Problem
Research questions and friction points this paper is trying to address.
social robot navigation
social conventions
behavioral adjustability
norm-grounded navigation
SoGuDiff: Socially Guided Diffusion for Steerable, Norm-Grounded Robot Navigation
🤖 AI Summary
This study addresses the absence of runtime interfaces for dynamically adjusting social norms in robot navigation by proposing a diffusion model-based social navigation framework. The method introduces a continuous style axis that enables independent or combined modulation of social behaviors, and employs a feasibility projection layer to decouple behavioral learning from motion constraints, thereby supporting real-time dynamic tuning during deployment. Experimental results demonstrate that the proposed framework strictly outperforms fixed-policy baselines and successfully reproduces diverse, distinct social styles in real-world scenarios. Ultimately, this work provides embodied agents with a flexible and controllable new paradigm for social interaction.
📝 Abstract
Beyond collision avoidance, socially competent robot navigation requires adherence to implicit social conventions that vary across contexts, cultures, and deployment requirements. Many conventional navigation policies learn a single normative behavior, either through reinforcement learning against a fixed reward function or imitation of human demonstrations, exposing no interface for adjusting that conduct at runtime. We present a diffusion-based navigation framework whose social behavior can be tuned at deployment: a desired style is specified, such as how closely the robot passes, which side it yields to, or how much it defers to groups, and the planner adapts accordingly. Continuous style axes can be followed independently or composed, spanning a behavioral space rather than discrete, primitive-based specifications. A feasibility projection layer separates learned social behavior from kinematic feasibility and collision avoidance. A single-axis sweep illustrates a tradeoff curve that strictly dominates the evaluated fixed-behavior baseline configurations, and stylistic differences are replicated in real-world demonstrations.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Diffusion Model
Socially Guided Navigation
Steerable Behavior
Feasibility Projection
Style Composition
🔎 Similar Papers
No similar papers found.
