(source: techstory.in )
Automation is going on in almost all the sectors of the world now-a-days. In every part of our lives, the usage of robots or automated systems is increasing. For example, if we talk about Space Technology, a planetary rover might decide to take samples or images on its own. Another one that is widely known are home automation like Google home, smart electronic devices, automatic vacuum cleaners, etc. So, it's very evident that the saturation of automation in all the industries is increasing rapidly. Because the demand is so high and the whole world is currently turning towards automation, the car industry is also taking part in the shift.
Autonomous or self-driving cars are very popular right now in the world. The fact that automated cars can reduce a lot of errors done by humans and will be a lot beneficial in other aspects too, makes it so important. A study shows that over 90 percent of road accidents are caused because of human error. There is much information suggesting that autonomous cars are one of the most important technologies which should be developed.
What is Automation in Cars?
Automation in cars means making the cars compatible to sense the surroundings and make decisions on the basis of that. The main concept revolves around the idea that the car doesn't need a human as a driver or even as a passenger. An automated car ought to do most of the things that a car does, on its own. Now, Automated cars are different from "self-driving" cars, combining sensors and software to control, navigate, and drive the vehicle.
So according to the Society of Automotive Engineers(SAE), currently there are 6 levels of driving automation in cars. They start from 0(fully manual car) to level 5(fully autonomous).
Details about the autonomy of cars:
- Level 0: Fully manually controlled by humans, these are the cars that have been mostly used throughout time.
- Level 1: Some of the systems of the car may be controlled by the car itself, like cruise control, automatic braking systems.
- Level 2: The car's automated systems are continuously increasing with the levelling up in automation. Here, the car's control can be fully taken over by the automated system. But the driver must be before the steering to intervene if the system fails to recognize a potential hazard.
- Level 3: The automation system takes over the full control of the car, drivers do not need to care for the driving tasks. Although the driver can still intervene if ever needed.
- Level 4: Driver fully leaves the driving tasks to the automation system. The car becomes fully-autonomous, although not in some scenarios. This level of automated driving is limited to very small cars that have been geofenced and have controlled environmental conditions.
- Level 5: Now the car is totally self-driven in any situation and any environmental conditions.
( Source: synopsis.com )
What's the difference between Autonomous, Automated and Self-driving cars?
Mostly when the aspects of these kinds of vehicles are taken up, the word automated is used. The main reason lying behind it is that autonomy goes beyond steering control or just sensing the environment. A fully autonomous car should be able to make decisions on its own, the car should be self-aware of that. On the other hand, a fully automated car takes the order and drives on its own.
Besides that, the terms autonomous and self-driving are also different. In self-driving, the vehicle can go on its own,maybe to almost all the places, but human passengers need to be there always. Self-driven cars can go up to Level 3 of automation but not higher than that. They are also limited to the places they can go and need geo fencing to know which place is reachable. Meanwhile, the fully autonomous vehicle can go up to level 5 of automation which allows it to go anywhere and can work independently.
How do autonomous cars actually work?
(Source: landmarkdivided.com )
The technological advancements in recent years have finally accumulated and made it possible to make autonomous a reality today. To be specific there are three main technologies that are making the cars autonomous:
1. IoT Sensors:
The fundamental concept of autonomous cars is that they will be sensing the surroundings and will be acting accordingly. So, when we talk about sensing the surroundings we will obviously need sensors to do that for us. There are different variants of sensors available today that make autonomous cars a reality for blind-spot monitoring, forward collision warning, radar, camera, LiDAR, and ultrasonic. They cumulatively work to make the magic happen which is giving the senses to an electromechanical object.
For Blind spot monitoring a car uses a set of sensors, which are mounted on the side mirrors or the rear bumper. They detect the cars in the adjacent lanes and give signals to the warning system to give warning to the autonomous system. In a lot of cars, cameras are also used to do this job.
(source: motortrend.com )
LiDAR (light detection and ranging) sensors can do both short and long-range sensing. It is used for 3-D mapping of the environment and the surroundings to make accurate detection of the surroundings. It is highly used to avoid collision and range-sensing.
So when the performance of all the sensors is combined, a leveraging strength is unleashed. The most valuable outcome which comes after the fusion is Safety, which is the top priority. To make that happen all the sensors complement each other and assist each other to get the best possible outcome. The ultimate motive is to use the strengths of the various vehicle sensors to compensate for the weaknesses of others and thus ultimately enable safe autonomous driving with sensor fusion.
(Source: blickfield.com)
2. IoT Connectivity:
The sensors aim to collect data, but it is of no use if it is not collected and managed safely. Now here cloud computing is used to analyze the environment around the cars including others cars, traffic etc. They become very handy while helping the car to monitor the surroundings and taking informed decisions. Autonomous cars also are needed to be connected to the web albeit edge computing hardware can solve small computing tasks locally.
