AI is a perfect complement to electric vehicles and can help development of e-mobility and monitoring of its extended ecosystem

EVs are said to be computers on wheels that are increasingly more dependent on advanced monitoring systems. Powered by semiconductor chips, these systems relay data on every aspect of the vehicles’ performance, including acceleration, braking and collision avoidance and the wear and tear of individual components.

Fig. 1: Semiconductor chips are the heart of all electronic components these days, including the ones used in EVs | Image: IndustryWeek

Artificial intelligence (AI) is therefore a perfect complement to EVs as AI is essentially built on recognising and predicting patterns. AI analyses enormous data sets and predicts how a certain parameter or phenomenon may evolve in the short- and/or long-term. Also known as Machine Learning, the process helps AI be an excellent analytical tool because:

  • It can analyse much more data than is humanly possible
  • It can spot patterns and draw inferences that would be too obscure for a human observer, precisely because it is able to analyse much greater volumes of data

These attributes are thus being deployed on modern vehicles that have hundreds of ECUs (electronic control units) to monitor on-board systems for optimal performance. AI will help the development of electric mobility not just in terms of improving the vehicles, but also by monitoring the extended e-mobility ecosystem.

How will AI help develop electric mobility?

  • Autonomous Driving: Tesla is leading EVs when it comes to developing autonomous driving. The system relies on a bevy of on-board cameras and sensors to help the vehicle understand its surroundings when on the move. Autonomous driving is gaining traction because it promises to be safer by removing the dependence on an inattentive human driver, and by following traffic rules with maximum fidelity.

    Fig. 2: Tesla’s Autopilot is the proprietary name for its self-driving technology, but it’s yet to be perfected | Image: Time

    The system therefore will need a constant feed of variables, such as the proximity to other vehicles on the road, their speeds, the need to speed up/slow down to drive in sync with the traffic, the vehicle’s position with respect to lane markings, the need to brake or speed up in an emergency situation etc. All of these inputs simultaneously could overwhelm a human driver, which is why an AI system manager that performs fast, split-second manoeuvres, would be a superior alternative. Of course this assumes that the system can take the right decisions exactly when needed, but given the amount of research being poured into self-driving EVs, this could be perfected within a few years.

    Fig. 3: Roof-mounted cameras and body-mounted sensors on Waymo’s self-driving cars in the US | Image: Wired

  • Battery management: An EV battery’s performance varies according to the driving conditions. Depending upon the amount of charge left in the pack, the ambient temperature and the load on the vehicle, the battery’s performance and hence its charging needs will be different. This would be difficult for most EV users to keep a track of unless they are particularly well-versed with how the battery behaves under all circumstances.
    AI, on the other hand, could analyse a multitude of factors and optimise not just the flow of charge from the battery pack and its cooling, but also interact with the charging station to optimise the recharging process. Tesla already uses AI to enhance driving ranges, extend battery life and to lower their energy consumption. Moveover, AI would also be useful for the driver to select the best possible time to recharge the vehicle, based on their location, the availability of charging stations nearby, the charging tariff and the time it would take to recharge the battery.
  • Energy efficiency: EV owners still suffer from range anxiety, which is why making the most out of a battery pack is paramount. AI could be useful in optimising the use of the battery pack by analysing the driving characteristics, the ambient temperature and the distance to the destination. Most importantly, it could warn the driver to stop at a safe location if the vehicle cannot make it to its destination. The same attribute could be used to optimise EV fleet operations to lower operational costs.
  • Predictive maintenance: EVs require maintenance just as much as ICEVs do. They lack a combustion engine but their motors, brakes , batteries and suspension components need to be examined and serviced at regular intervals.
  • User interaction: EVs come with a lot more electronic interfaces nowadays that rely on GPS connectivity, wi-fi connections with the driver, voice recognition and voice-activated commands. An adaptive AI that constantly learns more about the vehicles’ users can optimise their experience by bringing up functions in the vehicles based on the time of day, the destination and the level of battery charge, without the users themselves having to worry about them.

    Fig. 4: Data leaks are a real threat in using AI with EVs | Image: US Dept. of Energy

The risk of using AI with EVs

AI will thus be a major addition to the EV ecosystem. However, every electronic system is vulnerable to cyber attacks. These attacks can overpower the vehicle and take away all control from the driver, as well as leak the user’s information and identifying variables to cyber criminals. Data encryption and air-gapping the AI to the vehicle and its manufacturer alone will thus be critical to prevent a security incident.

Yet, overall, AI-assisted EVs should perform better than everyday EVs and it will make them easier to integrate with nationwide smart grids.