Optimization free control and ground force estimation with momentum observer for a multimodal legged aerial robot

Kaushik Venkatesh Krishnamurthy,Chenghao Wang,Shreyansh Pitroda,Eric Sihite,Alireza Ramezani,Morteza Gharib

2024-11-18

Abstract:Legged-aerial multimodal robots can make the most of both legged and aerial systems. In this paper, we propose a control framework that bypasses heavy onboard computers by using an optimization-free Explicit Reference Governor that incorporates external thruster forces from an attitude controller. Ground reaction forces are maintained within friction cone constraints using costly optimization solvers, but the ERG framework filters applied velocity references that ensure no slippage at the foot end. We also propose a Conjugate momentum observer, that is widely used in Disturbance Observation to estimate ground reaction forces and compare its efficacy against a constrained model in estimating ground reaction forces in a reduced-order simulation of Husky.

Robotics,Systems and Control

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is how to implement an optimization - free control framework and ground reaction force estimation in multimodal legged - aerial robots. Specifically, the paper proposes a new method to simplify and improve the control performance of such complex robots, especially in the following aspects:

1. **Avoid using heavy - duty on - board computers**: Traditional methods rely on powerful on - board computers to perform complex optimization calculations, which increases the load of the robot. This paper proposes an Explicit Reference Governor (ERG), which can bypass these complex optimization steps, thereby reducing the computational burden.

2. **Ensure that the feet do not slip**: By introducing the ERG framework, the velocity reference applied to the robot can be filtered to ensure that no slipping occurs when the feet are in contact with the ground. This is achieved by keeping the ground reaction force within the friction cone constraint.

3. **Estimate the ground reaction force**: In order to better control the robot, it is necessary to accurately estimate the ground reaction force. This paper proposes a Conjugate momentum observer to estimate the ground reaction force and compares it with a constrained model to verify its effectiveness.

4. **Combine with attitude controllers**: By integrating the thrust generated by external thrusters, the attitude control of the robot is improved, enabling the robot to switch smoothly between different modes (such as walking and flying).

### Main contributions
- **Optimization - free control framework**: A control method without an optimization solver is proposed, reducing the need for high - performance computing resources.
- **Conjugate momentum observer**: A new momentum observer is developed to estimate the ground reaction force, improving control accuracy.
- **Experimental verification**: The effectiveness of the proposed control and estimation framework is verified through simulation, demonstrating its potential in handling complex terrains and multimodal movements.

### Formula summary
- Dynamics equation of the momentum observer:
\[
\hat{\dot{p}}=\hat{M}\hat{\ddot{q}}+\hat{\dot{M}}\hat{\dot{q}}=\hat{u}_g + B_t u_t-\hat{h}+\hat{\dot{M}}\hat{\dot{q}}
\]
where \(\hat{p}\) is the estimated value of the generalized momentum, \(\hat{M}\) is the estimated value of the mass - inertia matrix, \(\hat{\ddot{q}}\) and \(\hat{\dot{q}}\) are the estimated values of the generalized coordinate acceleration and velocity respectively, \(\hat{u}_g\) is the estimated value of the ground reaction force, \(B_t u_t\) is the generalized force of the thruster input, and \(\hat{h}\) is the estimated value of the Coriolis force and gravity.

- Ground reaction force constraint:
\[
h_r=\begin{bmatrix}
-\text{sgn}(u_{gi,x})&0&\mu_s\\
0&-\text{sgn}(u_{gi,x})\mu_s\\
0&0&1
\] u_{g,i}+\begin{bmatrix}
0\\
0\\
-u_{\text{min}, gi,z}
\end{bmatrix}\geq0
\]

Through these methods, the paper aims to provide an efficient and practical solution for the control and ground reaction force estimation of multimodal legged - aerial robots.