The very idea of a vehicle moving on the streets without a driver at the wheels is amazing. And yet, we might see one in the future. However, the fruition of this amazing concept, among other things, depend on how it can leverage the big data. A lot has been happening on this front. The developments can be categorized into two categories: Deep Learning and big data-related developments. Again, both developments are interrelated. Deep Learning is dependent on Big Data. The Deep Learning technology is at the core of the Autonomous Vehicle technology. Cars learn from data on its way to becoming autonomous and the data is collected from various entities such as roads, driver and pedestrian behaviour and weather conditions. Marketplaces and platforms are at the core of supplying big data to the deep learning technology. This article discusses some innovative products and services built around the concept.
Read about the following amazing developments.
The BMW CarData Platform
The BMW CarData Platform facilitates customised vehicle-related services for BMW customers through third-party service providers. Third-party service providers such as insurance and infotainment companies will, based on BMW customer consent, access customer vehicle data from the platform and offer customised services to the customers. For example, insurance companies may offer customized premiums based on factors such as fuel consumption and mileage. BMW commits to full data confidentiality and customer consent before providing data access to service providers. Service providers, however, are denied vehicle access. To be able to use the platform, the BMW vehicle must first be configured for a special SIM card. The SIM card collects and transmits vehicle data to a secure server in the form of telematics data. Not all BMW vehicles are set up for the SIM CARD. According to Peter Schwarzenbauer, Member of the Board of Management of BMW AG, MINI, Rolls-Royce, BMW Motorrad, Customer Engagement and Digital Business Innovation BMW Group, “BMW CarData will take the connectivity of our vehicles to a new dimension. Our BMW ConnectedDrive customers will be able to take advantage of new, innovative and customized third-party services in a quick and easy manner. Protecting vehicle data is part of our understanding of premium in the highly-connected vehicle. This is what customers expect from us. In this way, we are allowing customers to decide what happens with their data. This is precisely the philosophy behind BMW CarData.” Now, as the CarData platform expands to other vehicles and become more versatile, imagine the volume of data that will be collected. When you think in terms of teaching an autonomous vehicle to drive safely and intelligently, the data can be an extremely important input for the deep learning technologies. The varied data allows the car to learn about diverse situations and act accordingly. So, while the Deep Learning technology is directly contributing to the autonomous vehicle technology, big data suppliers are playing a quiet but important role. Same applies to the Otonomo Marketplace which is discussed later in this article.
Automatic Driver Assistance System (ADAS)
The ADAS helps improve driving safety by providing various inputs to the driver. For example, it can detect driver drowsiness or predict collisions. ADAS is not exactly the idea to leverage big data but it is an intelligent system nonetheless. In fact, you can treat it as the precursor to the much-superior autonomous vehicle. It does have an element of autonomy because it can provide inputs real-time without any human intervention. Many ADAS products are available in the market and they are also being used. For example, the hill descent control, driver drowsiness detector and the collision avoidance are some of the common ADAS used.
Otonomo Cloud Marketplace
Since big data is treated as a major enabler of the autonomous vehicle, the Otonomo Cloud Marketplace could not only be a major step in that direction but also could be the indicator of a trend, that of commercialization of vehicle data. Otonomo is a marketplace or platform used for safe, secure, and simple aggregation, distribution, and consumption of vehicular data for all types of car manufacturers. Otonomo provides car data such as usage or depreciation data; car condition data, like mileage; car resource data such as average fuel consumption; and event data, such as automated service calls. The Otonomo Platform is accessed by various car manufacturers, Original Equipment Manufacturers (OEM) and service providers. The consumption of the data can potentially result in a better and safer driving experience. For example, drivers found to be breaking traffic rules regularly may be charged a higher insurance premium as a motivation. The platform is secure and confidential, and data is shared after obtaining explicit customer consent. According to Otonomo co-founder Ben Volkow, “The understanding that auto OEMs (original equipment manufacturers) have been accumulating crucial vehicle data at growing costs, but that the data was not being utilised or monetised despite its indisputable value, was probably the spark that led to the building of Otonomo”.
Deep learning is the process through which a vehicle can learn about road conditions from data and adapt accordingly. Deep learning is one of the most important factors governing the vehicle autonomy. Research is being done on enabling the vehicles learn from data and emulate human drivers and some encouraging advances have been noticed. Nvidia, a leading provider of interactive computer graphics, has been working on a product known as the PilotNet. PilotNet enables a car to learn from data and ply on the streets without any adverse events. The car derives its data from two sources: road images and driving behaviour data recorded by another car driven by a human. So, the car learns by observing what the human does. According to the Nvidia blog, “We trained our network to steer the car by having it study human drivers. The network recorded what the driver saw using a camera on the car, and then paired the images with data about the driver’s steering decisions.” Based on the inputs, the car can recognize various road entities such as traffic lights; road markings; bushes along road edges and boulevards. As development progresses, there are already a few encouraging signals. The car can evolve based on its learning from data and show increasing signs of maturity while plying on the streets. The product has already been tested and the car can successfully move after adapting to various road conditions. On the same note, a driverless shuttle has debuted on the streets of Las Vegas, offering rides to limited number of passengers. The driverless shuttle, made by a French company Navya, is still on a test drive but is expected to go the full mile as an autonomous vehicle.
Five levels of autonomy
To set the standards and definitions of an autonomous vehicle system, the National Highway Traffic Safety Administration in the US has adopted the maturity levels defined by the Society of Automotive Engineers’. This sets a reference or benchmark for creating autonomous vehicles. The autonomy level begins at zero and ends at 5 with zero indicating no autonomy and 5 indicating full independence from human intervention. The significance of this initiative lies in the official recognition and the terms of reference. It should be noted that the standards apply to the autonomous systems which will govern the vehicles, not the vehicles. According to Bryant Walker Smith, an expert in driverless cars and a professor at the University of South Carolina School of Law and School of Engineering, “”A Level 5 automated driving system could be in a vehicle with or without a steering wheel”.
While Deep Learning is at the forefront of driving the autonomous vehicle technology, big data is at the back end, playing its part. While a lot has been happening, autonomous cars have still a bit to go before coming to be regarded as a truly autonomous vehicle. Though experiments across the world have been encouraging, it is important to note that such experiments have been conducted in a constrained environment or on a pilot basis. The technology part apart, before the driver-less car truly becomes independent, there are other issues to be sorted out as well. For example, how do you manage the insurance part if a driver-less car is involved in an accident?