Imagine the future of mobility. Everything is autonomous and happens almost instantly. You don´t need to have a car, you don´t need to be able to drive. Fleets of autonomous vehicles are waiting to move you from A to B. But do you need in-depth tech knowledge, experience to handle complicated apps and communicate to dozens of computers on four wheels? This part doesn´t seem right. Just think of your grandparents trying to handle their Wifi, how can we expect them to be as common with modern technology as we are? The future of mobility needs to be democratic. Everybody - techies or grandparents, bankers without time, disabled people - have to be guaranteed the same access. Come on, let´s catch a ride!
What we did
Our approach to democratic access to mobility is to have the entire user experience (UX) of mobility in mind and rethink every aspect towards intuitive usage without any prerequisites, but for all demands. The goal is to gain a human-centered model, for every possible human. Consequently, in the middle of our model is the customer with his entire UX, requiring a ride, having the mobility and all the regarded services to his demand. To address everybody, the contact from between human and machine needs to be intuitive and possible under diverse circumstances. An app-based approach is thus only completely satisfying for those using an app. For those not using an app, this approach needs to be extended to a non-app-based communication. A main challenge is the detection of a person requiring transportation without usage of an app.
We belive, that in the future a fleet of intelligent vehicles communicates to each other. Individual mobility is of course still apparent, but the majority of transports will happen on a shared basis. With this fleet communication is possible to each car, whether it is already occupied or not. The customer may now send his signal to a taxi. It may be the techie via app, the grandparents via telephone, or the banker just via gestures. Via gestures? With a clever AI algorithm the car can detect if a pedestrian wants a ride!
For the desired intuitive communication it doesn´t matter if the cab is already occupied. In such a case the customers location is given to the next cab, which is happy to move to your location! As soon as it has arrived it will detect you, and ask you to enjoy your ride.
But to the UX belongs more than just calling a cab and driving from A to B. We imagine a NLP algorithm talking to the customer. It greets him in the vehicle and asks about the destination. It has knowledge about the city - take me to the best beer garden in the vicinity! With grandparents the conversation might cause the car to drive more slowly and carefully. With an emergency mode the fleet might optimize the concerning vehicles way to the next hospital. Of course, mobility comes at a price. How can a banker, working during the ride, have the same well-being as the grandparents who don´t use modern payment methods at all? As a result, a democratic technology needs to be suitable for different payment method. We imagine a traditional cash automat for the grandparents. Their UX should not be changed too much for them to still feel well. Cashless paying via card or e-pay methods should be allowed as well. Particularly for visitors of our future city this aspect is crucial. And considering the banker, it might be reasonable to minimize the customers input via digital technologies. Our idea is a user-profile with face-recognition and deposited payment account. In this way he can only jump into the cab he called via gesture, and directly go. His face is recognized and the payment is automatic.
To the real model also belong considerations regarding the implementation into real traffic. Of course, an abstract model of the future with fleets of Taxi is used. But still, some problems remain. For example the autonomous vehicle shall not be programmed to violate traffic rules if avoidable. How would the case of double-parked taxis look like, if there is no other parking possibility?
Still if the model was a reasonable approach for usage in a dense demand area, the utility for all areas, where cabs don´t drive by occasionally, needs to be sustained. In these cases additional benefit of non-app-based communication should be veryfied for such situation.
Additionally, we discussed various aspects regarding implementation of our model into the real world, or the future world we imagine.
How we built it
Challenges we ran into
Defining the festure to detect was quite hard and required more time than we thought. We expected many issues with regard to a realistic model, but diverse problems occurred, some of them not solvable for us. Particularly with regard to the reliable detection of a person some difficulties arised. As an example, the algorithm we used has problems if people with similar clothes - like black pullovers - are close to each other. Here it is hard to differenciate to which person an arm belongs, what plays an important role in communication to the cab. After all there are, of course, different aspects that remain unsolved towards a realistic approach. But many issues we faced were solvable in the end and made our model more realistic and thus more robust.
Accomplishments that we're proud of
First of all, we are proud of the overall performance of our team. In a short time we organized ourselves, created and managed a project and solved the problem in a team. Everybody was able to put his individual talents into our project. Technically spoken, the environment, that simulates the car with an camera and communicates to possible passengers is a component we are proud of. Our approach was to create a framework which shows some basic aspects about the real world, and we think that our hierarchy is a good approach. The car talks to our backend, that would also direct nearby cars in the fleet to the passenger, if the car which detected a passenger was occupied. Also, we built an interface for the fleet operator, showing all cars, their status (battery, and occupation) and the location on a map. Furthermore we simulated the cars to pick up waiting passengers. Overall we are really proud of this complete approach to the problem of mobility, that our team created.
What we learned
Since everybody from our team brought her or his individual talents to the project, we could really learn a lot from each other. While some could provide our group with strong programming skills, others managed the whole project and kept all aspects in mind. With regard to the project itself we definetely gained a lot of knowledge with regard to all the small aspects that need to be kept in mind while programming such an autonomous task. Even with all our considerations our approach is only a model and only in a model world. Technically spoken we of course learned a lot about creating a hierarchy and about the AI algorithm we used. Many aspects that might seem easy become almost unsolvable problems if converted to the real world. At all this weekend and the work on the AID challenge was an extraordinary experience for all of us, out of which we definetely took a lot for ourselves.
What's next for AID-cab
We would love to see that our approach is appreciated, so that some ideas might be further evolved. Perhaps our entire approach is such a good idea, that the principles might be further evolved. But first of all, we look forward to fruitfull discussions about our ideas and thoughts about the challenge. We have now built a first small knowledge and experience base for the area of autonomous driving. Thus learning from the hands-in experience from AID would definetely push us forward.
Log in or sign up for Devpost to join the conversation.