Reaching fast and autonomous flight requires computationally efficient and robust algorithms. To this end, we train Guidance & Control Networks to approximate optimal control policies ranging from energy-optimal to time-optimal flight. We show that the policies become more difficult to learn the closer we get to the time-optimal 'bang-bang' control profile. We also assess the importance of knowing the maximum angular rotor velocity of the quadcopter and show that over- or underestimating this limit leads to less robust flight. We propose an algorithm to identify the current maximum angular rotor velocity onboard and a network that adapts its policy based on the identified limit. Finally, we extend previous work on Guidance & Control Networks by learning to take consecutive waypoints into account. We fly a 4x3m track in similar lap times as the differential-flatness-based minimum snap benchmark controller while benefiting from the flexibility that Guidance & Control Networks offer.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Guidance & Control Networks for Time-Optimal Quadcopter Flight


    Beteiligte:

    Erscheinungsdatum :

    2023


    Medientyp :

    Preprint


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    Flight test of Quadcopter Guidance with Vision-Based Reinforcement Learning

    Siddiquee, Manan / Junell, Jaime / Van Kampen, Erik-Jan | AIAA | 2019


    Flight Test of Quadcopter Guidance with Vision-Based Reinforcement Learning

    Siddiquee, Manan / Junell, Jaime / Kampen, Erik-Jan Van | TIBKAT | 2019


    Quadcopter Control with Smooth Flight Mode

    Mendoza-Mendoza, Julio Alberto / Gonzalez-Villela, Victor / Sepulveda-Cervantes, Gabriel et al. | Springer Verlag | 2020


    Mission-Based Quadcopter Flight Simulation

    Hemmati, Vahid / Behnia, Mohammad / Mohammadi, Ahmad et al. | IEEE | 2024


    Quadcopter Guidance Law Design using Deep Reinforcement Learning

    Aydinli, Sevket Utku / Kutay, Ali Turker | IEEE | 2023