DTA

Digital Theses Archive

 

Tesi etd-11172021-151346

Type of thesis
Corso Ordinario Secondo Livello
Author
FERRINI, LORENZO
URN
etd-11172021-151346
Title
Dynamic obstacle avoidance for quadrotors using imitation learning
Structure
Cl. Sc. Sperimentali - Ingegneria
Course
INGEGNERIA - INGEGNERIA
Committee
relatore Prof. CASTOLDI, PIERO
Presidente Prof. FRISOLI, ANTONIO
Membro Prof. ODDO, CALOGERO MARIA
Membro Prof. CUCINOTTA, TOMMASO
Membro Prof. BERGAMASCO, MASSIMO
Membro Prof. CIARAMELLA, ERNESTO
Membro Prof. RICOTTI, LEONARDO
Keywords
  • control
  • depth-maps
  • Drone
  • Dynamic obstacle avoidance
  • imitation learning
  • perception
  • Reinforcement Learning
Exam session start date
16/12/2021;
Availability
parziale
Abstract
With this work I want to investigate the possibility of using imitation learning techniques to develop efficient obstacle avoidance algorithms. The core idea is to train an algorithm that uses previledged information using RL in simulation and then copy this behaviour with imitation learning to distill a policy that uses only the partial information disposable to the real drone.
Files