Model Predictive Path Integral Control using Covariance Variable Importance Sampling

Abstract

In this paper we develop a Model Predictive Path Integral (MPPI) control algorithm based on a generalized importance sampling scheme and perform parallel optimization via sampling using a Graphics Processing Unit (GPU). The proposed generalized importance sampling scheme allows for changes in the drift and diffusion terms of stochastic diffusion processes and plays a significant role in the performance of the model predictive control algorithm. We compare the proposed algorithm in simulation with a model predictive control version of differential dynamic programming.

8 pages

References in corpus (2)

• Path Integral Control and State Dependent Feedback
• GPU Based Path Integral Control with Learned Dynamics

Cited by in corpus (31)

• Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learning
• Deep Dynamics Models for Learning Dexterous Manipulation
• Deep Online Learning via Meta-Learning: Continual Adaptation for Model-Based RL
• Uncertainty Averse Pushing with Model Predictive Path Integral Control
• Keypoints into the Future: Self-Supervised Correspondence in Model-Based Reinforcement Learning
• 3D Neural Scene Representations for Visuomotor Control
• EDMP: Ensemble-of-costs-guided Diffusion for Motion Planning
• Model-Based Offline Planning
• Meta-Reinforcement Learning Robust to Distributional Shift via Model Identification and Experience Relabeling
• Combining Coarse and Fine Physics for Manipulation using Parallel-in-Time Integration
• Actionable Models: Unsupervised Offline Reinforcement Learning of Robotic Skills
• RoboNet: Large-Scale Multi-Robot Learning
• Multi-Agent Path Integral Control for Interaction-Aware Motion Planning in Urban Canals
• Off-Dynamics Reinforcement Learning: Training for Transfer with Domain Classifiers
• Reset-Free Lifelong Learning with Skill-Space Planning
• Topology-Guided Path Integral Approach for Stochastic Optimal Control in Cluttered Environment
• Amodal 3D Reconstruction for Robotic Manipulation via Stability and Connectivity
• Autoregressive Dynamics Models for Offline Policy Evaluation and Optimization
• Adaptive Online Planning for Continual Lifelong Learning
• Low Emission Building Control with Zero-Shot Reinforcement Learning
• O2A: One-shot Observational learning with Action vectors
• Terrain-Aware Kinodynamic Planning with Efficiently Adaptive State Lattices for Mobile Robot Navigation in Off-Road Environments
• Value Function Spaces: Skill-Centric State Abstractions for Long-Horizon Reasoning
• Solving Challenging Dexterous Manipulation Tasks With Trajectory Optimisation and Reinforcement Learning
• Meta-Model-Based Meta-Policy Optimization
• Hierarchically Integrated Models: Learning to Navigate from Heterogeneous Robots
• Maximum Entropy Model Rollouts: Fast Model Based Policy Optimization without Compounding Errors
• Information Theoretic Model Predictive Control on Jump Diffusion Processes
• Meta-Reinforcement Learning for Adaptive Motor Control in Changing Robot Dynamics and Environments
• NeuroSMPC: A Neural Network guided Sampling Based MPC for On-Road Autonomous Driving
• Discriminator Augmented Model-Based Reinforcement Learning