Autonomous vehicles
2. DNNs and algorithms analyse this data to form a live, digital representation of the world. The algorithms recognize, calculate, and predict a plethora of variables and scenarios to understand what is happening around the vehicle. Sensor fusion combines the individual readings of different sensors to identify singular objects. This can be thought of as ‘refining’ the data. Machine learning aids this phase and helps evolve the intelligence of the algorithms to improve accuracy in calculations and predictions. 3. Decision matrix algorithms function holistically, producing a physical decision by weighting these different pieces of data and values. 4. Once the most successful/safe outcome has been decided, the computer instructs mechanisms within the vehicle to operate and carry out the task (using actuators).
Feasibility
It is clear that the capabilities of autonomous vehicles lie within the limits of the technology it relies upon. In a self-driving vehicle, there are specific computer chips for machine learning functions and the ‘Deep Learning’ aspects of AI. In addition, each sensor and piece of hardware requir es multiple chips to function, as well as the algorithms which use their own computer chips in order to run. However, the self-driving car still needs to hold the required number of chips to function as a standard car, which only adds to the number of computer chips used (Eliot 2021). It is thanks to the miniaturization of electronics that these feats can be achieved, specifically, integrated circuits (ICs). ICs are composed of multiple transistors and other components on one silicon chip. The continuous sc aling down of these components on products has allowed us to achieve ‘ultra -large-scale integration’ , where we can now incorporate 10 billion transistors in just one silicon chip. Increasing the number of transistors increases the power of the system in terms of processing data, and the amount of storage space for said data. Moore’s Law states the number of transistors incorporated on one chip will double every 2 years (Granger 2017), and even though a ‘saturation point’ seems to be emerging in the number of transistors being integrated on a silicon chip, the development of nanotechnology could revolutionize the implementation of transistors and accelerate this technology to new heights. To put things in perspective, in 1984, 1 million transistors were implemented on a singular silicon chip. Today, 50 billion transistors can be integrated in a chip, an increase of 50,000%. This rapid rate of growth in the microelectronics industry could be the backbone for autonomous development in the automotive industry. Companies like Nvidia are leading the way for specialized computer chips in self-driving vehicles; the competition among leading companies to create the world’s most powerful chip (and sign deals and partnerships with automotive manufacturers) will only draw the possibility of level 5 autonomy closer. Additionally, the funding for autonomous development has been increasing in recent years, arguably driven by Tesla drawing people’s attention to their electric and semi -autonomous models. This heightened interest for the industry has led to more companies exploring the idea of their own fleet of autonomous cars. For example, General Motors announced their investment of over $20 billion USD into all-electric and autonomous vehicles ( Researchers make breakthroughs 2020). This willingness to invest and spend encourages investors from around the world to buy into the companies, improving the share prices for the industry and pushing forward the growth of autonomy.
21
Made with FlippingBook interactive PDF creator