Abstract:Tiny aerial robots show promise for applications like environmental monitoring and search-and-rescue but face challenges in control due to their limited computing power and complex dynamics. Model Predictive Control (MPC) can achieve agile trajectory tracking and handle constraints. Although current learning-based MPC methods, such as Gaussian Process (GP) MPC, improve control performance by learning residual dynamics, they are computationally demanding, limiting their onboard application on tiny robots. This paper introduces Tiny Learning-Based Model Predictive Control (LB MPC), a novel framework for resource-constrained micro multirotor platforms. By exploiting multirotor dynamics' structure and developing an efficient solver, our approach enables high-rate control at 100 Hz on a Crazyflie 2.1 with a Teensy 4.0 microcontroller. We demonstrate a 23% average improvement in tracking performance over existing embedded MPC methods, achieving the first onboard implementation of learning-based MPC on a tiny multirotor (53 g).

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to achieve efficient and real - time model predictive control (MPC) on computationally resource - constrained miniature multi - rotor unmanned aerial vehicles (such as the Crazyflie 2.1 which weighs only 53 grams) to meet the control challenges of these small aircraft in complex dynamic environments. Specifically, the paper addresses the following issues: 1. **Limited computational resources**: Due to their small size and low cost, miniature multi - rotor unmanned aerial vehicles are usually equipped with computational hardware with limited performance. Although traditional learning - based MPC methods (such as Gaussian process MPC) can improve control performance, they require a large amount of computation and are difficult to implement real - time control on airborne devices. 2. **Complex dynamic characteristics**: The dynamic characteristics of miniature multi - rotor unmanned aerial vehicles are complex and nonlinear, and there are uncertainties (such as air resistance or other environmental interferences), which make it difficult for traditional linear MPC to achieve the ideal control effect. To solve these problems, the paper proposes a new framework named Tiny Learning - Based MPC. By combining differential flatness and an efficient solver design, it achieves high - frequency (100 Hz) predictive control on resource - constrained embedded hardware. The main contributions of this method include: - **Development of the Tiny Learning - Based MPC framework**: Combining learning - based MPC and differential flatness, and using a custom - made second - order cone programming (SOCP) solver to achieve efficient predictive control. - **First deployment of learning - based MPC on miniature multi - rotor unmanned aerial vehicles**: Using the Teensy 4.0 microcontroller on the Crazyflie 2.1 to achieve the sending of body - rate and total - thrust commands 100 times per second. - **Significant improvement in tracking performance**: Compared with existing embedded MPC methods, Tiny Learning - Based MPC shows better tracking performance in both simulations and experiments, with an average improvement ranging from 15% to 32%. Through these improvements, Tiny Learning - Based MPC can achieve efficient and accurate control of miniature multi - rotor unmanned aerial vehicles under very limited computational resources, and is suitable for application scenarios such as environmental monitoring, search and rescue.

Tiny Learning-Based MPC for Multirotors: Solver-Aware Learning for Efficient Embedded Predictive Control

TinyMPC: Model-Predictive Control on Resource-Constrained Microcontrollers

Real-time Neural-MPC: Deep Learning Model Predictive Control for Quadrotors and Agile Robotic Platforms

A Computationally Efficient Learning-Based Model Predictive Control for Multirotors under Aerodynamic Disturbances

Enhancing Sample Efficiency and Uncertainty Compensation in Learning-based Model Predictive Control for Aerial Robots

Flying Quadrotors in Tight Formations using Learning-based Model Predictive Control

Computationally Efficient Data-Driven MPC for Agile Quadrotor Flight

Embedded Hierarchical MPC for Autonomous Navigation

Towards Model Predictive Control for Acrobatic Quadrotor Flights

Efficient Learning-based Trajectory Tacker for Quadrotor at High-speed Flight

Performance, Precision, and Payloads: Adaptive Nonlinear MPC for Quadrotors

Proto-MPC: An Encoder-Prototype-Decoder Approach for Quadrotor Control in Challenging Winds

Deep Model Predictive Optimization

Low-Level Control of a Quadrotor With Deep Model-Based Reinforcement Learning

Variable-Frequency Model Learning and Predictive Control for Jumping Maneuvers on Legged Robots

Gaussian Process Model Predictive Control of An Unmanned Quadrotor

Model Predictive Control for Micro Aerial Vehicles: A Survey

Aggressive Trajectory Tracking for Nano Quadrotors Using Embedded Nonlinear Model Predictive Control

Towards Optimization of Energy Consumption of Tello Quad-Rotor with Mpc Model Implementation

SE(3) Koopman-MPC: Data-driven Learning and Control of Quadrotor UAVs

Closed-Loop Model Identification and MPC-based Navigation of Quadcopters: A Case Study of Parrot Bebop 2