Today’s cars are constantly processing and transmitting data and receiving data from servers about the world around them. This data is used for navigation, to manage the engine or to deliver communication and infotainment services to passengers. In addition, cars are the major contributors to emission of CO2, NO2 and other gases. If the above data is used in a smart way one can reduce these emissions. Today, the data generated by cars is shared over with other vehicles, traffic infrastructure and private and public entities. These “connected cars” are part of the evolving Internet of Things (IoT). The growing amount of personal data generated by connected cars raises the interest of insurers, automakers, law enforcement authorities and other third parties.
Overall there is a lack of clear guidelines around data collection for L2+ operated cars for the purpose of automating/optimizing the traffic. Self-driving cars provides great opportunities for traffic management systems to collaborate and achieve more efficient traffic management. By knowing the destination of passengers, travel time can be minimized by better integrated to optimize routes, speeds and traffic lights. Such systems can also provide emergency corridors for high priority vehicles such as police cars, fire engines, and ambulances.
The pollution problem with fine particulate matter (PM 2.5) is increasingly correlated with respiratory diseases and correlated mortality rates. Even though there is a lack still of comprehensive causation models, the measured mortality rates are higher than the current Covid-19 pandemic, meaning: “Life expectancy in Europe by about 2.2 years with an annual, attributable per capita mortality rate in Europe of 133 / 100,000 per year” according to oxford university press.
Emission from Driving
A system with real-time data updates has already paved the path to Level 4 or even for Level 2 operated cars: adaptive cruise control features that minimize energy usage using traffic forecasts and known mobility patterns while also avoiding traffic congestions by adapting speeds. A better efficiency of car operation directly supports reduction of overall emission to the environment. Advancements in car operation with the system is magnified in Level 4 / 5.
Emission from Production
IVL Swedish Environmental Research Institute has investigated lithium-ion batteries climate impact from a life cycle perspective. Their report shows that the battery manufacturing is a contributor to emissions. For every kilowatt hour of storage capacity in the battery generated emissions of 150 to 200 kilos of carbon dioxide already in the factory. Even before buying the car emissions occurred, corresponding to approximately 5.3 tons and 17.5 tons of carbon dioxide. About half the emissions arising from the production of raw materials and half the production of the battery factory. The mining accounts for only a small proportion of between 10-20 percent. Therefore, a continuous upgrade of production methods along with pro-active data driven efficiency improvement is essential to curb the emission curve.
Nevertheless, these clear facts need to be contested with what current technology are demonstrating and are expected to enable until 2030.
With expected automatization towards L4/L5, the utilization rate of cars could be dramatically improved to 50% mark for the operation of taxis with the ease of connectivity, which is a 10-fold improvement vs. current 5% utilization of the car. At that operation level along accurate predictive maintenance to maintain lifetime of cars the need to maintain car current production rates of almost 100M cars per year is dubious.
A study from transportenvironment.org benchmarks the impact of battery production on the overall CO2 profile of the car over its lifetime. The study compares CO2 emitted when electricity is produced with fuel burnt. Electric cars in Europe emit, on average, almost 3 times less CO2 than equivalent petrol/diesel cars.
Depending on the Energy mix of each EU country, the overall per lifetime CO2 emissions can be reduced by up to 79%, with an average of 63% in the EU. Most of the impact being the improved efficiency of energy conversion of the electric car during its operation and the depending on the mix of renewable or CO2 neutral (e.g. nuclear) energy resources deployed in the grid.
Improvements to reach the aggressive CO2 emission roadmaps of 2030:
The 20g CO2 eq/km corresponding to the production of cars and battery cells could be reduced by 10x by enabling fleets to operate at that increased utilization rate.
Developments to increase the reliability and predictability of storage capabilities for renewable but volatile energy sources, provided by newcomers such as Twaice.
Further 10-20% and beyond improvements in the overall range of the car based on adaptive cruise control schemes and blending of hybrid cars on top of current performances.
Currently predicted 30% reduced ranges for L4 operated cars needs to be addressed by less power-hungry machine learning models. Current training requirements of new models are impacting dramatically the CO2 emission in cloud environments
This makes it clear that operating the car with fully deployed L2+ autonomous driving and hybrid / electric powertrain are the path to reach the aggressive guidelines set by the EU in 2030. In order to achieve this appropriate improvement of the energy efficiency, the training robust machine learning models is required.
It is essential that legislation on data aligns and does not hinder these developments in machine learning models.
One can contrast these goals with current government policies. In Germany, Bavaria’s Prime Minister has already requested a program for the automotive industry, together with manufacturers such as BMW and the VW Group. With the possibilities to provide a state premium for new cars including fuel/diesel powertrain cars. As an after math of corona pandemic, set to create a quick restart for the most important industrial sector. While the German Association of the Automotive Industry (VDA), asked for a premium to be paid for new petrol and diesel cars. The state bonus is only intended to boost the purchase of environmentally friendly new cars. To create a double chance with the new environmental bonus. As an economic measure, to boost the economy back and at the same time accelerate the switchover of customers.
Emission reduction through Software
Transition to electro mobility complemented by better utilization of cars will help pull to down the emission curve. Another major contributor to help reduce the energy consumption and thereby emission through the software that compute the essential data generating sensors. An increased efficiency in energy consumption and an efficient operation will help in neutralizing the CO2 impact of cars. The energy efficient processing this is key for enabling autonomous driving and getting a lower production cost of car. Efficiency improvement in computing power consumption and reducing the computational power needed can save up to 1 ton of CO2 for every 3 cars each year as described in our previous blog. Concrete measurements done with real hybrid cars on real roads exposed a 18% - 25% energy efficiency gain; of which smarter “engine blending” delivers 9% (see our blog). Furthermore, real-time data inference along with connected features also supports to make full use of adaptive cruise control functions that further improves energy consumption by 10-20%.
A well-known fact in most AD development stacks deployed today is that 80% of data collected and processed in data servers is discarded afterwards, meaning that meaningful data selection in the car will become an essential component.
The growth to connected level 4 cars is a long process. While engineers make the cars technology truly autonomous, the essential component towards a sustainable utilization of cars in mobility ecosystems is an increase in their efficiency. It is imperative to make sure that cars L4 function remain as CO2 neutral as possible i.e. without any decrease in the car’s mileage range. Efficient edge computing of sensor data to achieve L4 functionality is part of the solution.