Picking up and Dropping Off — Autonomous Vehicles and the Last Mile
When you order a cab, you can normally rely on the driver’s common sense to choose the best place to pick you up and drop you off. Sometimes, this driver will be aided by signs that determine curbside rules, other times they will just have to use their intuition. Such intuition will be required by autonomous vehicles (AV) too, except in this case it won’t be intuition.
At their core, AVs are disrupting the whole concept of parking and its infrastructure. A complete revolution of ownership patterns is likely to mean that parking in a traditional sense is replaced by more short-term pick-up and drop-off scenarios. Here at Parkbob, our efforts to make it easier for human drivers to park, will eventually be vital for AVs. This will be important in order to provide users with the best and safest experience whilst helping city governments to maintain orderly streets.
Implicit and explicit rules
A fundamental difficulty for AVs is replacing the ability of the human brain to interpret rules. This is particularly important in the last mile, where the pick-up and drop-off will take place. Whereas a human can gauge whether or not a pick-up/drop-off spot is suitable based on their understanding of the context, an AV is much less flexible and may struggle to accurately determine the best option. This is especially true when it comes to rules and regulations, where based on our coverage of 61 cities on three continents, 30 percent of all rules and restrictions are implicit, with no explicit sign.
An example of this is the observation of lane usage information. Bus lanes, temporary peak hours, street sweeping, emergency lanes and temporary constructions all need to be observed. Without an intuitive driver, AVs therefore require more detailed input with up-to-date street and curbside data. 2nd lane drop-offs for example, may be allowed in certain cities but not in others. As the answer is often found in city codes and traffic rules but not on signs, interpretation of the existing rules is required. Therefore, the cameras and other sensors found in an AV that can read road signs, are not sufficient in themselves.
If it wasn’t complicated enough already, some signs may even have a different meaning entirely in another jurisdiction. It is by no means a given that the same sign will mean the same thing in China as it does in Germany. Therefore, the interpretation of implicit rules becomes key. Add to this different languages and in some cases poorly printed temporary signs, and it can become very difficult to interpret the rules, even if there is a sign in place.
Missing piece of the map
Without being able to rely on context, AVs require HD mapping that mean that even if a sign or object like a traffic light is obscured, the vehicle still knows it is there. One of the problems is that accurate HD mapping is expensive and the rules that govern urban traffic are constantly changing. For example, temporary events like street cleaning may restrict parking or the rules concerning drop-off zones may be changed completely.
At Parkbob, we provide the missing piece for SD and HD maps by providing accurate curbside information and in doing so address the blind spot that these maps have. The vehicle communicates the real world situation to the “parking brain”, creating a feedback loop. Parkbob then matches the information from signs, collected by the vehicle, with our map. In case there is a difference, there might have been a change in curbside rules which is not yet reflected in our data. At this moment, there is no “brain” that could process, interpret and update a parking map accordingly meaning the collected data is “lost”. By providing the full set of rules, implicit and explicit, Parkbob enables AVs to take more accurate decisions.
In an urban mobility landscape composed of full automation, street signs may become redundant entirely with city governments relying solely on digital rules and regulations. However, till we get there, vehicles need to be able to interpret sign information and then be able to match this to a map.
In doing so, AVs will also bring more consistency when it comes to observing curbside rules. Ride hailing firms for example keep historical data on where passengers enter or leave a vehicle in order to have a better understanding of access points. However, this does not take into account the legal aspect. It is well known that drivers “adapt” the rules to suit their situation, a luxury that autonomous vehicles won’t have. Their decisions will be based on legal options only, with the potential extension towards emergency stopping zones, such as bus stops and fire entries. These would only be allowed in special circumstances like when a vehicle malfunctions or when a passenger needs to jump out quickly because of sickness.
In the last mile, AVs will require more than just HD mapping without human intuition. Here at Parkbob, we aim to use the information from digitizing the last mile, to help AVs to take accurate decisions. By codifying the human experience, we aim to fill in the gaps that make the pick-up and drop-off of users safe and legal.
Nilüfer Cipa is Head of GIS at Parkbob. You can connect with her on LinkedIn.
Parkbob organizes the world’s curb-side data and transforms it into actionable information to enable better mobility decisions. Want to know more? Visit the services-section on our website or get in touch with usdirectly in order to explore the possibilities of a cooperation!