AU2020103161A4 - An novel method for a smart car - artificial intelligence based autonomous steering control system with voice assistance - Google Patents

Abstract

To minimize road danger caused by minor driver failures, the idea of an intelligent steering system was developed. In an emergency, the smart steering system manoeuvres the steering wheel, which otherwise may lead to a vehicle accident or obstacle on the lane. Intelligent steering systems are an integral part of smart parking, lane management and other collision prevention systems. Depending on the rising demand of automatic systems for vehicles, order for the intelligent steering systems is expected to increase in the projected period. Increased social purchasing power parity and a young generation's preference for automated vehicle systems are expected to be the main drivers of the intelligent market for control systems. Features such as voice inputs and hand movements from sophisticated, intelligent steering systems, make it extremely appealing to customers. The production of self-employed cars is anticipated to hamper the demand for the smart steering system by eliminating the steering wheel from the driver's cab. A technology-based, steering wheel type, input technology, vehicle, and region are the basis of the intelligent steering system market. The smart steering system market can be segmented into RADAR and LiDAR-based on technology. A large proportion of the smart steering system market is possibly in the RADAR market. Due to its economic quality, it is preferable in most automobiles. These devices are used for the identification by the steering actuator of obstacles along the road of the vehicle that controls the steering wheel of the vehicle to prevent a collision. x~fobt Maeag w"SWw rednavwkn tAMa Fig: Autonomous Vehicle by using multi sensor .... M .......... - ..w Fig: Al model for autonomous vehicle including data collection, planning and act

Description

x~fobt Maeag

w"SWw

rednavwkn

tAMa

Fig: Autonomous Vehicle by using multi sensor

.......... .... - wM ..

Fig: Al model for autonomous vehicle including data collection, planning and act

Editorial Note 2020103161 There is 11 pages of Description only.

TITLE OF THE INVENTION ARTIFICIAL INTELLIGENCE-BASED AUTONOMOUS STEERING CONTROL SYSTEM WITH VOICE ASSIST APPLICANT

The following specification particularly describes the invention and the manner in which it is to be performed

ARTIFICIAL INTELLIGENCE-BASED AUTONOMOUS STEERING CONTROL SYSTEM WITH VOICE ASSIST FIELD OF THE INVENTION

Artificial Intelligence ( Al) is a machine intelligence tool which gives massive potential to

intelligent revolution in the industry. It enables data/information collection, the identification

of alternatives, selection of other options, specific activities, decision-making, decision

making and intelligent prediction.

On the other hand, Internet of Things ( IoT) is the core element of the industry 4.0 revolution

and involves a worldwide infrastructure for storage, application, sensing, advanced services

and networking technology to capture and process data/information. With the combination of

high-speed, resilient, low latency networking, the Al and IoT technologies, the complementary real-world and digital know-how of industry 4.0 is transformable to a truly smart AV. It was shown that 90 percent of auto accidents are caused by human error, and ten times the average driver is the safest. Automated vehicle safety is critical, and users need an appropriate level of risk 1000 times smaller. One of the unbelievable advantages of AVs is:

(1) an improvement in the safety of the vehicle, (2) a reduction in injuries, and (3) a reduction

in fuel usage and, (4) a lack of driver time and business opportunities. AVs can, however, use

large-scale sensors and system data/information.

The difficulty of AV data/information (1 GB per second of processing) used during

Advanced Driver Assistance Systems ( ADAS), and entertainment is growing. Hardware and

software specifications using sensors, actuators, and software must also be designed to cope

with functions identical to the superhuman brain as aimed at by Al. AV sensors and

equipment produces information such as time, data, motion detection, navigation, fuel

consumption, voice recognition, acceleration vehicle speed, deceleration, cumulative miles;

voice search; eye and conductor monitoring engines; image recognition, feel analytics, voice

recognition, gesture and virtual assistance. Similarly, data are produced by AV sensors and

devices. The overall estimates for 100,000 vehicles are also 100 terabytes per year.

