Composite Gaussian Processes Flows for Learning Discontinuous Multimodal Policies

TL;DR

This paper introduces CGP-Flows, a hybrid semi-parametric model combining OMGPs and CNFs, to effectively learn complex, multimodal, and discontinuous control policies for robotic tasks with improved efficiency and success rates.

Contribution

The paper presents a novel hybrid model, CGP-Flows, that combines OMGPs and CNFs to better capture complex robotic policies involving multimodality and discontinuities.

Findings

CGP-Flows outperform baseline methods in simulated robotic tasks.

CGP-Flows achieve higher success rates with statistical significance.

The model demonstrates effectiveness in real-world robotic experiments.

Abstract

Learning control policies for real-world robotic tasks often involve challenges such as multimodality, local discontinuities, and the need for computational efficiency. These challenges arise from the complexity of robotic environments, where multiple solutions may coexist. To address these issues, we propose Composite Gaussian Processes Flows (CGP-Flows), a novel semi-parametric model for robotic policy. CGP-Flows integrate Overlapping Mixtures of Gaussian Processes (OMGPs) with the Continuous Normalizing Flows (CNFs), enabling them to model complex policies addressing multimodality and local discontinuities. This hybrid approach retains the computational efficiency of OMGPs while incorporating the flexibility of CNFs. Experiments conducted in both simulated and real-world robotic tasks demonstrate that CGP-flows significantly improve performance in modeling control policies. In a simulation task, we confirmed that CGP-Flows had a higher success rate compared to the baseline method, and the success rate of GCP-Flow was significantly different from the success rate of other baselines in chi-square tests.