Full-Body Torque-Level Non-linear Model Predictive Control for Aerial Manipulation

Josep Martí-Saumell,Joan Solà,Angel Santamaria-Navarro,Juan Andrade-Cetto

2023-07-27

Abstract:Non-linear model predictive control (nMPC) is a powerful approach to control complex robots (such as humanoids, quadrupeds, or unmanned aerial manipulators (UAMs)) as it brings important advantages over other existing techniques. The full-body dynamics, along with the prediction capability of the optimal control problem (OCP) solved at the core of the controller, allows to actuate the robot in line with its dynamics. This fact enhances the robot capabilities and allows, e.g., to perform intricate maneuvers at high dynamics while optimizing the amount of energy used. Despite the many similarities between humanoids or quadrupeds and UAMs, full-body torque-level nMPC has rarely been applied to UAMs. This paper provides a thorough description of how to use such techniques in the field of aerial manipulation. We give a detailed explanation of the different parts involved in the OCP, from the UAM dynamical model to the residuals in the cost function. We develop and compare three different nMPC controllers: Weighted MPC, Rail MPC, and Carrot MPC, which differ on the structure of their OCPs and on how these are updated at every time step. To validate the proposed framework, we present a wide variety of simulated case studies. First, we evaluate the trajectory generation problem, i.e., optimal control problems solved offline, involving different kinds of motions (e.g., aggressive maneuvers or contact locomotion) for different types of UAMs. Then, we assess the performance of the three nMPC controllers, i.e., closed-loop controllers solved online, through a variety of realistic simulations. For the benefit of the community, we have made available the source code related to this work.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the application of full - body torque - level non - linear model predictive control (nMPC) in Unmanned Aerial Manipulators (UAMs). Specifically, the goals of the paper are:

1. **Describe how to use full - body torque - level nMPC technology in the field of aerial manipulation**: The paper explains in detail the different parts in the Optimal Control Problem (OCP), from the dynamic model of UAM to the residual terms in the cost function.

2. **Develop and compare three different nMPC controllers**: These three controllers are Weighted MPC, Rail MPC and Carrot MPC, which differ in the structure of OCP and the way of updating at each time step.

3. **Verify the proposed framework through extensive simulation cases**: The paper first evaluates the trajectory generation problem, that is, the optimal control problem solved offline, involving different motions of different types of UAMs (such as aggressive maneuvers or contact motions). Then, through a series of realistic simulations, the performance of the three nMPC controllers (closed - loop controllers solved online) is evaluated.

4. **Provide open - source code**: For the benefit of the community, the author has made the source code related to this work public.

### The main contributions of the paper include:
1. **Detailed theoretical and technical descriptions**: Including the theory of fast OCP solvers, the definition of cost functions and the establishment of dynamic models.
2. **Propose a new OCP strategy - Carrot - nMPC**: This strategy fully respects the optimality criteria used to define tasks and is compared with existing methods from other robotic fields.
3. **Thorough verification through realistic simulations**: Covering multiple platforms, arms and tasks, especially those involving high - dynamics, providing evidence of computing time and hardware requirements, proving that these methods can be implemented in practical applications in the near future.
4. **Open - source software tools**: All software tools used in the experiments have been made public for the community to use.

### Background and motivation:
- **Similarities between UAMs and humanoid robots**: UAMs and humanoid robots are very similar in many aspects. For example, both have arms connected to movable bodies, can perform various operations (such as pushing, grasping, manipulating, etc.), and both need to counteract gravity.
- **Control challenges**: Although UAMs and humanoid robots face similar challenges in control, currently, full - body torque - level nMPC is rarely applied in UAMs.
- **Differences in control methods**: Existing UAM control methods are usually divided into decentralized control and centralized control. Decentralized control controls the platform and the robotic arm separately, while centralized control performs full - body control on the entire UAM.

### Structure of the paper:
- **Part II**: Outlines the current status of UAM control.
- **Part III**: Introduces the main methods for solving OCP, with particular attention to fast solvers.
- **Part IIII**: Explains how to construct the dynamic model of UAM, including the case of contact with the environment.
- **Part V**: Describes how to construct the cost function, its structure and different residual functions.
- **Part VI**: Introduces different strategies from the OCP for offline trajectory generation to the OCP for online robot control.
- **Part VII**: Verifies the proposed method through experiments, including UAM tasks in complex situations and performance comparisons of different nMPC techniques.
- **Part VIII**: Conclusions.

Through these contributions, the paper aims to promote the control technology of UAMs towards more advanced and complex tasks, especially in high - dynamic and environmental - interaction situations.