At present even with AI driver assisted cars they are crashing with risk of error estimated in thousands of miles and not millions of miles. While a driving system may have been trained to the equivalent error tolerance in millions of miles and thus far more than the service life of the product. It is expected that travel is more complex than the trained systems and thus the significant difference between simulated events and real-world decision-making. AI is fundamentally a weighted decision-making matrix used to assess input source data such as from sensors and then filter and decide on a decision at each node of the matrix. To create the weighting factor the matrix is ‘trained’ through iteratively passing database data of example data through the matrix until the error tolerance is within specification of the AI system. The challenge is that the error tolerance could be within a specified value; however the matrix may not represent the real-world decisions that the product requires; this is where vehicle to infrastructure communication is required.
V2I communication reduces the limitation of sensors in vehicles, namely trying to predict the motion of the object in or going into the road or change direction which may be predicted by driver behavior and acceleration changes that may not get picked up by visual sensors. In addition to this, assessing the state of traffic lights may help reinforce an AI’s decision-making success; while it may not be needed in simple situations if you are in the middle of a brightly lit commercial district or the sensor is obscured then when to stop or start driving could be compromised. With V2I communication the two systems can reduce a risk of an incident significantly.
The clever thing about vehicle to infrastructure communication is that it can help when road repair occurs and speeds change with poorly visible signs. Highways in the near future may have systems that not just notifies the driver visually but also communicates to the driver as an alert or to the AI the speed change or change in driving conditions.
Vehicle to infrastructure communication can also be used in the opposite direction to just being car centric allowing highways or busy city routes to us AI to preempt traffic jams, increase spaces between cars to stop delays from traffic jams and buildup of traffic behind cars waiting to move along with rerouting traffic to ensure they get to destinations promptly. This could take different forms from either the AI of the car or the vehicle to the infrastructure communication system conducting the majority of the processing.
AI vehicles that use AI processors that are modular are usually designed to use AI processors in cascading data approach or in parallel or with an arrangement of both depending on need. This is important as it means that vehicles of the future may be designed either to integrate both systems fully or have them divided into discrete control modules. Vehicle to infrastructure communication will be part of the patchwork of innovation that will go into running transportation in the world. Currently V2I communication has received a warm welcome in China and Europe and it is likely that it will propagate quickly as the new US infrastructure bill is implemented in 2022. A transportation system should use the best communication systems and automation needed to stop crashes and improve travel time and experience. As different fuel technologies are rolled out across the world it may be also used to help automate refueling activities as part of journeys with AI predicting they the best time is to stop without the driver having an active input. No matter what the future will look like, it will be safer and easier than ever to get around.