Reference-Free Sampling-Based Model Predictive Control
Fabian Schramm, Pierre Fabre, Nicolas Perrin-Gilbert, Justin Carpentier
AI summary
Problem
Existing sampling-based MPC methods typically rely on handcrafted gait references, predefined contact sequences, or GPU acceleration, limiting their adaptability and computational efficiency for online robot control.
Approach
The method extends Model Predictive Path Integral (MPPI) control using a cubic Hermite spline parameterization that jointly optimizes position and velocity control points, paired with a diffusion-inspired noise annealing schedule and reference-free cost functions.
Key results
- Discovers diverse emergent gaits without gait priors or offline training
- Achieves real-time control on standard CPU hardware using only 30–70 sampled trajectories
- Demonstrates robust sim-to-real transfer on a Go2 quadruped and complex humanoid behaviors in simulation
- Eliminates GPU acceleration requirements typical of state-of-the-art MPPI methods
Why it matters
Enables adaptable, computationally efficient whole-body control for legged and humanoid robots without relying on rigid gait templates or specialized hardware.
Abstract
We present a sampling-based model predictive control (MPC) framework that enables emergent locomotion without relying on handcrafted gait patterns or predefined contact sequences. Our method discovers diverse motion pat- terns, ranging from trotting to galloping, robust standing policies, jumping, and handstand balancing, purely through the optimization of high-level objectives. Building on model predictive path integral (MPPI), we propose a cubic Hermite spline parameterization that operates on position and velocity control points. Our approach enables contact-making and contact-breaking strategies that adapt automatically to task requirements, requiring only a limited number of sampled trajectories. This sample efficiency enables real-time control on standard CPU hardware, eliminating the GPU acceleration typically required by other state-of-the-art MPPI methods. We validate our approach on the Go2 quadrupedal robot, demonstrating a range of emergent gaits and basic jumping capabilities. In simulation, we further showcase more complex behaviors, such as backflips, dynamic handstand balancing and locomotion on a Humanoid, all without requiring reference tracking or offline pre-training.