Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lander.AI: Adaptive Landing Behavior Agent for Expertise in 3D Dynamic Platform Landings

Robinroy Peter, Lavanya Ratnabala, Demetros Aschu, Aleksey Fedoseev, Dzmitry Tsetserukou

TL;DR

The paper tackles autonomous drone landing on moving platforms under wind disturbances by introducing Lander.AI, a TD3-based DRL agent trained in a gym-pybullet-drones environment with domain randomization and a potential-field reward. The method integrates a carefully designed observation space, a three-action PID-like control output, and a reward structure combining attractive and potential-field terms, validated through extensive simulation and real-world tests on Crazyflie 2.1 drones using Vicon indoor localization. Compared to a Crazyflie EKF-PID baseline, Lander.AI achieves higher landing precision and robustness across static and dynamic scenarios, including complex trajectories, demonstrating strong transfer from simulation to real hardware. These results indicate Lander.AI’s practical potential for safe, autonomous landings in inspection, delivery, and emergency response applications, while underscoring DRL’s capacity to address intricate aerodynamic challenges in dynamic environments.

Abstract

Mastering autonomous drone landing on dynamic platforms presents formidable challenges due to unpredictable velocities and external disturbances caused by the wind, ground effect, turbines or propellers of the docking platform. This study introduces an advanced Deep Reinforcement Learning (DRL) agent, Lander:AI, designed to navigate and land on platforms in the presence of windy conditions, thereby enhancing drone autonomy and safety. Lander:AI is rigorously trained within the gym-pybullet-drone simulation, an environment that mirrors real-world complexities, including wind turbulence, to ensure the agent's robustness and adaptability. The agent's capabilities were empirically validated with Crazyflie 2.1 drones across various test scenarios, encompassing both simulated environments and real-world conditions. The experimental results showcased Lander:AI's high-precision landing and its ability to adapt to moving platforms, even under wind-induced disturbances. Furthermore, the system performance was benchmarked against a baseline PID controller augmented with an Extended Kalman Filter, illustrating significant improvements in landing precision and error recovery. Lander:AI leverages bio-inspired learning to adapt to external forces like birds, enhancing drone adaptability without knowing force magnitudes.This research not only advances drone landing technologies, essential for inspection and emergency applications, but also highlights the potential of DRL in addressing intricate aerodynamic challenges.

Lander.AI: Adaptive Landing Behavior Agent for Expertise in 3D Dynamic Platform Landings

TL;DR

The paper tackles autonomous drone landing on moving platforms under wind disturbances by introducing Lander.AI, a TD3-based DRL agent trained in a gym-pybullet-drones environment with domain randomization and a potential-field reward. The method integrates a carefully designed observation space, a three-action PID-like control output, and a reward structure combining attractive and potential-field terms, validated through extensive simulation and real-world tests on Crazyflie 2.1 drones using Vicon indoor localization. Compared to a Crazyflie EKF-PID baseline, Lander.AI achieves higher landing precision and robustness across static and dynamic scenarios, including complex trajectories, demonstrating strong transfer from simulation to real hardware. These results indicate Lander.AI’s practical potential for safe, autonomous landings in inspection, delivery, and emergency response applications, while underscoring DRL’s capacity to address intricate aerodynamic challenges in dynamic environments.

Abstract

Mastering autonomous drone landing on dynamic platforms presents formidable challenges due to unpredictable velocities and external disturbances caused by the wind, ground effect, turbines or propellers of the docking platform. This study introduces an advanced Deep Reinforcement Learning (DRL) agent, Lander:AI, designed to navigate and land on platforms in the presence of windy conditions, thereby enhancing drone autonomy and safety. Lander:AI is rigorously trained within the gym-pybullet-drone simulation, an environment that mirrors real-world complexities, including wind turbulence, to ensure the agent's robustness and adaptability. The agent's capabilities were empirically validated with Crazyflie 2.1 drones across various test scenarios, encompassing both simulated environments and real-world conditions. The experimental results showcased Lander:AI's high-precision landing and its ability to adapt to moving platforms, even under wind-induced disturbances. Furthermore, the system performance was benchmarked against a baseline PID controller augmented with an Extended Kalman Filter, illustrating significant improvements in landing precision and error recovery. Lander:AI leverages bio-inspired learning to adapt to external forces like birds, enhancing drone adaptability without knowing force magnitudes.This research not only advances drone landing technologies, essential for inspection and emergency applications, but also highlights the potential of DRL in addressing intricate aerodynamic challenges.
Paper Structure (17 sections, 9 equations, 11 figures)

This paper contains 17 sections, 9 equations, 11 figures.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Simulation Environment Setup
  5. Deep Reinforcement Learning Framework
  6. Lander.AI Agent Architecture
  7. Reward Function
  8. Training Protocol
  9. Real-World Validation Setup
  10. Indoor Localization System (Vicon)
  11. Crazyflie Drones
  12. Baseline Comparison
  13. Crazyflie On-board PID Controller with Extended Kalman Filter
  14. Experiments
  15. Experimental Design
  16. ...and 2 more sections

Figures (11)

  • Figure 1: Composite image of the Lander.AI drone agent landing on UR10 robotic arm equipped with landing pad and air disturbance impeller.
  • Figure 2: Gym-pybullet Simulation Environment Setup.
  • Figure 3: Architecture of Deep Reinforcement Learning Model Illustrating Inputs, Hidden Layers, and Action Mechanisms.
  • Figure 4: Origin View of Lander.AI's Reward Function Emphasizing Safety and Behavior.
  • Figure 5: Lateral View of Lander.AI's Reward Function Showcasing Safety and Behavioral Rewards.
  • ...and 6 more figures