Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
153 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

An Adaptive PID Autotuner for Multicopters with Experimental Results (2109.12797v1)

Published 27 Sep 2021 in eess.SY, cs.AI, cs.RO, and cs.SY

Abstract: This paper develops an adaptive PID autotuner for multicopters, and presents simulation and experimental results. The autotuner consists of adaptive digital control laws based on retrospective cost adaptive control implemented in the PX4 flight stack. A learning trajectory is used to optimize the autopilot during a single flight. The autotuned autopilot is then compared with the default PX4 autopilot by flying a test trajectory constructed using the second-order Hilbert curve. In order to investigate the sensitivity of the autotuner to the quadcopter dynamics, the mass of the quadcopter is varied, and the performance of the autotuned and default autopilot is compared. It is observed that the autotuned autopilot outperforms the default autopilot.

Citations (16)

Summary

  • The paper proposes an adaptive PID autotuner for quadcopters that uses Retrospective Cost Adaptive Control (RCAC) to optimize controller gains iteratively without requiring precise dynamic models.
  • Experimental results show the adaptive autotuner significantly improved position tracking performance, achieving a 38.4% error reduction in simulation and 32.3% in physical flight tests compared to the default autopilot.
  • The adaptive system demonstrated robustness to varying conditions like payload changes, highlighting its potential for enhancing UAV reliability and performance in dynamic real-world applications.

Adaptive Digital PID Control of a Quadcopter: An Overview

This paper presents an investigation into the development and evaluation of an adaptive PID autotuner for multicopters, specifically focusing on quadcopters. The paper addresses the challenge of controlling the inherently nonlinear and unstable dynamics of quadcopters by proposing a methodology that utilizes retrospective cost adaptive control (RCAC) to optimize a digital PID controller. The proposed adaptive autotuner is incorporated into the PX4 flight stack and tested against the default PX4 autopilot using various flight scenarios.

Research Goal and Methodology

The primary goal of this research is to improve the performance of quadcopter control systems without relying on precise mathematical models of the quadcopter's dynamics. Traditional methods for multicopter control, such as feedback-linearization, backstepping, and several adaptive approaches, necessitate accurate dynamic models, which can be difficult to obtain and maintain due to varying operational conditions and external disturbances. This paper circumvents this requirement by employing an adaptive digital control strategy that iteratively optimizes the control gains via a learning flight trajectory.

The adaptive autotuner replaces fixed-gain PID controllers with adaptive counterparts, leveraging the RCAC algorithm to adjust gains in real-time. These adjustments are based on minimizing a retrospective cost function, which measures the deviation between desired and actual system performance. The algorithm iteratively updates control parameters during flight, enabling the system to adapt to dynamic changes in conditions and properties, such as payload variations.

Key Findings and Results

The paper showcases the efficacy of the adaptive autotuner through both simulation and physical flight tests. Results indicate that the autotuned autopilot consistently outperforms the default PX4 autopilot, demonstrating improved position tracking across various flight scenarios. Numerical results from the simulation showed a 38.4% reduction in tracking error, while physical flight tests conducted at the University of Michigan's M-Air facility reported a 32.3% improvement.

The adaptive system's robustness to changes in quadcopter mass, simulating conditions such as payload shifts, is also tested. This adaptability underscores the potential of the proposed approach for real-world applications, where such variations are commonplace.

Implications and Future Directions

The paper introduces an effective solution for the adaptive control of quadcopters, eliminating the dependency on precise modeling, thereby simplifying the deployment process for varying configurations and payloads. It emphasizes the potential for adaptive control algorithms, like RCAC, to enhance UAV stability and agility in dynamic environments.

Future research efforts could focus on extending this adaptive approach to more complex UAV models, involving asymmetric configurations or malfunction scenarios where traditional PID controllers struggle. Additionally, exploring the integration of machine learning techniques to further optimize control strategies in real-time could yield potential advancements in autonomous UAV operations.

In conclusion, this paper provides a substantial contribution to the field of UAV control, presenting an adaptive control mechanism that enhances the reliability and performance of quadcopter autopilots, paving the way for broader applications and further technological advancements in unmanned aerial systems.

Youtube Logo Streamline Icon: https://streamlinehq.com