Opportunities and Challenges of AI in Transportation
Increased autonomy provides major opportunities in road transportation. It lowers the costs of intercity transportation, as the need for privately owned vehicles and parking lots is reduced. The costs of operating a public transportation service can be reduced, as fewer drivers are needed to keep a bus fleet operational. Sensible 4 has already demonstrated that autonomous last mile transportation is feasible in city environments in our pilot deployments in Norway.
In addition to cost savings, autonomous vehicles also make road transportation safer. One of the most important safety features of autonomous vehicles is the ability to continuously observe a full 360 degree view around the vehicle. This means that the autonomous vehicle is always attending the road and the surroundings. Compared to human drivers, the autonomous vehicle is never tired nor checking the cellphone for the latest notifications.
A third opportunity offered by autonomous vehicles is situational intelligence. The vehicles collect by nature large amounts of information of the environment where the vehicle is driving. This data could be used for high definition city map creation, land surveying, or road quality monitoring.
As the Opportunities Are So Clear, How Come Vehicles Are Still Driven Manually?
Let’s consider an example of travelling from Helsinki to Oslo. The distance sure is long, nearly 2000 kilometers or 1250 miles, but finding a route using a GPS navigator is simple. Also the behaviour of the vehicle and traffic rules are well known, so in principle we should be able to create a computer program to drive the journey. Unfortunately, as we dig a bit deeper into the problem, we encounter multiple details which still are not fully solved.
One detail to consider is the adaption of the planned route to the local environment. A simple roadwork or a traffic accident may require the vehicle to re-plan the route using only local observations of the situation, which can still be challenging for an AI.
Another issue is robust vehicle control. As long as the vehicle behaves normally, things usually work fine. But flat tires can happen, the road may be unusually slippery, and the sensors may get covered with dirt or snow. The AI needs to be able to understand when things are not working as expected, and to be able to act accordingly. However, guaranteeing that the AI works correctly in all the possible rare situations is still difficult.
When Can We Expect That These Problems Are Solved?
At Sensible 4, we are launching our full stack self driving software, Dawn, in 2022. The product allows our customers to transform any regular vehicle into Level 4 self driving vehicle, meaning that a single operator at a Remote Control Centre can supervise the operation of multiple autonomous vehicles simultaneously.
Our software is able to solve various challenges in autonomous driving. One example is local re-planning of routes, which the remote operator will check before the vehicle implements them. Other examples include robust vehicle control and measurement systems, which allow the vehicle to drive in unusually harsh weather conditions, and system diagnostics, which alert the operator whenever anything unusual happens.
If you are interested in the company, our software and products or the related research, please feel free to contact us:
- Dr. Antti Kangasrääsiö, Head of Research, email@example.com
Written by Dr. Antti Kangasrääsiö, Head of Research at Sensible 4