Competition in January 2022
There were two tasks, both defined in a toy town. The positioning of the streets was changed on the day of the competition, so solutions should generalize to any possible streetmap.
The track was located on 2nd floor, end of the hallway pointing north towards Narva street.
Task1: Autonomous driving - obstacle avoidance. The car must complete a circuit in the town without hitting any of the objects nor the walls. The speed must be sufficient to complete the circuit in no longer than 2 times the time it takes a human to complete. Each team has multiple attempts. In this task, the town is turned into a circuit with no intersections or with all but one direction blocked by obstacles.
If multiple teams complete the challenge, the task complexity is increased by adding more obstacles more centrally on the road.
Task 2: Autonomous driving - route completion. The car must demonstrate the ability to travel a given set of routes from predestined points A to points B in the toy town. The routes are given to the teams at the beginning of the Task 2 competition and they can either encode it into their driving system or enter high-level commands [turn left, go straight, go right] on the fly when the car is driving. The team that completes most routes wins. Hitting the walls or obstacles constitutes a crash and the route is considered incomplete.
We used the DonkeyCar S1 platform. These cars come equipped with a frontal camera that is the main sensor here. They also have an IMU (inertial measurement unit), that can but doesn't need to be used. The car is 10:1 scale compared to a real car. Changing car hardware is not allowed in this competition, this is a software competition. The existing codebase is really nice and lets you get to collecting data in 10 minutes and training a neural network model in less than an hour.
For some teams, participation in the competition yielded 6EAP if you register to Autonomous Vehicles Project LTAT.06.012, for some this was a Machine Learning course project, for some a warm-up for a thesis using the toy cars.
Example AI driving: below is a link to a video of a Neural Network optimized to imitate human driving. The network has been shown 20 laps of clean driving in the given circuit.
This is AI driving, you model should behave like this, but also deal well with intersections and obstacles.
The detailed rules of the self-driving competition can be found here: RULES
Teams: (can have up to 5 members + supervisor)
Team1: Thamara L., Maximilian N., Shumpei M + Supervisor: Leo Schoberwalter
Team2: Rustam A., Aral a. Gianluca R.
Team3: [dropped out]
Team4: Risto K.
Team5: Kristjan R.
Team6: Mike C.
Team7: Hans Kristjan V, Karl-Ingmar A, Hendrik Parik
Team8: [dropped out]
Team9: [dropped out]
TEAMS CAN MERGE AT ANY POINT IN TIME!