[Paper Review] Towards a Unified Planner For Socially-Aware Navigation

Santosh Balajee Banisetty, David Feil-Seifer|arXiv (Cornell University)|Oct 1, 2018

Social Robot Interaction and HRI24 references4 citations

TL;DR

This paper proposes a Unified Socially-Aware Navigation (USAN) architecture that dynamically adapts robot navigation to varying social contexts using a context classifier, intent recognition, and a multi-objective local planner (PaCcET). By integrating these components into the ROS navigation stack, the system achieves socially appropriate trajectories—such as joining a queue or positioning in a group—outperforming traditional planners in both performance and social acceptability.

ABSTRACT

This paper presents the framework for a novel Unified Socially-Aware Navigation (USAN) architecture and explains its need in Socially Assistive Robotics (SAR) applications. Our approach emphasizes interpersonal distance and how spatial communication can be used to build a unified planner for a human-robot collaborative environment. Socially-Aware Navigation (SAN) is vital for helping humans to feel comfortable and safe around robots; HRI studies have shown the importance of SAN transcends safety and comfort. SAN plays a crucial role in perceived intelligence, sociability and social capacity of the robot, thereby increasing the acceptance of the robots in public places. Human environments are very dynamic and pose serious social challenges to robots intended for interactions with people. For the robots to cope with the changing dynamics of a situation, there is a need to infer intent and detect changes in the interaction context. SAN has gained immense interest in the social robotics community; to the best of our knowledge, however, there is no planner that can adapt to different interaction contexts spontaneously after autonomously sensing the context. Most of the recent efforts involve social path planning for a single context. In this work, we propose a novel approach for a unified architecture to SAN that can plan and execute trajectories for an autonomously sensed interaction context that are human-friendly. Our approach augments the navigation stack of the Robot Operating System (ROS) utilizing machine learning and optimization tools. We modified the ROS navigation stack using a machine learning-based context classifier and a PaCcET based local planner for us to achieve the goals of USAN. We discuss our preliminary results and concrete plans on putting the pieces together in achieving USAN.

Motivation & Objective

• Address the lack of adaptable social navigation in robots that currently operate only in single, predefined contexts.
• Improve human-robot interaction by ensuring robots respect personal space and social norms during navigation.
• Develop a modular, extensible architecture for socially-aware navigation that integrates perception, context understanding, and motion planning.
• Enable robots to autonomously detect interaction contexts and select appropriate navigation objectives accordingly.
• Enhance robot acceptance in public spaces by aligning navigation behavior with human social expectations and proxemic rules.

Proposed method

• Augment the ROS navigation stack with a machine learning-based context classifier to detect interaction scenarios such as 'meeting', 'walking together', or 'waiting in a queue'.
• Integrate a PaCcET-based local planner that optimizes for multiple social and performance objectives (e.g., distance, proximity, group alignment) in real time.
• Use a cardinal objective selector to dynamically choose relevant social goals based on the detected context, such as joining a line or positioning within a group.
• Apply multi-objective optimization to generate trajectories that balance social norms (e.g., proxemics) with traditional navigation metrics like time and distance.
• Design a modular system allowing plug-in of custom perception and planning modules for different robot platforms and environments.
• Validate the architecture through simulation and planned experiments on PR2 and Pioneer robots using both performance and social acceptability metrics.

Experimental results

Research questions

• RQ1 How can a robot autonomously detect and adapt to different social interaction contexts during navigation?
• RQ2 What combination of social and performance objectives is most effective for generating human-friendly trajectories in diverse social scenarios?
• RQ3 Can a unified planner outperform context-specific planners in terms of social acceptability and navigation efficiency?
• RQ4 How do human participants perceive and react to robot navigation behaviors that respect social norms compared to traditional, non-social planners?
• RQ5 What metrics best capture the social quality of robot navigation in human-robot interaction settings?

Key findings

• The proposed USAN architecture successfully integrates context classification, intent recognition, and multi-objective planning to generate context-aware trajectories.
• In a 'waiting in a queue' scenario, the PaCcET-based planner steered the robot to join the line, while the traditional planner moved it to the front—demonstrating socially appropriate behavior.
• The system enables robots to position themselves appropriately in group settings, such as forming an O-formation to interact with all members.
• Preliminary results show that the modular architecture supports dynamic objective selection based on context, improving social alignment.
• The team plans to validate the system through in-person and observer studies using validated scales for comfort, naturalness, and sociability.
• The approach shows potential to overcome the limitation of existing SAN systems that are restricted to single-context operation.