AImotion Challenge

Startseite

Unser Angebot

Events

News

The AWARE Center invites you to take part at our international AI@motion Challenge!

You are a student (Bachelor’s or Master’s degree) from Germany, Brazil or Finland, interested in Artificial Intelligence? Perfect! We are looking for teams (2-4 members each + supervising professor optionally) which are ready to develop a special AI software to run the MobiBox.

The MobiBox is a handheld that includes all sensors and intelligence needed to move/steer all kinds of vehicles (future vision). It can be plugged into an autonomous air taxi, a self-driving car, a city scooter, a bicycle or just carried around whenever you move. You thus have one smart, intelligence device and all the vehicles, scooters etc. are plain hardware pieces.

What you can win...

Coming from Brazil or Finland: You will get a trip to Ingolstadt to participate in our upcoming AImotion Network Conference (November) and will get the chance to bring the MobiBox from the simulation to the real world, being able to be connected to a play car, a real drone and a scooter. The MobiBox has to fulfill challenging tasks already learned within the simulation on the real hardware and demonstrating it at the THI campus using a TurtleBot.

Coming from Ingolstadt: You will win a trip to an AI event taking place in one of our partner countries Brazil or Finland! A top class jury of science and industry representatives will rate your results and decide for the best solution.

Goals of AI@motion Challenge

The MobiBox takes commands in natural speech and is able to execute them in a simulated world while being connected to either a drone, a car or a scooter. The commands are of kind:

Car	Drone	Scooter/Bicycle
Turn left at the next crossing Turn right after the man with the dog Keep straight until the blue house	Find a safe place to land at the next parking area Fly up to a height of 10 meters Avoid hitting the upcoming trees by flying around on the left side	Note the pedestrian 5 meters in front of us Red traffic light ahead

Tasks within AI@motion Challenge

The goal is that the MobiBox can deliver enough information about the environment so that the transportation devices drone, car, scooter/bicyle can navigate safely within their environment. Be creative in how to achieve that!

Detect objects and classify them
Segmentation of the image so that the transportation device knows where it can fly or drive
Speech recognition and transformation into control commands
Matching of spoken words to detected objects
The MobiBox should be able to understand natual language, i.e., commands that a human just queries out of a situation

The results of the teams will be tested by taking a journey within the virtual world of the Airsim Simulator. The journey starts at a secret point and commands will be given by speech to tell the mobility device where to go to. This procedure is done three times, for each of the two transportation devices drones and scooters in an appropriate manner.

For a creative design and name of the MobiBox
For funny conversations and interactions of the MobiBox with the user
For efficient use of computation power
For reusing the same components for different tasks

Hints for the AI@motion Challenge

You have to use the open-source simulator to generate training and test data for drones and cars: Airsim
Training in a simulator and applying it to real world is a challenging task where you need to overcome the domain gaps.
You are open on the sensors used within your MobiBox, but please consider that the sensor set for car and drone should be the same. Please also note that you need to generate training data somehow. Thus, it is recommended to use a single monochromic or a stereo camera. Most simulators support this sensor setup and are able to provide annotation for bounding boxes and semantic segmentation for it automatically.