BACKGROUND OF THE INVENTION

These data will further increase as connected vehicles (CVs) are becoming increasingly

available. As the AV goes forward, industrial producers and distributors have new

opportunities so that businesses can use Al to boost their client's value. In processing this

data/information through Al, Machine Learning ( ML) algorithms are the most robust

approach. The ML algorithms help to establish behaviour patterns for specific drivers profiles

and give vehicle owners with their corresponding application precisely what they need both

in the vehicle and through mobile phones. You do so by recalling your actions and evaluating

your driving background and the situation on the lane. Via various IoT networks including

Local Area Network ( LAN), Wide Area Network ( WAN), Wireless Sensor Network

(WSN), and Personal Areas Network (PAN), Al can manage AV broad information, but

some additional conditions such as traffic, feelings talk and experience will have to be

collected via various IoT networks. This considerable data/information needs to have certain

substances that allow them to collect data and exchange it, such as embedded electronic

devices, sensors, automobiles, constructions, software and network connectivity. These IoT

enabled AVs use a range of integrated technologies, such as enhanced protection, fuel

consumption reductions and vehicle safety, to provide many successful assistance in real

time. The 4.0 industry and IoT can be turned into a significant stimulus by reducing system

failure, enhancing quality control, increasing efficiency, and simultaneously decreasing costs.

IoT technology's potential and projections are incredible.

Morgan Stanley Research's study was claiming that, with many superior technologies, main

features and services, at least nine industrial companies would benefit from AVs.

(1) Original Equipment Manufacturers (OEM), (2) Auto dealers, (3) Autonomous trucks, (4)

Chemical engineering, (5) Electric utilities, (6) Semiconductor, (7) IT hardware-software, (8)

Telecom and communications, and (9) Beverage and restaurant sectors.

Artificial Intelligence: Benefits and Challenges

Industrial revolution trend, including technology, automation and data exchange. Present

businesses have new competitive and demand problems and have to undergo a drastic

transition to the evolution of Industry 4.0. Artificial Intelligence ( Al) offers better decision

making and decision-making dynamics for industry 4.0 and results in increased efficiency of

the enterprise, decreases the fault of the machine, improves quality management, increases

the productivity and lowers costs.

While Al is likely to modify the world today, it does have its limits. All's key difficulties are

learning from experience, and expertise can not be translated into learning. Furthermore, the

results indicate any inaccuracies in the data and are very difficult.

Artificial Intelligence: History and Approaches

Al is focused on broad data combinations. With iterative processing, it processes data very

rapidly through intelligent algorithms that allow the programme to learn from data

functionality or patterns. Since the vision error recently declined significantly (< 5 percent)

relative to human vision error, Al has become more prevalent.

Autonomous Vehicle

An independent vehicle (AV), instead of being driven by humans, is a vehicle that can guide

itself. The AV has grown into a form of driverless vehicle that is the art of driving with

computers for the future. AVs is targeted due to: (1) improved safety of cars, (2) reduced

injuries, (3) fuel consumption, (4) release of driver time and business opportunities, (5) new

market opportunity, and (6) decreased emissions and dust particulates. Around 10 million

AVs are expected to be on the road by the year 2020, and AVs are expected to generate a

cumulative revenue of 7 trillion USD by 2050.

Automated Vehicle: Levels and History

Advanced Driver Assistance Systems (ADASs) vehicles have six tiers for automated

vehicles. The travel stage of automation to fully self-sufficient automobiles is: Null stage

no automation and the people carry out all the complex driving tasks such as speeding or

slowing down, driving, braking etc. Level 1: driver help via an acceleration/deceleration

system, or steering with driver knowledge.

Stage two - partial automation of cars, where both acceleration/determination and steering

functions are mixed. Stage 3: Automation of driving mode by the condition that the

automated driving device executes with precise precision when the driver answers requests.

Level 4-high automation enables all driving tasks to be carried out under certain

circumstances, even if a human driver does not respond to a request. The vehicle performs all

the driving jobs/functions under all situations, level five - full automation.

Level 4 and 5 allow AVs to perform all driving capabilities together with several systems and

sensors to control a driverless car. The six automated vehicle levels (detailed in the section

"AV Objective Sensors," ADAS Technologies, Sensors and Controls) were developed.

Artificial Intelligence in Autonomous Vehicle

Step 1: Data Collection

AVs are designed to generate a large amount of data from their vehicles and their

surroundings using multi-sensors and equipment, including radars, cameras and

communication. These AVs include lane, road facilities, all other vehicles, parking, traffic

information, and environmental information similar to that of the human driver, and any other

item on or close to the road. These data are then forwarded as modified information to be

processed by AV. This is the first AV contact with particular vehicle circumstances and

conditions.

