How AI Could Help Boost Battery Life?

You may no longer have to worry about your smartphone’s battery life! 

No, there is no super battery on the market just yet. 

The Android team and Alphabet’s DeepMind are using Artificial Intelligence (AI) Technology to add a new “Adaptive Battery” feature to manage battery life. 

The AI will constantly monitor the battery consumption of your phone and will close apps that you don’t use for months but still run in the background. The new system will also keep adjusting your phone’s brightness according to usage.

The method uses software-generated data instead of power-intensive sensors, such as GPS or microphones.

AI forecasts depend on the app's users have and when they use those. Factors, including the time and period of phone usage, also influence the result.

How the Technology will Impact Phone Manufacturers and Consumers

MIT professor Richard Braatz and an assistant professor at Stanford William Chueh have led researchers to design a machine learning technique that can make a prediction of battery health with 10x higher accuracy than the existing industry standards.

This development could help to make safer and more durable batteries for electric vehicles (EVs) and consumer electronics. This particular problem is limiting the widespread adoption of electric vehicles.

The issue of anticipating the remaining lifespan of lithium-ion batteries also troubles mobile phone users. Over a period, the performance of batteries declines due to complex and subtle chemical processes.

The new method sends electrical pulses to the batteries for monitoring and measuring the response. The researchers have performed more than 20,000 experimental measurements to train the model, which is the largest dataset for this type of experiment.

The machine learning then processes the measurements to predict the battery’s health and potential lifespan. This method is non-invasive and only requires a simple add-on to any current battery system. 

The method only consumes battery on the apps you often use, and there has been a 30 per cent reduction in CPU usage on average when waking up apps. All your data will remain private on your phone, as AI will solely run on your device.

The AI Also Optimizes Fast Charging

One crucial focus area of researchers was to find a more efficient way to charge batteries in 10 minutes. This feature can make batteries run much longer in phones and speed up the mass adoption of electric vehicles. 

To create the dataset, the researchers repeatedly charged and discharged the batteries until each battery reached the end of its life cycle. 

While optimizing fast charging, the researchers were also looking for answers to the following questions: 

  • When a smartphone automatically updates itself without affecting its usage? 

  • What is the right time for a device to download shows, and podcasts and synchronize photos? 

  • Is there a better way for phones to launch their data synchronization without leaving users with only 10 per cent battery life?

Earlier, Stanford graduate students Nicholas Perkins, Norman Jin, and Peter Attia found a way to accurately predict the useful life of lithium-ion batteries during the research. 

The rapid advancement in computational power and data generation in machine learning has enabled it to complete various tasks, including the prediction of material properties. Experiments show how it is possible to predict the future behaviour of complex systems.

Usually, the capacity of lithium-ion batteries remains stable for some time. Then it goes downhill sharply. The point of decline varies widely, which consumers experience first-hand.  

In the said research project, the batteries lasted between 150 and 2,300 cycles. Such a variation was partially due to testing different fast-charging methods and manufacturing differences of batteries.

So far, after all the investments of time and money in developing batteries, the progress results are still measured in decades. In this type of research work, scientists are eliminating battery testing, which is one of the most time-consuming steps.

Closing Words

To sum it up, although it can be challenging for AI to forecast when some arbitrary events occur, it still shows great promise when the occurrences are more regular. For example, many consumers check their mobile phones before going to sleep and lock their devices until morning.

Based on the newly-developed AI system, smartphone operating systems and apps can minimize unnecessary energy consumption because of downloading content, especially when consumers are not supposed to use their phones.

During periods of non-usage, smartphones could schedule tasks that require more computation, such as app and OS updates, or some other activities that can affect the user experience. 

Instead of checking for the latest updates every five minutes, your cell phone can only check once.

In the near future, large-scale implementation of this type of AI system on smartphones could help save more battery life throughout the day and increase their usage.