The 5 levels of electric vehicle charging automation
A taxonomy for human-machine collaboration in EV charging: the driver’s view.
If you are a little bit into the future of cars you, are surely familiar with the five levels of autonomous driving. They describe how vehicle driving technology will step-by-step advance from human driving to full machine driving. Based on these levels it is more or less clear for innovators what needs to be achieved next so that the machine makes the right decisions while driving: when to accelerate, when to turn right, when to slow down, etc.
But what about automating electric vehicle (EV) charging decisions? If we as humans, at a time when vehicles already drive by themselves, do not still want to decide where, when and how to charge next by ourselves, we need to invest in automating every EV charging task. To help the industry with organizing these R&D endeavors, we propose a five level framework for EV charging automation.
From human to machine decision making in EV charging
The fundamental question behind each level of autonomous driving is: Who is making the decisions for a given set of tasks? It starts with the human making all driving decisions, e.g. turning the wheel at the right point in time. Step-by-step more tasks get executed by software until the machine is able to perform all driving tasks completely automated (and autonomously)[1]. Research has proposed many different frameworks on how to describe the various steps towards completely automated task execution by machines[2]. We have reviewed the different approaches and focused on generating a model that readers in the industry can easily relate to and apply to their own work:
In the following we will go deeper into each of the steps and provide some EV charging related examples.
Level 1: Human decision making
Principle: Humans are fully responsible for the entire decision making process.
Example: Humans decide where to charge next based on their own experiences and knowledge provided from other humans.
Reason for next level: Humans can’t collect all information by themselves.
Level 2: Assisted human decision making
Principle: Machines provide relevant info that humans cannot gather or calculate. Humans make better decisions based on more information.
Example: Driver decides where to charge based on availability predictions.
Reason for moving to the next level: Too many uncertainties for humans to make a good decision.
Level 3: Recommendation-based human decision making
Principle: Machines offer various options, humans evaluate and make decisions.
Example:Driver decides on different options of charging trips provided by an app.
Reason for moving to the next level: Human decision making takes too much time.
Level 4: Delegated machine decision making
Principle: Machines get control over a defined scenario and make decisions, humans only get informed to be able to take back control if necessary.
Example: The EV decides by itself which charge options are best, the driver gets informed.
Reason for moving to the next level: Informing humans is redundant.
Level 5: Automated machine decision making
Principle: Machines fully control themselves and make automated decisions, humans are completely out of the loop.
Example: An autonomous electric taxi defines its own driving and charging routine.
Where are we today?
At the moment of writing, we see solutions in levels one to four in the market. What is important to recognize is that the quality of automation depends a lot on the use-case and its environment. The more complex the use-case and its environment are, the harder it is to achieve higher automation.
On Level 1: Human decision making we see basic home, destination or public charging without the use of any digital service. In some cases, this level is still enough, but this will soon not hold true any more. Humans simply cannot collect all the information about charging options (e.g. public charging stations along a route) by themselves anymore. Thus providers relying on level one solutions might not be able to satisfy their customers any longer.
On Level 2: Assisted human decision making we see for example charging apps that provide different kinds of information to the user (location, availability, socket types, etc.). This level still fulfills the needs of the EV driver in many use-cases, but due to the sheer amount of information and possible outcomes of decisions (e.g. which station to pick for the next charging sessions), level 2 will soon not be enough. Based on our research, still many solutions struggle to get level 2 implementations sufficient enough to move on to the next level (e.g. not enough (or correct) information about charge points is gathered). This is a major blocker in the industry at the moment.
On Level 3: Recommendation-based human decision-making we identified for example routing apps which give recommendations where to charge along a route or at a destination. We see some solutions on this level already, but the recommendations provided are often not trusted 100% by the drivers yet. There might be several reasons behind this, two of the most important ones might be that the recommendation system (1) doesn’t have sufficient information about the charging options (e.g. predicted charging availability) or (2) doesn’t learn enough automatically from past driver decisions.
On Level 4: Delegated machine decision making we can already see some home or work charging apps that adjust charging speed automatically according to electricity prices or local green energy production. Often they are limited to very “confined” use-cases, as not enough data is available (e.g. charging rates adjustment algorithms don’t respect spontaneous user needs) or regulatory questions are not solved yet.
What does the future hold?
We are looking forward to companies moving their solutions slowly up to level 5 of EV charging automation. This surely can be a challenging task, but with the right team and competencies in product, software and AI technology any provider can achieve higher levels. With fronyx, players in the EV charging industry can even simplify this process by getting access to cutting-edge AI solutions based on simple software integrations. We help you to get the advantages of AI technology for EV charging automation and to keep your focus on providing the best possible customer experience. Reach out to us to learn more.
What do you think about this model? Comment below or reach out to us on Linkedin. We are looking forward to your thoughts.
[1] Some research in the field differentiates between autonomy (“How much human guidance is necessary?”) and automation (“Which decisions can be done by a machine?”). We have focused on automation in this article, as autonomy will only become important in higher levels of the model.
[2] Bitkom Study, Umlaut Study
