LLA-MPPI: Rapidly Adaptive Whole-body Control of Legged Robots with GPU-Accelerated Parallel Simulations

Key points

• Real-time whole-body controllers for legged robots typically plan through a fixed nominal model and degrade when the deployed dynamics change.1
• The method converts whole-body adaptation into selection over a bank of GPU-batched contact simulators with different physical or structural parameters.1
• A whole-body MPPI planner optimizes controls through the selected model.1
• Hardware validation on a Unitree Go2 shows the robot walking under a payload added mid-run, walking after one leg is disabled, and pushing a box to its goal while increasing its mass on the fly.1

Sources (1)

• [1]LLA-MPPI: Rapidly Adaptive Whole-body Control of Legged Robots with GPU-Accelerated Parallel Simulations
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 05:30 PM

  We present Look-back and Look-ahead Adaptive Model Predictive Path Integral control (LLA-MPPI).

  Real-time whole-body controllers for legged robots typically plan through a fixed nominal model and degrade when the deployed dynamics change.

Extractive summary: sentences quoted from the sources.