Where is everyone going?
AI-powered traffic pattern predictions
Status: Assigned
Student: Camilio DelfGaauw
Problem Statement:
Self-driving cars relay heavily on local imaging and AI to determine outside obstacles; but if each vehicle would be aware of each other the overall traffic management would be greatly relaxed. Nonetheless, even knowing where everyone else keeps the problem of discovering each other trajectories and avoid collisions. Here comes AI, can we build a dynamic machine learning model able to predict trajectories?
Tasks:
The student will use the traffic simulator SUMO to generate realistic mobility scenarios and will build a Deep (Reinforcement) Learning model to avoid collisions.
Work:
10% Theory, 70% Simulations, 20%Writing.
Contact:
Alessandro Chiumento (a.chiumento@utwente.nl)