Neural Lander: Stable Drone Landing Control using Learned Dynamics

Abstract

Precise near-ground trajectory control is difficult for multi-rotor drones, due to the complex aerodynamic effects caused by interactions between multi-rotor airflow and the environment. Conventional control methods often fail to properly account for these complex effects and fall short in accomplishing smooth landing. In this paper, we present a novel deep-learning-based robust nonlinear controller (Neural Lander) that improves control performance of a quadrotor during landing. Our approach combines a nominal dynamics model with a Deep Neural Network (DNN) that learns high-order interactions. We apply spectral normalization (SN) to constrain the Lipschitz constant of the DNN. Leveraging this Lipschitz property, we design a nonlinear feedback linearization controller using the learned model and prove system stability with disturbance rejection. To the best of our knowledge, this is the first DNN-based nonlinear feedback controller with stability guarantees that can utilize arbitrarily large neural nets. Experimental results demonstrate that the proposed controller significantly outperforms a Baseline Nonlinear Tracking Controller in both landing and cross-table trajectory tracking cases. We also empirically show that the DNN generalizes well to unseen data outside the training domain.

7 pages, 5 figures, https://youtu.be/FLLsG0S78ik

References in corpus (8)

• Spectral Normalization for Generative Adversarial Networks
• Understanding deep learning requires rethinking generalization
• Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks
• Safe Model-based Reinforcement Learning with Stability Guarantees
• Exploring Generalization in Deep Learning
• A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks
• Safe Controller Optimization for Quadrotors with Gaussian Processes
• Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Data

Cited by in corpus (48)

• Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds
• Autonomous Drone Racing: A Survey
• NeuroBEM: Hybrid Aerodynamic Quadrotor Model
• Learning Efficient Navigation in Vortical Flow Fields
• Learning for Safety-Critical Control with Control Barrier Functions
• Bayesian Learning-Based Adaptive Control for Safety Critical Systems
• Physics-Inspired Temporal Learning of Quadrotor Dynamics for Accurate Model Predictive Trajectory Tracking
• Neural Contraction Metrics for Robust Estimation and Control: A Convex Optimization Approach
• Episodic Learning with Control Lyapunov Functions for Uncertain Robotic Systems
• Learning-based Predictive Path Following Control for Nonlinear Systems Under Uncertain Disturbances
• Inclined Quadrotor Landing using Deep Reinforcement Learning
• Learning-based Robust Motion Planning with Guaranteed Stability: A Contraction Theory Approach
• Blending Diverse Physical Priors with Neural Networks
• Neural Stochastic Contraction Metrics for Learning-based Control and Estimation
• Online Optimization with Memory and Competitive Control
• Learning to Fly via Deep Model-Based Reinforcement Learning
• Learning to Assist Drone Landings
• Meta-Adaptive Nonlinear Control: Theory and Algorithms
• Reinforcement Learning for Shared Autonomy Drone Landings
• ShieldNN: A Provably Safe NN Filter for Unsafe NN Controllers
• Safe Learning-based Gradient-free Model Predictive Control Based on Cross-entropy Method
• Differentiable Moving Horizon Estimation for Robust Flight Control
• SO(2)-Equivariant Downwash Models for Close Proximity Flight
• The Power of Predictions in Online Control
• Meta-Learning-Based Robust Adaptive Flight Control Under Uncertain Wind Conditions
• A Control Lyapunov Perspective on Episodic Learning via Projection to State Stability
• Robust Regression for Safe Exploration in Control
• Deep Learning-Based Spatiotemporal Multi-Event Reconstruction for Delay Line Detectors
• Deep KKL: Data-driven Output Prediction for Non-Linear Systems
• An Analysis of the Expressiveness of Deep Neural Network Architectures Based on Their Lipschitz Constants
• Deep Shape from Polarization
• Neural-Swarm2: Planning and Control of Heterogeneous Multirotor Swarms using Learned Interactions
• Neural-Swarm: Decentralized Close-Proximity Multirotor Control Using Learned Interactions
• Vid2Param: Modelling of Dynamics Parameters from Video
• Visual Physics: Discovering Physical Laws from Videos
• Provably Correct Training of Neural Network Controllers Using Reachability Analysis
• Hybrid aerial ground locomotion with a single passive wheel
• Guaranteed Trajectory Tracking under Learned Dynamics with Contraction Metrics and Disturbance Estimation
• Adaptive Neural Trajectory Tracking Control for Flexible-Joint Robots with Online Learning
• Coordinated Control of Deformation and Flight for Morphing Aircraft via Meta-Learning and Coupled State-Dependent Riccati Equations
• Disentangling Observed Causal Effects from Latent Confounders using Method of Moments
• Quadrotor Trajectory Tracking with Learned Dynamics: Joint Koopman-based Learning of System Models and Function Dictionaries
• Enhancement for Robustness of Koopman Operator-based Data-driven Mobile Robotic Systems
• A Data-driven Hierarchical Control Structure for Systems with Uncertainty
• Low-Latency Event-Based Velocimetry for Quadrotor Control in a Narrow Pipe
• Modeling Electrical Motor Dynamics using Encoder-Decoder with Recurrent Skip Connection
• Safe Learning-based Tracking Control for Quadrotors under Wind Disturbances
• Physics-infused Learning for Aerial Manipulator in Winds and Near-Wall Environments