Autonomous driving vehicles are a central component of the Mobility of the future. What do the concrete deployment scenarios look like and how do autonomous driving vehicles work from a data science perspective? This blog article is a deep dive into the matter and shows all facets of what is technically possible.
Assuming I want to go to my friend Katja's birthday party from home in about 10 - 15 years' time, this could look like this:
In the run-up, I phone my friend Max, who also wants to come to the party from a little further away. Since our routes partly overlap, we would like to ride this part of the route together. In order to plan this, we tie our Cognitive AI Agents into the conversation. They both suggest that we travel with a autonomously driving robot taxi and also identify the area where we can best meet to continue the journey together. As we know the exact time, we can pre-book the robotaxi at a cheaper rate and process the payment directly through the AI Agent.
Next we discuss what to get our friend and after we have agreed, I order the gift. Since the shop automatic deliveries supports, I add the gift to my Robotaxi booking so that the Robotaxi can collect the gift from another ride near the shop in advance.
On the day of the party, the autonomously driving robot taxi arrives at the ordered time and recognises me by my faceso that I can just get straight in. During the journey I take a short nap, for which I darken the windows and put the seat in a sleeping position. Max had to leave his workplace early to get to the party on time. Therefore, he continues his work while driving in his robot taxi to our meeting point.
Due to a unforeseen roadblock my robotaxi has to take a small diversions. Our robotaxis exchange this information and set a new meeting point, where we both arrive almost at the same time. On the ride together to our destination, we play cards and nibble on some of the snacks I have baked.
After we get out and take the gift out of the car, it continues on to the next customer or the next charging station. We haven't booked a ride home yet, because you never know how long the party will last and the waiting time for a free robot taxi is usually less than 15 minutes.
Is the scenario described and autonomous driving vehicles realistic or just a crank? In the following we want to examine this in more detail and answer the question from a Data science perspective rate.
Inhaltsverzeichnis
Breakdown from the perspective of a data scientist
From the perspective of a Data Scientist autonomous driving vehicles Large travelling computers represent. As such, they are connected to the internet and can transport people or objects. The story from the introduction touches on many topics that will now be analysed and presented in more detail methodically and technologically.
Cognitive AI agents are learning chatbots that can do more than just talk to passengers
The cognitive AI agent ultimately represents a Chat or speech robot which enables human conversation through artificial intelligence. Building on Basic Conversational Dialogues and Natural Language Processing (NLP) he constantly learns through the guided dialogues. By means of Deep Learning and reinforcement learning methods, it can react to both historical and current context in a conversation and adapt to previously unknown situations.
In the example presented, the cognitive AI agent performs a Mobility optimisation by requesting different mobility providers with the given context of date, departure time, destination and the condition to travel part of the way together. He can book the best solution from the suggestions directly with the provider.
Facial recognition and biometric recognition become standard in autonomous driving vehicles
The Identification through biometric characteristics will increase strongly in the future and can replace concepts such as passwords or even keys. Biometric recognition owes its breakthrough to technological progress, which makes it possible to biometric characteristics with reasonable effort and in Very high quality and evaluate them.
The robot taxi identifies me in the example by applying a Face recognition algorithm. This can use the images from the cameras that the autonomously driving vehicle needs for navigation anyway. This identification means that you no longer need a key to access the car and the billing of the driving time can be done securely and automatically.
Electric autonomous driving vehicles are large, moving computers
The car of the future is basically a giant computer on the move, equipped with high computing power is equipped. For the car to be able to drive autonomously, it must be able to get a picture of its environment at any time by reading cameras, radar or lidar sensors. The accruing Data must be processed and made available to the actuators in steering, engine control or brakes in the required form in order to be able to execute necessary controls.
In the implementation, various methods are used for this purpose. Machine Learning and Filter methods which are repeatedly optimised and calibrated with new data. For these updates, it is necessary that future vehicles are firstly able to store larger amounts of data and send it to a backend and secondly have the possibility to update software. Especially for the second point, it is important that hardware and software are more separated.
The hardware, too, should be more modular so that a Upgrade of computing power or storage space for data is possible in the aftermath. The transmission of larger amounts of data, e.g. from situations in which a Anomaly detected can be done during the battery charging process either via a WiFi connection or also via a suitably equipped power cable. Overall, the Communication of the car with other participants very important and can be roughly divided into three domains.
- The Remote Domain
The Remote domain has a central backend for the mobile transmission of smaller amounts of data and important Security Updates. This route is also used to current traffic messages replaced, like the roadblock in the example above. The Backend also receives the larger data streams sent via WiFi from the vehicle and sends, for example, larger Updates to the vehicle only if there is a WiFi connection. - The local ad hoc domain
Within theseAd hoc domain finds the Communication with the immediate surroundings communication between the autonomously driving vehicle and the road users within a certain radius. For this purpose, a common WiFi standard has been defined by the car manufacturers, which is used for the Data exchange is used. This way, vehicles can exchange their speeds, destinations and, if necessary, warnings locally. This eliminates the need for traffic lights for cars at intersections. In addition, there are significantly fewer traffic jams because speeds are adjusted in time, and on the motorway, vehicles can be grouped together in columns to reduce air resistance. - The domain in the vehicle itself
Within the Vehicle domain devices can be connected via Bluetooth, NFC or WiFi. Since the vehicle interior can be flexibly adapted to the Wishes of the driver it is possible, for example, to transmit a corresponding configuration to the vehicle. Another example is the activation of Entertainment programmes by transmitting the required data for the journey to the car.
A Robotaxi rarely comes alone - The Robotaxi Fleet
The fleet management of the robotaxis is based on Data Science Use Cases. Most important is the Fleet optimisation in terms of customer availability and waiting time with optimal utilisation of the fleet. For this purpose, with historical usage dataMachine learning models are trained on the day of the week, the time of day, the influence of major events, the weather, and the start and end points of a booking, which are then used to predict the expected demand.
In this way, autonomously driving vehicles of the fleet distribute themselves optimally in the city depending on the time of day and usage history in order to minimise waiting times for passengers. In addition, the autonomously driving vehicles independently drive to free charging stations at the optimal time in order to be available in any case during heavy traffic. The same applies to the maintenance of the vehicles.
Through Predictive Maintenance they are monitored around the clock and any impending technical problems are detected in good time so that maintenance can be carried out at a suitable time. For predictive maintenance, again Machine Learning Methods which, through training on historical data, identifies either unique fault patterns or anomalies in the car's operating condition.
What of this is already possible today?
Many points have already been made today in relation to the Connected Car implemented, or will be implemented in the near future. For Tesla vehicles, autonomous driving is already possible today in certain situations. Many other car manufacturers are currently testing systems on approved routes in order to be able to offer autonomous driving vehicles in the future.
Similarly, Uber has launched several initiatives to deploy autonomous driving vehicles in major cities in the US for taxi services. In the next 5 to 10 years, autonomous driving vehicles will hit the streets in large numbers. A good look at the current state of the You can find Connected Car here.
0 Kommentare