Sign in Subscribe
Machine learning

Adaptive Control System for Autonomous Drones

Strategic Insights

— 1 min read

Researchers at MIT have developed a machine learning-based adaptive control algorithm for autonomous drones to minimize deviation from their intended trajectory in the face of unpredictable forces like gusty winds. The new technique does not require prior knowledge of the structure of disturbances and can learn all it needs to know from a small amount of observational data collected during flight.

Key Takeaways:

  • The adaptive control system uses a neural network model to approximate disturbances and automatically determines the most suitable mirror-descent function to reduce tracking error.
  • The method achieves 50% less trajectory tracking error than baseline methods in simulations and performs better with new wind speeds it didn't see during training.
  • The system can adapt to different types of disturbances and can automatically make choices that will be best for quick adaptation through meta-learning.
  • The researchers use a technique called meta-learning to train the system to adapt to different types of disturbances and objectives.
  • The adaptive control system can continuously recalculate, in real-time, how the drone should produce thrust to keep it as close as possible to its target trajectory while accommodating uncertain disturbances.
  • The system outperformed baseline approaches in simulations and real-world experiments, especially in challenging environments with stronger wind speeds.

Statistics:

  • The adaptive control system achieved 50% less trajectory tracking error than baseline methods in simulations.
  • The system can adapt to different types of disturbances and can automatically make choices that will be best for quick adaptation.
  • The researchers used 15 minutes of flight time to collect observational data for the neural network model.
  • The system can handle wind speeds that are stronger than those seen during training with success.
  • The margin by which the method outperformed the baselines grew as the wind speeds intensified.

Sources:

  • NewsRx LLC, "Adaptive Control System for Autonomous Drones," June 29, 2025, VerticalNews.
  • Tang, S., Sun, H., et al., "Learning for Dynamics and Control Conference Presentation," Learning for Dynamics and Control Conference.

Read more