I took my 15-year-old to her first big concert recently – Roger Waters’ Us + Them tour. If you are wondering where am I going with the blog on automotive technologies and talking about Roger Waters, stay with me for a second. She and I both loved the concert, however, as I drove to and from the concert, I struggled with the stress of congested traffic and getting in and out of the parking lot. During those high-stress moments I imagined how wonderful it would be to have an autonomous car that would pick me up and drop me off and navigate through the mess while I could relax. Like a cab, minus the driver that I would have to compensate and tip. There are many technological, infrastructural and legal hurdles we have to cross before these cars can be unleashed onto roads. However, one can’t help imaging how convenient will it be to have a self-driving vehicle to drive us.
Automakers are investing heavily in technologies to realize a “driverless” car dream. The vehicles of the future are going to be electric, connected and autonomous. We have made much progress and come a long way in gaining valuable knowledge that is helping us realize this dream. Here are some technological advancements that are critical to get a fully autonomous car on the road.
Computers have shrunk into sizes that can fit into the palm of our hands. The computing power has increased exponentially, enabling efficient operation of very complex operating systems, data processing and machine learning. A similar trend can also be seen in the sensors, actuators, and processors used in cars. The sensors have reduced in size, the data resolution as well as screen resolution has increased, accuracy has improved and of course happening at lower energy consumption. We are able to reduce cost of manufacture due to mass production. As an example, the extremely large (and very expensive) LiDAR sensors on top of the autonomous cars have been miniaturized into a lower cost and compact size that can be hidden away inside the bumpers.
Machine learning is another technology which has seen increased applications, partially due to its computing hardware improvements. As improvements and tweaks in these machine learning algorithms are being made, the hardware also needs to be flexible as well as configurable in order to accommodate these changes. Automotive FPGAs are the ideal candidates for that.
Machine Vision, combined with Machine Learning, is key hardware technology towards driverless vehicles. Many of the laptops and smartphones now come equipped with facial and object recognition feature for logging in users. This technology will help recognize pedestrians, lanes, and objects like trees and traffic signs for autonomous vehicles. Vehicles with Level 4 and Level 5 automation will be equipped with hundreds of such sensors, which means that hardware has to be able to process information in real-time and use machine learning to make split-second decisions similar to humans.
Operating System Software
The key component that will enable full autonomy in the cars is software. Software technologies used in a vehicle’s operating system help us make cars that are configurable and improve their autonomy. Once again, software-based machine learning applications are adding the intelligence aspect to the cars. The immense complexity of the modern operating system, especially for autonomous cars will be matched only by its security, completeness and safety. In fact, these aspects of the operating systems will be crucial for the vehicle’s ultimate acceptance by the masses.
Operating system complexity increases as one moves from Level 1 to Level 5 of the automation. It needs to be very flexible, highly configurable and constantly updated, taking advantage of the developments and updates in the infrastructures and applications. The safety and security features will require frequent updates because the current model of updating software every six months or every year will not work. We are at the cusp of emergence of the new ecosystems around cars.
Automotive industry has not tackled a problem of this scale before. It has to focus on the deployment of a software that is developed, tested, secure, and reliable. It is expected that the characteristics of this software will surpass the software currently used in some of the state of the art operating systems. The automakers will play the leadership role in electrification/ hybridization of the vehicles and improve ability to deploy such an operating system. They will also have to gain the public’s trust that the software they produce is of the highest quality and reliability autonomous vehicles require.
Another requirement for the application software for vehicles is that it must be intuitive to use. It has to perform data gathering, processing, sensor fusion, aggregation and presentation. The operating systems have to be able to acquire and process all the information and present it to an average driver in a useful and simple interface. The end result will be a rich human-machine interface (HMI), featuring a set of streaming entertainment options that are accessible by the vehicle’s occupants while being transported.
Navigation and Mapping
Digital mapping, different from cartography, is the digital representation of our physical world. Current mapping software on our smartphones combine cartography along with environment information or big data. Although the current algorithms are good enough to get us to our destination, they are only good for autonomy Level 3 vehicles. Fully autonomous transports will require much higher resolution and accuracy of say 1 cm (unlike current accuracy of about 1 meter).
Mapping has to evolve from being static maps to more interactive, frequently updated maps that not only take into account the various objects on the road that will have an impact on driving, e.g. a new construction, trees, changes in traffic, road signs, accidents, but can also accurate locate your position on the road. Lane departure and GPS technologies have to accurately identify your position and effectively move you to the best lane to make your exit or avoid obstacles on the road.
Autonomous vehicles may need live updates to push every few seconds. Some innovative startups have already come up with unique ways to improve the user’s navigation experience. A good example is the Waze app, where the company uses crowdsourcing that capture live data from its various users that help them provide real-time information to their users. But there is much more innovating to be done to meet the need of an autonomous car.
High Bandwidth Communication
Last but not least is communication or big data for the autonomous vehicles. The advancements in hardware, software, and mapping are all leading smart cars to acquire, process and understand data from variety of sensors. The intelligence of these cars will be to make use of the data to drive autonomously.
The sheer volume of data generated and the communication between vehicles, and between vehicle and infrastructure will be ginormous. These smart cars need to communicate constantly with each other, making their presence known and sharing live information about the road conditions and hazards. Just imagine a future with thousands of such vehicles on our roads.
The future cars will combine the local edge computing as well as cloud computing. These cars will be servers on wheels – collecting data, processing, storage and communication servers on wheels. A new breed of companies has emerged and will continue to emerge that will combine knowledge of auto manufacturers and Silicon Valley. Just visit any auto show, you will notice computer processor manufacturers are now major players in autonomous car market.
The advancements in technologies are helping evolve the human dream of smart autonomous cars. As we are moving forward, the technologies enabling autonomous cars have to be flexible, responsive, configurable and able to handle large data while learning complex algorithms. But until this tech is perfected and ready for the road, I will continue to dream of a car that will drive me around, especially to concerts.