IoT and Machine learning enable data-driven sustainable mobility

By Lev Levine, Product Director, CO2 & Energy Management, ZF Group

Environment, sustainability, and governance (ESG) is a big focus for consumers, businesses, and governments to the extent that corporate sales, profits, access to investor capital, and ability to retain top talent depend on the company’s ability to measure and reduce their CO2 footprint. Regulations also play a big role in creating a risk of the high cost of incompliance. 

Mobility assets are a major source of carbon footprint, typically generating 10% to 90% of the company emissions, depending on the industry. These include not only company fleets but also the upstream and downstream supply chain, travel, and employee commute.

How to enable the transition to sustainable mobility in a business environment? How to implement this transition in an economically feasible fashion?
To answer these questions, companies commonly need to look at three domains:
(1) accelerating the adoption of digital technologies, (2) building an open “carbon” ecosystem coverage, and (3) leveraging new business models. This article will focus mainly on the technology aspect of sustainable mobility transitions.

In modern vehicles, all the systems, e.g., powertrains, suspensions, electronics, and safety, generate high volumes of data from sensors and ECUs. These data can be extracted from the vehicle systems given data privacy permissions. Vehicles have high-capacity data processing and data storage capabilities, that can be used to aggregate and preprocess the data in-vehicle and deliver edge applications. Vehicle data and metadata can also be communicated to and from the Cloud for big data applications.

There are multiple ecosystem data sources that are particularly important for sustainability management, such as supply chain operators, power grid companies, EV infrastructure operators, fueling operators, local governments, and parking operators, to name a few.

With all the IoT data potentially available come a few challenges. First, robust data protection and consent management processes compliant with GDPR should be in place to enable drivers to securely share their vehicle and behavior data. Another issue is related to data quality: the inconsistencies in dimensions and frequency of data across OEMs make it difficult to perform the analyses. Also, the connectivity and data transport vary from OEM to OEM: some OEMs support API integrations, but limit the usage of third-party HW, while others do the opposite. Finally, integrating the ecosystem partners can be costly and technically challenging, especially in an environment of missing or multiple communication protocols and data standards. It’s important to map the relevant data points (features) to the sustainability targets (labels) and build the data pipeline that can normalize, protect, and ingest the data across a variety of mobility assets.

The next step is to understand which types of operational and business insights can be extracted from the data to generate sustainable outcomes?
There are five fundamental decisions that could be made to manage sustainable mobility: (1) fleets electrification decisions, (2) Electric charging decisions, (3) Driver behavior decisions, (4) Vehicle system technology decisions, and (5) Travel decisions.

Fleet electrification decisions are based on the carbon footprint consequences of the vehicle lifecycle: manufacturing, operation, and disposal. Selecting the right vehicles for the fleet upgrade depends on the daily driving distances, urban vs. highway driving, whether heavy-duty or passenger vehicles, and other factors. Ironically, the Battery Electric Vehicles (BEV) are not always the lowest CO2 footprint choice due to the sustainability impact at the manufacturing and disposal phases. Plug-in Hybrid Electric Vehicles (PHEVs) could, in some cases, deliver better sustainability outcomes and economics. The decisions on which vehicles to use can be complex, costly, and should be data-driven.

Electric vehicle charging decisions are focusing on optimizing the location, time, duration, power, and direction of charging. Here the goal is to maximize the usage of energy generation with a minimum carbon footprint to charge the electric vehicle while optimizing the cost and vehicle downtime. The rational EV charging decisions can be complex and should be data-driven.

Driver behavior decisions could include speed, acceleration, routing, charging schedule, and driving mode (electric vs gasoline) optimizations. While representing a high-potential emission reduction opportunity, they do not require high-CapEx investment into new vehicle systems, but rather access to data and the ability to build and serve high-accuracy AI-models. It is also important to use gamification, behavior economics, and incentives to encourage the driver to follow the recommendations.

Vehicle system technology decisions are made about which vehicle systems to use and how to optimize the performance of these systems to reduce the lifecycle carbon footprint. These decisions are either part of OEM product engineering or aftermarket fleet technology applications and are data-driven.

Finally, travel decisions are focused on whether to travel or collaborate remotely, and which routes and modes of transportation to use. The goal here is to minimize the carbon footprint of the trip and/or per cargo box while containing the travel time and cost. Such decisions usually require big data analytics and AI modeling.

Mobility data insights that help drive sustainability decisions are inefficient and, in many cases, impossible to obtain by building rule-based algorithms. There can be hundreds or thousands of columns and millions or billions of records in each data set, the data sets can be streaming, and patterns can change over time. Developing data pipelines and building machine learning models that can be continuously retrained and served in the cloud and at the edge (in-vehicles) are key to embracing the IoT – generated data and delivering accurate classification and prediction results to support sustainable mobility decisions. IoT and machine learning technologies have been benefiting the automotive sector for many years. Now that sustainable mobility is at the top of the agendas of businesses, governments, and consumers, it is time to focus on leveraging these technologies in developing next generation solutions that can deliver emission reduction while generating operational efficiencies, savings, and new revenue opportunities.