Artificial Intelligence (AI) is is a key technology in the development of autonomous or self-driving vehicles. Not only is AI capable of collecting and analysing data through sensors and cameras but it is also capable of adapting to situations and learning through machine learning.
When AI is paired with a vehicle, it allows the vehicle to become autonomous, enabling it to make its own decisions without human intervention through the continuous collection and analysing of data.
Levels of automated driving systems
The capabilities of an autonomous vehicle are categorised into 6 levels (Level 0 being the lowest). This categorisation, developed by the Society of Automotive Engineers is explained through the graphic below:
Graphic source: www.sae.org
This categorisation is important as it serves as a general guideline for how technologically advanced a vehicle is.
In order to create a Level 5 autonomous vehicle, AI developers and vehicle manufacturers will have to work together in these 4 main areas – navigation system, path planning, environment perception and car control.
The main principle in developing an autonomous vehicle is to build a vehicle that is capable of moving to from point A to point B without human intervention and that can only be achieved if the vehicle has a sense of direction.
For a car to be considered self-driving, it must first be able to automatically locate its position and perform the path planning to destination. For this objective, a global positioning system (GPS) is equipped to receive the location information from the satellite.
Other than GPS, the AI fitted in autonomous vehicles also utilises sensors such as accelerometers and gyroscopes to determine the angle and the speed of travel, enabling the vehicle to make navigation decisions and path planning accurate, faster and smarter.
Picture source: www.researchgate.net
Once the vehicle obtains data of its location, it can then utilise path planning algorithms such as Dijkstra’s algorithm* and Bell-Ford algorithm* to create a path to its destination.
- Dijkstra’s algorithm – an algorithm for finding the shortest paths between nodes in a graph, which may represent, for example, road networks.
- Bell-Ford algorithm – is an algorithm that computes shortest paths from a single source vertex to all of the other vertices in a weighted digraph. It is slower than Dijkstra’s algorithm but more versatile as it is capable of handling graphs in which some of the edge weights are negative numbers.
For the AI to determine an appropriate route to its destination, it will need various road data such as traffic intensity, building information and information on road works. To achieve this, an electronic map – a digital memory of the map embedded into the AI, is used to store these data.
Generally, an electronic map contains three layers of information to allow accurate decision-making. The three layers are:
- Active layer
- Contains physical data of the map such as road shape, elevation and angles, lane width, trees, buildings and so on.
- Dynamic layer
- Stores real-time data that is obtained from sensors, radars and cameras surrounding the autonomous vehicle’s environment (this includes other vehicles, buildings and road sensors).
- Analysis layer
- This layer enables the AI to constantly analyse all the data obtained from the previous two layers to create and form the best route to its destination.
With the usage of proper path planning algorithms and analysis of real-time data, the vehicle can automatically decide the easiest, shortest and fastest route to a particular destination.
Picture source: www.viavisolutions.com
To provide necessary information for an autonomous vehicle to make decisions, the vehicle is required to perceive its surrounding environment on its own. This is achieved by equipping vehicles with various sensors, radars and cameras that can capture vital data to ease and increase the accuracy of the autonomous vehicle’s decision-making ability.
In order to be fully autonomous, the vehicles will also capture data from its surroundings through sensors, radars and cameras that are strategically placed in various locations such as traffic lights, buildings and even public transportation systems. This will form a network of continuous data where every connected vehicle, transportation system and infrastructure can obtain and analyse data from – eventually developing into a Smart City.
Generally, sensors are used to measure reaction time. It does this by transmitting a laser to a particular object and calculating the time taken for the laser to reflect. Radars, are used to measure distance and cameras are equipped to provide autonomous vehicles a visual perception (3D data) of its surroundings.
These apparatus are already widely used by current car manufacturers, especially in the application of autonomous emergency braking system, adaptive cruise control and automatic headlights.
However, the next step in this area is to enable the vehicles to capture and share the data with each other in real-time so that the data is accessible by all autonomous vehicles, making it easier for the AI to learn and adapt to new and complex situations.
A vehicle’s control system is the core component of a vehicle as it controls the various functions of the vehicle, which includes the anti-lock braking system, electronic stability control, cruise control system and so on.
This “vehicle control system” consists of the electronic control unit (ECU) and communication bus. The ECU implements the control algorithm whereas the communication bus realizes the communication function between ECU and mechanical parts.
In order for a vehicle to be fully autonomous, the AI must be able to know when to use these various features embedded within the vehicle. Hence, the data contained within the three layers of the electronic map are fed to the AI.
Then, the AI performs calculations based on the data received to transform raw data into meaningful data, which is then passed to the vehicle control system. The vehicle control system then uses these meaningful data to control the vehicle’s direction and speed, including other features of the vehicle such as lights, signalling and honking.
The continuous development of technologies will only make it easier to realise a Level 5 autonomous vehicle. However, it is more important for vehicle manufacturers to continuously work closely with AI developers in order to supplement each other’s growth in order to progress towards future mobility.