When it comes to connectivity, the internet has become one of the most important commodities now-a-days. So mostly around the world, 4G services are used for connectivity. It is quite good but not the best that technology can provide. So with the advancement of network connections, we will be able to implement 5G, which will heavily influence the autonomous cars' world.
5G promises to be close to 1,000 times faster than 4G LTE at peak throughput, which will make connection woes such as high latency and long response times a thing of the past. If it is equipped with autonomous cars, 5G networks will allow for seamless communication between two or more cars, but it doesn’t stop there. Gradually the world is being dominated by IoT devices, where everything, be it a motorized vehicle or a traffic light, will be connected to a high-speed network of some sort, enabling all sorts of new and exciting functionality to be a possibility. Gradually the world is being dominated by IoT devices, where everything, be it a motorized vehicle or a traffic light, will be connected to a high-speed network of some sort. This connectivity enables a lot of new and exciting functionality to be a possibility.
(Source: landmarkdivided.com)
3. Software Algorithms:
Now the data is collected, shared but it is needed to analyze them properly and then find the best course of action. So this task is actually performed by control algorithms and software. The complexity of this part is higher than any other system which is used to make the autonomous car. The algorithms need to decide flawlessly, otherwise "A flaw" can result in a fatal accident like Uber's Self-driving accident.
Some of the most used software algorithms:
Regression Algorithms
Regression algorithms leverage the repeatability of the environment by which it creates a statistical model of the relation between a picture and therefore the position of a given object in their image. Now, the statistical model is often learned offline and provides fast online detection made possible by image sampling. Besides that, it is often extended to other objects without requiring extensive human modelling. As an output to the web stage, the algorithm returns an object position and confidence in the presence of the thing. All of those algorithms also can be used for long learning, short prediction. The sort of regression algorithms which will be used for self-driving cars are Bayesian regression, neural network regression and decision forest regression, among others.
Pattern Recognition Algorithms (Classification)
Using ADAS, the pictures obtained through sensors possess all the kinds of environmental data obtained; filtering of the pictures is required to acknowledge instances of an object category by ruling out the irrelevant data points. Pattern recognition algorithms are good at excluding these unusual data points which are irrelevant to the model. Recognition of patterns during a data set is a crucial step before classifying the objects. These sorts of algorithms also can be defined as data reduction algorithms.
These algorithms help in reducing the info set by detecting object edges and fitting line segments (polylines) and circular arcs to the sides. Line segments are aligned to edges up to a corner, then a replacement line segment is started. Circular arcs are fit sequences of line segments that approximate an arc.
Clustering
Sometimes the pictures obtained by the system aren't clear and it's difficult to detect and locate objects. It’s also possible that the classification algorithms may miss the thing and fail to classify and report it to the system. The rationale might be low-resolution images, only a few data points or discontinuous data. This sort of algorithm is sweet at discovering structure from data points. Like regression, it describes the category of problem and therefore the class of methods. Clustering methods are typically organized by modelling approaches like centroid-based and hierarchical. All methods are concerned with using the inherent structures within the data to best organize the info into groups of maximum commonality. The foremost commonly used sort of algorithm is K-means, Multi-class Neural Network.
Decision Matrix Algorithms
This type of algorithm is sweet at systematically identifying, analysing, and rating the performance of relationships between sets of values and knowledge. These algorithms are mainly used for deciding. Whether a car must take a left turn or it must hit the brakes depends on the extent of confidence the algorithms carry in the classification, recognition and prediction of subsequent movement of objects. These algorithms are models composed of multiple decision models independently trained and whose predictions are combined in to form the general prediction. While reducing the likelihood of errors in deciding. The foremost commonly used algorithms are gradient boosting (GDM) and AdaBoosting.
(Source: visteon.com)
Conclusion:
There are several challenges in making autonomous cars a reality like expensive LiDAR, poor weather conditions, understanding traffic law, etc. One of the most important potential differences is that it will dramatically lower CO2 emissions. In a recent study, experts identified three trends that, if adopted concurrently, would unleash the complete potential of autonomous cars: vehicle automation, vehicle electrification, and ridesharing. By 2050, these “three revolutions in urban transportation” could:
- Reduce traffic congestion by 30%
- Cutting down of transportation costs by 40%
- Enhance walkability and livability.
- Lower down urban carbon dioxide emissions by 80% globally.
This technology might influence all of our lives whether we own a self-driving car or not. Our society can definitely enjoy autonomous vehicles, it will help the elderly people and the people with physical impairment. Soon, everyone is going to be ready to own a self-driving car. we hope that there’ll be fewer crazy drivers on our roads soon.