[Paper Review] Socially-Aware Navigation: A Non-linear Multi-Objective Optimization Approach

Santosh Balajee Banisetty, Scott Forer+3 more|arXiv (Cornell University)|Nov 10, 2019

Social Robot Interaction and HRI4 citations

TL;DR

This paper proposes a non-linear multi-objective optimization approach using Pareto Concavity Elimination Transformation (PaCcET) to enable socially-aware navigation in mobile robots. Integrated into the ROS navigation stack, the method optimizes for social norms like personal space and proxemics across diverse human-robot interaction scenarios, demonstrating successful navigation in hallways, galleries, queues, and group interactions with both simulated and real-world robots.

ABSTRACT

Mobile robots are increasingly populating homes, hospitals, shopping malls, factory floors, and other human environments. Human society has social norms that people mutually accept; obeying these norms is an essential signal that someone is participating socially with respect to the rest of the population. For robots to be socially compatible with humans, it is crucial for robots to obey these social norms. In prior work, we demonstrated a Socially-Aware Navigation (SAN) planner, based on Pareto Concavity Elimination Transformation (PaCcET), in a hallway scenario, optimizing two objectives so that the robot does not invade the personal space of people. This paper extends our PaCcET based SAN planner to multiple scenarios with more than two objectives. We modified the Robot Operating System's (ROS) navigation stack to include PaCcET in the local planning task. We show that our approach can accommodate multiple Human-Robot Interaction (HRI) scenarios. Using the proposed approach, we achieved successful HRI in multiple scenarios like hallway interactions, an art gallery, waiting in a queue, and interacting with a group. We implemented our method on a simulated PR2 robot in a 2D simulator (Stage) and a pioneer-3DX mobile robot in the real-world to validate all the scenarios. A comprehensive set of experiments shows that our approach can handle multiple interaction scenarios on both holonomic and non-holonomic robots; hence, it can be a viable option for a Unified Socially-Aware Navigation (USAN).

Motivation & Objective

• To address the challenge of robots violating social norms during navigation in human-populated environments, such as invading personal space or treating people like obstacles.
• To extend prior work on socially-aware navigation (SAN) from two-objective to multi-objective optimization, enabling broader applicability across complex human-robot interaction (HRI) scenarios.
• To integrate a non-linear multi-objective optimization framework (PaCcET) into the Robot Operating System (ROS) navigation stack at the local planning level for real-time, socially compliant trajectory generation.
• To validate the approach across diverse, realistic HRI scenarios involving multiple humans, dynamic group formations, and environmental constraints such as activity zones and social goals.
• To lay the foundation for a Unified Socially-Aware Navigation (USAN) system that generalizes across interaction types while maintaining social acceptability and safety.

Proposed method

• The authors employ the Pareto Concavity Elimination Transformation (PaCcET) to convert a non-convex Pareto front into a convex one, enabling efficient solution of non-linear multi-objective optimization problems in local robot planning.
• The method integrates PaCcET into the local planner of the ROS navigation stack, allowing the robot to simultaneously optimize for performance (e.g., distance, time) and social objectives (e.g., personal space, group proximity, social goal adherence).
• Key objectives include minimizing distance to the goal, maximizing distance from humans (to respect personal space), and maintaining appropriate interpersonal distances in group interactions.
• The approach accounts for dynamic changes in human positions, group formations (e.g., O-formation), and environmental context (e.g., hallway walls, activity zones) during real-time trajectory planning.
• The system was implemented and tested on a simulated PR2 robot in the Stage 2D simulator and on a real Pioneer P3DX mobile robot in physical environments.
• Social metrics such as comfort, sociability, naturalness, predictability, and legibility were used to evaluate the social impact of the robot’s behavior.

Experimental results

Research questions

• RQ1 Can a non-linear multi-objective optimization framework like PaCcET effectively handle multiple social and performance objectives in robot navigation?
• RQ2 How well does the PaCcET-based planner generalize across diverse HRI scenarios, including single-person interactions, queues, group interactions, and complex environments like art galleries?
• RQ3 To what extent does the integration of PaCcET into the ROS navigation stack enable socially compliant navigation on both holonomic and non-holonomic robots?
• RQ4 How do changes in human positions, group formations, and environmental constraints affect the planner’s ability to maintain social norms?
• RQ5 Can the proposed approach be extended to support high-level decision-making using knowledge graphs for autonomous scenario understanding?

Key findings

• The PaCcET-based SAN planner successfully generated trajectories that respected personal space and social norms in all tested scenarios, including hallways, art galleries, queues, and group interactions.
• The system demonstrated robustness to dynamic changes, such as shifting human positions and varying group formations (e.g., O-formation), maintaining socially compliant behavior throughout.
• The approach was validated on both a simulated PR2 robot in Stage and a real Pioneer P3DX robot, confirming its feasibility in real-world deployment.
• The planner outperformed the traditional ROS local planner, which treated humans and obstacles identically, resulting in unsafe proximity to people.
• The method enabled consistent social behavior across sub-scenarios, such as stationary humans, passing, and group avoidance, indicating strong generalization capability.
• The study paves the way for a Unified Socially-Aware Navigation (USAN) system capable of handling diverse, complex HRI contexts with high social acceptability.