Composite Gaussian Processes Flows for Learning Discontinuous Multimodal Policies

TL;DR

CGP-Flows fuse sparse Overlapping Mixtures of Gaussian Processes with Conditional Normalizing Flows implemented via Neural-ODEs to learn robotic policies that are multimodal and locally discontinuous while remaining computationally efficient. The method uses a sparse OMGP base to capture multiple plausible action modes and a CNF-based transformation to produce the final policy distribution, with variational inference governing the joint model. Empirical results across ball-shooting, object-swiping, and real-world grasping demonstrate superior accuracy and favorable computational characteristics compared to OMGPs and NGGPs, including statistically significant improvements in policy success. This approach offers a practical, scalable path for learning complex robotic policies in real-world environments where multiple solutions and abrupt modality changes coexist.

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.

Figures (38)

• Figure 1:
• Figure 2:
• Figure 3:
• Figure 5: GP policy
• Figure 6: OMGP policy