Diffusion-based Planning with Learned Viability Filters

2025-02-26

AI Summary

Diffusion models for embodied agent motion planning often generate sampled trajectories violating implicit hard constraints—e.g., fall prevention and collision avoidance—due to the lack of explicit feasibility modeling during sampling. Method: We propose a learnable Feasibility Filter: a neural network that explicitly models and predicts the success probability of any diffusion-sampled trajectory over future timesteps. Multiple filters can be composed to jointly enforce diverse implicit constraints. Our approach integrates diffusion priors, online resampling, and compositional constraint discrimination—without modifying the diffusion process or training objective. Results: Evaluated on high-difficulty humanoid locomotion tasks—including box climbing, wall crossing, and complex obstacle avoidance—the method achieves robust, real-time planning. It significantly outperforms guidance-based diffusion approaches in inference speed and, for the first time, enables efficient, decoupled modeling and enforcement of implicit physical feasibility constraints.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Diffusion models can be used as a motion planner by sampling from a distribution of possible futures. However, the samples may not satisfy hard constraints that exist only implicitly in the training data, e.g., avoiding falls or not colliding with a wall. We propose learned viability filters that efficiently predict the future success of any given plan, i.e., diffusion sample, and thereby enforce an implicit future-success constraint. Multiple viability filters can also be composed together. We demonstrate the approach on detailed footstep planning for challenging 3D human locomotion tasks, showing the effectiveness of viability filters in performing online planning and control for box-climbing, step-over walls, and obstacle avoidance. We further show that using viability filters is significantly faster than guidance-based diffusion prediction.

Problem

• Ensuring future success in diffusion-based motion planning
• Composing multiple viability filters for complex tasks
• Improving speed over guidance-based diffusion prediction

Innovation

• Diffusion models for motion planning
• Learned viability filters enforce constraints
• Faster than guidance-based diffusion prediction