Tailoring Solution Accuracy for Fast Whole-body Model Predictive Control of Legged Robots

Charles Khazoom,Seungwoo Hong,Matthew Chignoli,Elijah Stanger-Jones,Sangbae Kim

DOI: https://doi.org/10.1109/LRA.2024.3455907

2024-09-12

Abstract:Thanks to recent advancements in accelerating non-linear model predictive control (NMPC), it is now feasible to deploy whole-body NMPC at real-time rates for humanoid robots. However, enforcing inequality constraints in real time for such high-dimensional systems remains challenging due to the need for additional iterations. This paper presents an implementation of whole-body NMPC for legged robots that provides low-accuracy solutions to NMPC with general equality and inequality constraints. Instead of aiming for highly accurate optimal solutions, we leverage the alternating direction method of multipliers to rapidly provide low-accuracy solutions to quadratic programming subproblems. Our extensive simulation results indicate that real robots often cannot benefit from highly accurate solutions due to dynamics discretization errors, inertial modeling errors and delays. We incorporate control barrier functions (CBFs) at the initial timestep of the NMPC for the self-collision constraints, resulting in up to a 26-fold reduction in the number of self-collisions without adding computational burden. The controller is reliably deployed on hardware at 90 Hz for a problem involving 32 timesteps, 2004 variables, and 3768 constraints. The NMPC delivers sufficiently accurate solutions, enabling the MIT Humanoid to plan complex crossed-leg and arm motions that enhance stability when walking and recovering from significant disturbances.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to effectively handle inequality constraints and provide sufficiently accurate solutions when implementing whole - body nonlinear model predictive control (Whole - body NMPC) for legged robots (such as humanoid robots) under real - time conditions. Specifically, the paper focuses on the following aspects:

1. **Improving real - time performance**: Traditional NMPC methods are difficult to achieve real - time performance in high - dimensional systems, especially when inequality constraints need to be satisfied. The paper proposes a method based on the alternating direction multiplier method (ADMM) to quickly solve low - precision quadratic programming sub - problems, thereby improving real - time performance.

2. **Reducing self - collision**: The paper introduces control barrier functions (CBFs) to enhance the ability to avoid self - collision. By applying CBFs in the initial time step of NMPC, the occurrence of self - collision can be significantly reduced without increasing the computational burden.

3. **Dealing with modeling errors and delays**: Through a large number of simulation experiments, the paper evaluates the closed - loop performance of solutions with different precisions in the presence of inertial modeling errors, dynamic discretization errors, and computational delays. The results show that the actual robot system may not always benefit from high - precision solutions because factors such as modeling errors and delays will limit the effectiveness of high - precision solutions.

4. **Verifying hardware performance**: The paper has carried out extensive hardware experiments on the MIT humanoid robot to verify the effectiveness of the proposed NMPC method. The experimental results show that this method can run reliably at a frequency of 90 Hz, handle problems containing 32 time steps, 2004 variables, and 3768 constraints, and enable the robot to perform complex cross - legged and arm movements, improving walking stability and the ability to recover from external disturbances.

In summary, the main objective of this paper is to provide an effective method to handle inequality constraints in the whole - body NMPC of legged robots while ensuring real - time performance, and verify its feasibility and effectiveness in actual robot systems through experiments.