# Deep reinforcement learning reveals fewer sensors are needed for autonomous gust alleviation

**Authors:** Kevin PT. Haughn, Christina Harvey, Daniel J. Inman  
**arXiv:** [2304.03133](https://arxiv.org/abs/2304.03133)  
**Date:** 2023-04-06  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.03133v1/

## Key points

- The use of UAVs in cities is becoming increasingly necessary due to their potential for various applications
- Complex urban landscape and unpredictable wind gusts pose a challenge to their safe and effective operation
- Current methods for mitigating gusts rely on traditional control surfaces and computationally intensive modeling, resulting in slower response times
- Researchers led by Kevin PT have developed a new approach using deep reinforcement learning to create an autonomous gust alleviation controller for a camber-morphing wing
- This method reduced the impact of gusts by 84%, using real-time pressure signals from on-board sensors
- Using only three pressure taps was just as effective as using six, indicating that this reduced-sensor fly-by-feel control could enable UAV missions in previously inaccessible locations.

## Summary

The use of uncrewed aerial vehicles (UAVs) in cities is becoming increasingly necessary due to their potential for various applications. However, the complex urban landscape and unpredictable wind gusts pose a challenge to their safe and effective operation. Current methods for mitigating gusts rely on traditional control surfaces and computationally intensive modeling, resulting in slower response times. Researchers led by Kevin PT have developed a new approach using deep reinforcement learning to create an autonomous gust alleviation controller for a camber-morphing wing. This method reduced the impact of gusts by 84%, using real-time pressure signals from on-board sensors. Remarkably, the study found that using only three pressure taps was just as effective as using six, indicating that this reduced-sensor fly-by-feel control could enable UAV missions in previously inaccessible locations.