At the conference in Ludwigsburg that I attended, Volkmar Denner, head of management at Bosch, talked about the future of connected mobility. Bosch is the largest tier-1 automotive supplier in Europe. However, they are not limited to automotive and also supply products in the IoT space (in fact, they supplied 27M IoT products in 2016) and you might even have a washing machine or a dishwasher made by them. Bosch has 20,000 software engineers, of which 4,000 are working on IoT. One interesting datapoint: about half of all Bosch electronic products today have an IP address, by 2020 it will be all of them. Some of the changes in vehicles are clear. There remain big technical challenges, but the implications of electrification of the power train, and autonomous driving, are fairly clear. But the implications of connectivity are much vaguer. If you look at the pain points of mobility today, many of these are potentially improved with connectivity and appropriate services: 30% of all urban traffic is people searching for parking spaces The average commuter spends 42 hours in traffic jams each year Goods distribution is inefficient, with 50% of delivery costs incurred in the 'last mile' Asset utilization is poor with private cars sitting idle about 95% of the time Dropping down a level, connectivity gives three things: See what others can see (traffic jam warning, vehicle approach warning) Collect and analyze data (real time parking spot maps, preventive maintenance) Match supply and demand (ride sharing, ride hailing, traffic flow management) The Internet Enters the Vehicle; the Vehicle Enters the Internet One big change coming is where the car fits into its environment. A non-connected car, like most vehicles today, are a big expensive product that controls its environment. There has been some change with smartphones, where people often don't care what features their sound system or GPS system provides in the car, they just want to use the one they are familiar with in their phone (for example, i have never listened to the radio in the year and a half I have had my current car). The big change, from the point of view of the car companies, is that the car is one part of a person's digital lifestyle. In some ways it is a third living space (home and office being the other two) but it draws information from both, such as knowing when you have to be where from your calendar, or controlling appliances at home from the vehicle. The big change is that the car has to fit into the rest of the environment, it is not the master with everything else as a slave any more. Communication Infrastructure How to use communication is an area of debate. Obviously, autonomous driving functions need to be handled on the vehicle. You can't make the decision about whether a light is red or green depend on good connectivity at that very moment. But, many other functions can be handled with low latency communication, such as vehicle-to-vehicle (V2V) communication, which probably goes via a local basestation rather then directly between vehicles. The there is the entire internet behind that to various cloud providers such as Bosch's own IoT cloud. The tradeoff is that there is essentially unlimited computing power in the cloud, but that is where the latency is longest. On-vehicle there is limited computing power, for cost and power reasons, but the latency is as short as it can be. But we are not starting with a blank slate. The architecture today has separate electronic control units (ECUs) for each function. Each typically contains a microcontroller and some software. The first big change is that autonomous driving requires a central vehicle computer that groups the main decision making. This will be connected to the cloud but mainly for slow processes such as updating maps or the software. With faster, more reliable communication, some functions can be moved off-vehicle to use cloud computational resources. Artificial Intelligence Artificial intelligence and deep learning are key capabilities, and are a core skill for the whole of Bosch. They have already set up a center of AI. The big challenge is how to transfer the know-how we have as humans to the vehicle. There are three main ways to do this. First, training neural networks with large amounts of raw data (such as traffic signs, pictures of roads, and so on). Second, Bayesian networks and graphs to model situational understanding. And thirdly, reinforcement learning. We can see how this works in practice with pedestrian protection. First, neural networks recognize that there is a person in the roadway. Second, probability-based decision-making decides what to do: apply the brakes, swerve, slow down. Then depending on what happens, additional information can be saved so that handling pedestrians improves over time. One area that this can be important is handling different cultures with the same code. Pedestrians in Germany or Japan never cross on a red light (well, almost never). In US and UK, people do it all the time if they judge that traffic is far enough away. In Italy, traffic lights seem to be merely hints. These things also change over time, measured in years, so the algorithms need to gradually adapt. The bottom line is that if you and I buy the same car, and use them in very different environments, they will adapt differently. The New Ecosystem There will be a lot of companies involved in future ecosystems around cars. The car of the future will be connected, of course, and so become part of the IoT. In time, there will be new user experiences and new business models. The car will become a third living space, along with home and office. To make this all work will require a lot of openness and flexible models of cooperation. But it will be centered on the user, not the vehicle, a network of capabilities, not a master car with slave systems around it.![]()
![]()
↧