Sunday 11 April 2021

Machine with Intelligence

 In the Name of Allah, the Most Gracious, the Most Merciful 

 


 

How to make machine with intelligence?

Human has constantly been trying to recognize how our mind works, and make machines to act like our brain. It way we had been trying to make machines that could think. It was no longer easy like making machines which could do simple such things as calculating or lifting heavy stuff. It become no longer sufficient to meet human’s will. What we need are a device that may classify pics, recognize bizarre cells among healthy cells, or play chess or go together with us! Of course, it isn't always clean. It is not making machines anymore, it's far greater like mimicking how human thinks.

 


 

Got some idea from nature!

understanding human mind with math did no longer assist it. assume, is it feasible to construct living creature’s biology with only logical math of 0 and 1? however, someday, one precise concept got here out. It was simply cloning human mind like neurons not designing digital circuits. additionally, there was a big improvement of computer hardware like GPU. GPU has extra simple employees than CPU so it became perfect to do simple calculations what we need in neural networks, computer approach works like neurons!


 

 

Like Neural Networks

As Korean, the biggest event in last decade about AI was ofc, the great Google Deepmind Challenge match, the Go match held in 2016 between AlphaGo and Sedol Lee for 5 days. The match between the best Go artificial intelligence program and the best human talent in Go attracted attention, and the final result was AlphaGo’s victory over Sedol Lee with four wins and one loss. This incident imprinted the concept of artificial intelligence on Koreans. Artificial intelligence can be easily found in our lives. Artificial intelligence is also used to determine whether Netflix’s recommendation system, Google’s automatic completion function, Apple Watch users are exercising or not, and whether smartphones will catch Wi-Fi or use a communication network. And the basic of all artificial intelligence is neural networks.

 


 

Traditional Computers

In truth, computers are a sort of calculator. computer systems can do math operations in no time. So we use computers while we need the functions of a calculator, inclusive of getting the sum of numbers, calculating taxes, or processing numerical records. watching videos and being attentive to tune through a computer or cellphone additionally feels like a math operation to a laptop. all of the records is made of numbers, so what computers do is handiest rushing up the 4-pronged operations we’ve discovered. but, dashing up the four-pronged operation does not imply that computer systems have intelligence. 

 


 


Having Intelligence?

Then how can we say what having intelligence is? let us think about the photograph below.




can you apprehend what is it within the photograph? clearly, you don’t want to assume. it's far a bird. We, individual, can say it with self assurance and it's far even tough to get it wrong.

What about this photo? what's it about? how many birds on this photograph? There are sincerely 3 birds within the picture.




Like this, we humans can process a large quantity of information that an image has and become aware of the objects that are not there. however, it is not easy for computer systems. surely, it's far very tough.


TasksComputerHuman
Calculating a lot of numbers easy hard
Understanding imageshardeasy    

 

The point is, human desires computer systems do works rather than us, no longer simplest just calculating numbers, however also know-how photographs. due to the fact computer can work quicker and longer than human.



And the element that makes computer can do those hard tasks is artificial intelligence. due to the fact computer is includes digital circuits, it is locating and making new algorithms to solve tough problems for machine. Make machines do things like human.

Saturday 27 March 2021

Autonomous Driving - Trends 2021

 In the Name of Allah, the most Gracious, the most Merciful

Overseas autonomous driving technology status

The development of autonomous driving technology is focusing on commercialization and product production rather than technology R&D, and the pace of development is accelerating. Global market company Navigant Research published a report on the competitiveness of 19 major global companies developing autonomous driving technologies and platforms. Competitiveness is compared based on vision, market entry strategy, partnership, production strategy, technology, sales/marketing/distribution, product quality and reliability, and classified into leading, competition, challenge, and subgroups. GM, Google Waymo, Daimler, and Ford formed the leading groups, PSA, Toyota, Volvo, Baidu, Navya, and Hyundai were competing groups, while Apple, Uber, Tesla, and Honda were evaluated as challenging groups. 



Google Waymo

Google's subsidiary Waymo is trying to solve problems such as lane keeping, which requires constant observation of the driver to create a system that can fully autonomously drive without a driver. Currently, it has completed an autonomous driving system of level 4 of the American Society of Automotive Engineers (SAE), which requires the driver's observation only under certain conditions. Waymo built a system that recognizes and reacts to obstacles and events around the car using lidar, cameras, radar, and other ancillary sensors. 