Step 2: Path Planning

AV device massive data will be saved and added in each route into a database called Big

Data from past driving experiences. An Al agent also uses a big data for making meaningful

strategic decisions. The track planning control strategy of the AVs helps self-driving vehicles

to determine the shortest, most comfortable and most economically profitable routes between

points A and B, using previous driving experience that enables the Al agent to make far more

precise future decisions. All the static and manoeuvrable obstacles that must be found and

bypassed by a vehicle make seeking routes difficult. Path planning strategy means finding a geometric path from an initial setup to an initial configuration, so that and set up and state on the track are feasible (if time is taken into account). Road planning management strategy involves Maneuvre Planning which is designed to decide on a vehicle at the best high level when contemplating how a vehicle is to go from one viable state to the next in real-time and follow the film limits of a vehicle based on its vehicle dynamism and as limited. The route planning strategy includes manoeuvre planning. In the driving and driving situation, the AV knows precisely what to do.

Step 3: Act

The AV can detect road objects, navigate through the highways, parking lots, obstacles,

entertainment, traffic lights, bikes, pedestrians, work areas, weather conditions and other

vehicles without the interposition of the human driver and goes to the destination safely based

on the decisions taken by the Al agent.

The AVs also come with Al-based control systems and functional systems such as steering

control, pedal speed, voice and speech recognition, brake pedal monitoring, visual

monitoring, protection system, gesture controls, fuel consumption and other driver

support/monitoring systems. This AV process loop will take place repetitively, including data

collection, route planning and execution. The more data loops are generated, the smarter the

Al agent is, and the more accurate the decisions are made, especially in complex driving

situations.

Autonomous Vehicle Challenges

Road Conditions: Road conditions are highly evolving and vary from point to point. Smooth

and marked large roads are available in some places. However, several other places have

significantly degraded road conditions - no lane marking. Lanes are not defined; potholes,

mountain and tunnel roads do not have very simple and identical external directions.

Weather Conditions: Another spoilsport is the weather. Sunny, clear weather or precipitous

and precipitous weather can occur. In all kinds of weather conditions, AVs should operate.

No space for malfunction or downtime is available at all.

Traffic Conditions: AVs should be driven on the road under all kinds of conditions of

transport. They would have to drive on the road with other AVs, and there would be several

people at the same time.

There are a lot of feelings wherever people are involved. Traffic can be extraordinarily

moderate and autonomous. However, it is also the case that people violate traffic laws.

Unforeseen circumstances can lead to an object. Even the motion of few centimetres per

minute is necessary for dense traffic. One can not wait indefinitely for traffic to clear and

provide preconditions for traffic to start moving. If more of these cars wait on the road to

clean up the traffic, this may eventually lead to a traffic impasse.

Accident Liability: Liability for injuries is the most critical part of AVs. Who is responsible

for AV accidents? In the case of AVs, the programme will be the main driving force of the

car, and all the crucial decisions will be made. Although the initial designs have an individual

positioned physically behind the steering wheel, Google's latest techniques do not include a

dashboard and a wheel. How is the person in the vehicle expected to inspect the vehicle for

an unsightly incident in these designs, where the car has no power, such as a steering wheel, a

brake pedal, or an accelerator pedal? Also, the creation of AVs primarily means that the

occupants are in a comfortable role and can not pay close attention to traffic conditions.

When their attention is needed, it might be too late to avoid the situation when they need to

act.

Radar Interference: For navigation, AVs use lasers and radar. The lasers are attached to the

rooftop while the sensors are connected to the vehicle's body.

Integrating Artificial Intelligence with Edge Computingfor Autonomous Vehicles

To have an Al model for AVs that use edge computing, we have to change the conventional

cloud model where all data storage and analytics take place in the cloud. This traditional

model is presented in the cloud to incorporate the control mechanisms and database system.

To develop this model, the method and preparation needs to be split into two modules that are

done in combination with the edge and the cloud. Al-based AV using edge computing, where

data from the AV are moved for pre-processing and decision-making to the edge node. The

data from the IoT sensors are processed locally on the edge while data from the edge nodes

are collected and transmitted to the cloud for global offline and time-conscious decisions.

Thus, time-sensitive decisions like identification of obstacles or avoidance of crashes are

made at the edge node in a much shorter time. The data on route, traffic and driving trends in

the cloud was analysed to enhance road safety and driving experience. The Al models in the

edge node can be dynamic and modified in compliance with the policies, applicable

regulations and customer requirements. The amount of data transferred in the cloud is lower

than the amount produced by IoT sensors in AVs as data is pre-processed, filtered and

cleaned in the edge node before cloud download. The bandwidth and costs can be

considerably decreased.

