improve safety for all road users, increase reliability, and reduce time and costs associated with transportation.
The management of vehicles involves various subsystems such as an alarm control system, information display system, real-time video surveillance system, analysis system, and power system. With the use of intelligent embedded motion controllers, LED panels with traffic lights and warning systems, vehicle detection detectors, and surveillance cameras, different road functional groups can be identified.
INTERNATIONAL SCIENTIFIC AND TECHNICAL CONFERENCE "DIGITAL TECHNOLOGIES: PROBLEMS AND SOLUTIONS OF PRACTICAL IMPLEMENTATION IN THE SPHERES" APRIL 27-28, 2023
The vehicle detectors collect information on traffic flow, such as the number of vehicles, their speed and location, and transmit this wirelessly to the smart embedded computer. The embedded computer, as part of the traffic controller, analyzes this data to determine the state of the traffic flow and sends control signals to the traffic lights and warning boards to ensure a smooth traffic flow and avoid congestion. The traffic controller can also inform drivers of traffic jams and suggest alternative routes to avoid them. The video surveillance system is also part of this setup.
The purpose of the video surveillance system is to monitor and regulate high-risk areas of a road. It enables both real-time and periodic monitoring of the area with the option to save the video data for future reference. The system also allows for live transmission of camera footage to a monitoring center via digital channels, and the use of the Streamlogic compression algorithm allows for efficient use of storage space and wireless data transmission.
Cities' social, economic, and environmental achievements and problems are largely determined by their transport systems. Ground modes of transportation make up over 95% of urban transportation, and improving their efficiency and safety is critical.
Overloading of transport networks is a major issue, resulting in negative consequences such as excessive traffic volumes. Various approaches based on the patterns of traffic flow are used to solve this problem, but many aspects of traffic flow formation remain insufficiently studied. Modernizing external means of organizing traffic can be expensive, so it may be more effective to optimize driver behavior while driving.
This can be achieved using mobile devices' existing infrastructure, allowing for a new approach to traffic flow optimization. Various options for optimizing driver behavior and traffic flow have been proposed, including a system with intelligent traffic lights and cars equipped with wifi transmitters, and a system that uses a smartphone's camera to determine the optimal speed to overcome intersections. AI is already being used in road transport. For example, Google and Yandex use their algorithms to update road maps as soon as the company receives a photo with a new object in a given area.
AI-based autopilots are currently being developed. Neural networks allow you to recognize objects around the car, as well as calculate the trajectory of movement from the marking image. In total, there are 6 types of autopilots in the automotive industry: - Level 0, the driver does all the work. - Level 1, "hands on". The system assists the driver in driving. For example, cruise control, adaptive cruise control and automatic parking. - Level 2, "hands off". The system itself controls the car, the driver monitors the correct operation of the autopilot and is ready to intervene. - Level 3, "eyes off". Same as level 2, but the driver is not required to react immediately, he needs to intervene within the time determined by the system. - Level 4, mind off. Does not require constant attention. Automatic control is carried out in certain geofences or situations. - Level 5, "steering wheel optional". No human intervention is required.
Theoretically, the drones of Yandex and Waymo from Google can go to the 5th level, but no one is ready to give absolute automation to their cars, as this entails a huge responsibility. Now companies are developing in the field of driver monitoring systems. For example, Attention Assist from Mercedes-Benz or DSM project SOWA.The main principles are the assessment of the driver's behavior and the fixation of the gaze and tracking the state of the eyes. The face recognition system highlights the eyes, nose and mouth in the image, captures the blink rate and other features of the face, and informs the driver about changes in his condition.
INTERNATIONAL SCIENTIFIC AND TECHNICAL CONFERENCE "DIGITAL TECHNOLOGIES: PROBLEMS AND SOLUTIONS OF PRACTICAL IMPLEMENTATION IN THE SPHERES" APRIL 27-28, 2023
Application of a fuzzy control module to solve the problem of car parking Let us apply the modified module to solve the same problem fuzzy control with the possibility of learning. In the first phase, the structure of the module was chosen. For the first input signal - the position of the car on the x axis - it is proposed to use five membership functions, and
for the second input signal - the angle $, under which the car is located to the y-axis - seven membership functions. As a result of the generation of fuzzy rule messages based on a pairwise comparison of all membership functions of the first and second variables, 5x7=35 rules were obtained. The fuzzy control module created on the basis of these rules is shown in Fig. 1. The centers of the membership function of the output variable (the angle of rotation of the car wheels
@) placed at point 0, which is equivalent to the absence of conclusions (conclusions) in the rules. The initial membership functions formed in this way were used in the fuzzy control module by setting the corresponding initial values of the parameters and weights.
During the first 38 epochs, the network was trained using a standard
n = 0 25
backpropagation algorithm with a learning rate For the next 70 epochs, the network
trained according to the same algorithm, but the value of the learning coefficient automatically
n = 01
decreased (to the level ).
The membership functions corrected in the process of learning are shown in Figs.
3.
Fig. 1. Initial membership functions: a) for the position of the car x, b) for the angle of the
$
car
INTERNATIONAL SCIENTIFIC AND TECHNICAL CONFERENCE "DIGITAL TECHNOLOGIES: PROBLEMS AND SOLUTIONS OF PRACTICAL IMPLEMENTATION IN THE SPHERES" APRIL 27-28, 2023
Fig. 2. The trajectories of the car from three initial positions: (x, (f) = (-100, - 600 ) (l00,1200 )u (o, 1800 )
The correctness of the network functioning was checked by modeling for three different starting positions of the car. The motion trajectories for each case are shown in fig. 3.
They use rangefinders (laser, ultrasonic), gyroscopes. The most interesting AI systems are object recognition systems. Such systems can use convolutional neural networks (CNN) (Fig. 3), Haar masks and RNNs (Recurrent Neural Networks). Convolutional neural networks use convolution algorithms, their result and convolution kernels are shown in Fig. 2.
Fig. 3. Examples of kernels and results of convolution over them Prospects and opportunities for AI in transport On public transport, it will be possible to completely abandon drivers, because automated systems will be able to perform all their duties. Also, AI is able to select the optimal number of public transport on the line, which will reduce fuel costs and environmental damage. Ocado has
INTERNATIONAL SCIENTIFIC AND TECHNICAL CONFERENCE "DIGITAL TECHNOLOGIES: PROBLEMS AND SOLUTIONS OF PRACTICAL IMPLEMENTATION IN THE SPHERES" APRIL 27-28, 2023
optimized one of its warehouses by implementing warehouse robots that are controlled by a single system to avoid robot collisions. A similar system can be implemented in rail transport.
REFERENCES
1. Luger, J.O. Artificial intelligence: strategies and methods for solving complex problems / J.O. Luger. - M.: Dialectics, 2016. - 864 p.
2. Nilson, N. Principles of artificial intelligence / N. Nilson. - M.: Radio and communication,
2015. - 373 p.
3. Russell, S. Artificial intelligence: a modern approach / S. Russell, P. Norvig. - M.: Williams,
2016. - 578 p.
4. Slagl, J. Artificial intelligence / J. Slagl. - M.: Mir, 2016. - 320 p.
5. Tei, A. Logical approach to artificial intelligence / A. Tei, P. Gribomon, et al. - M.: Mir, 2015. - 432 p.