Waymo's autonomous driving software can be divided into three main stages: observation stage, behavior prediction stage, and planning stage. In the observation stage, objects around the road are recognized, what they are, and their speed, direction, and acceleration are measured. In the behavior prediction step, the recognized objects predict how they will behave around the road. Waymo can make predictions based on the experience of driving millions of miles. In the planning stage, based on the information obtained in the previous two stages, the car will plan which route to take. In their experience, they think defensive and timid driving is the safest. 


The Waymo system can create a map and locate a car through real-time information with a deep understanding of the physical environment such as road type, distance, and dimension. In addition, it is equipped with a system that collects and interprets data obtained through trial driving. A system was also developed to protect against hacking caused by internet connection. 



Tesla Autopilot

Tesla's Autopilot is Tesla's autonomous driving system and hardware that provides lane maintenance, car control, autonomous parking, and lane change with driver confirmation. Tesla aims to provide fully autonomous cars. 


Advanced autopilots adapt to traffic conditions, keep their lanes, and change lanes on their own without driver intervention. If you move from another arterial road to another arterial road and get close to your destination, you can get out of the arterial road by yourself and park your car automatically.



Current status of open self-driving platforms abroad

Baidu Apollo

Baidu unveiled the autonomous driving platform'Apollo 3.0' in July 2018. 'Apollo 3.0' is the level 4 level of the American Society of Automotive Engineers (SAE). Level 4 is a level that does not require driver intervention, but the driver must be on board in case of an emergency. 'Apollo 3.0' is equipped with a program that allows developers and partners to develop autonomous vehicles within three months. The program consists of autonomous parking, unmanned self-driving delivery, and unmanned self-driving shuttle service. 


In the case of valet parking, all vehicles can use the autonomous parking service at a cost of about $1509 (about 1.68 million won) by adding only a camera and an ultrasonic radar. You can also use autopilot kits, security systems, and operational scheduling solutions. In addition, interesting technologies such as Baidu's voice recognition software, facial recognition technology, fatigue symptom monitoring technology, and customized services are also installed. 


Based on this technology, Baidu plans to pilot ten 14-seater self-driving minibuses, Apollon, in Japan and China's Beijing, Shenzhen, Wuhan, and other cities this year in cooperation with Softbank's subsidiary SB Drive. Apollo is equipped with Baidu's autonomous driving platform Apollo. 



Autoware Foundation Autoware

Autoware was jointly developed by Nagoya University and Nagasaki University in Japan, and was released in August 2015, and is still being actively developed. Autoware is an application software developed based on the robot software platform ROS (Robot Operation System). When a 3D point map created using LiDAR is given in advance, Autoware compares the 3D point cloud data obtained from the LiDAR while driving to determine the location of the vehicle in real time. Once the location is determined, the driving route to the destination can be planned and autonomous driving can be performed by controlling the steering wheel, accelerator pedal, brake pedal, and gear stick. 

 
The figure schematically shows the operation process of Autoware. First, in the recognition step, the position of the host vehicle is identified on a three-dimensional point map using a lidar, and an obstacle is recognized with a camera. Second, in the determination step, a safe driving route is planned based on the identified information and the input final destination information, and a driving lane, an entrance lane at an intersection, etc. are determined. Third, in the driving step, the target steering angle and target speed of the vehicle are determined in real time according to the given driving path and transmitted to the vehicle control computer. Finally, the vehicle control computer manipulates the wheel, accelerator pedal, and brake pedal to achieve the delivered goal.

Overseas self-driving technology verification process status

Mcity Test Facility University of Michigan, USA

The Mcity Test Facility is a test site designed to test and simulate autonomous vehicles and technologies in urban and suburban environments. Mcity worked with the Michigan Department of Transportation to design a unique test facility that simulates the complex vehicles encountered in urban and suburban environments. Mcity is located on the University of Michigan's North campus site with approximately 16 acres of road and transportation infrastructure. It has about 5 lanes with obstacles such as intersections, traffic signs, signals, sidewalks, buildings, street lights, and construction barriers. 
 

Centre of Excellence for Testing & Research of AVs (CETRAN) 싱가폴

CETRAN (Centre of Excellence for Testing & Research of AVs), led by Singapore's Nanyang Technological Univercity, does not directly develop autonomous driving technology, but how the autonomous driving system should work, how to test it, and autonomous driving. Study how international standards for automobiles should be created. CETRAN considers the safety of automobile passengers first, and their goal is to make society trust autonomous vehicles such as public transportation. 
 

 


American Center for Mobility (ACM) Michigan, USA

The American Center for Mobility (ACM) offers a variety of autonomous driving test environments over 500 acres. ACM is a global center established for testing, verification, training, and product standard development for autonomous vehicles. It can be used by all businesses, governments, universities, etc