Energy and Emissions Implications of Autonomous Vehicles

In the United States, the use of light-duty passenger cars leads to almost 20% of GHG

national emissions (EPA, 2013a). It is also a large contributor to standard air pollution, such

as smog and ground-level ozone, and accounts for around 60% of petroleum consumption

(Davis, Diegel and Boundy, 2013). As with protection, congestion and land use, the transition

to AVs could, at least on a long-term basis, have a significant effect on energy usage, GHG

emissions and traditional air pollution impacts in the transport sectors. It will be focused on

three factors to assess whether AVs boost or deteriorate energy usage and environmental outcomes: the fuel efficiency in AVs and the carbon intensity and life cycle emissions profile in AV fuel will result in the overall shift in VMT resulting from the use of AVs. About these three variables, we address the possible scope and direction of change AVs might have.

Politicians and other actors will use this knowledge to understand how short-term policies

can influence the potential energy and environmental outcomes for AVs.

Parking is the source for many cities of significant steady urban profits. AVs will kill this

source of income by making nearby parking unnecessary. Although tax-revenue-generating

uses can substitute parking, a shift to AVs can disrupt municipal finances substantially.

Others noted possible problems with social equity posed by AV innovations, arguing that an

emphasis on AVs would distract us from the transit system (Arieff, 2013). The AVs could

only reinforce our individualist auto-centred society through the hungry public transport of

drivers, rather than improving the transport that can benefit all people. One of the major

public transportation attractions is that a smartphone can be read or used. In private vehicles,

less people can use public transit while these practices carried out. In turn, this could lead to

lower fare income and public transit to decrease or raise rates, which could build a vicious

spiral in reducing motorcycling. AVs are also expected to at least initially be considerably

more costly than traditional cars - exacerbating the risk of crash between the rich and the

poor. However, these findings are not predetermined, and several policy instruments are

available to address. Jobs' going to be lost, too.

WORKING PRINCIPLE

The software will recognise the signs and store them with Google Maps and GPS in a

common database. The GPS software part registers travel signs and directions. Any vehicle

moving in traffic with the software records the new signs that have been identified and updates the trust and recognition level for other drivers. A different software variable can recognise the dividing lines between the paths. It uses three cameras to precisely measure the location of the car on the lane, where the lane sides are, or to give a new route for the next few seconds, even without traffic signs. Another part has been used to identify other vehicle fingerprints from webcam images with Artificial Intelligence. Parallel to and propagation of the algorithms is implemented. To ensure data processes and control on three separate computers with multiple core systems, I built a semaphore management software framework.

Besides, this innovation includes a homemade LIDAR - 3D radar and OpenGL software to

create a 3D world for the car to navigate. The car would choose to prevent hurdles when

using it. The 3D radar helps to improve the trust of decisions in the software framework. I

also used another training technique in the network, drive car with EEG sensor about 100 km,

and took over 1 000 human brain samples when he took some action such as Acceleration,

Break, left or right turn. All the situation was mapped with brain samples.

The control is obtained by tapping into the Controller Area Network (CAN) bus of the

vehicle. We reflect the steering command 1/ r, where r is the turning radius in metres, to

ensure that our device is separate from car geometry. We use 1 / r instead of r to avoid the

singularity of straight driving (the turning radius is endless). 1 / r smooth transitions from left

to right by zero turns (negative values). The training data includes individual video clips,

coupled with the respective steering control (1 / r). Training with human driver data alone is

not enough; the network needs to learn how to recover from errors and how to get the car out

of the traffic slowly. This improves training data with other pictures showing the vehicle in

various changes from the middle of the track and turns from the direction of the lane. Photos

can be taken from left and right cameras for two particular off-centre shifts.

The viewpoint simulates further changes between the camera utilising the image

transformation from a nearest camera. Precise shift of the perspective involves awareness of

the 3D scene that we have no, so we approach the change by assuming every point below the

horizon is flat and every spot above the horizon is endlessly removed. It works well for flat

land, but it provides distortions for objects which cling over the ground like vehicles, poles,

trees and buildings for a complete rendering. These distortions luckily do not pose a big

problem for network preparation. In two seconds, the steering label for the images is adapted

quickly to one which correctly guides the car back to the desired position and orientation.

Editorial Note 2020103161 There is 1 page of Claims only.

We can claim:

(1) The use of data to automate learning,

(2) Developing knowledge skills for current goods,

(3) The adaptation to programmable by data of smart learning algorithms,

(4) A logical data analysis,

(5) The improvement of data accuracy.

Editorial Note 2020103161 There is 1 page of Drawings only.

Fig: Autonomous Vehicle by using multi sensor

Fig: AI model for autonomous vehicle including data collection, planning and act