Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Online Context Learning for Socially Compliant Navigation

Iaroslav Okunevich, Alexandre Lombard, Tomas Krajnik, Yassine Ruichek, Zhi Yan

TL;DR

This work addresses the challenge of socially compliant robot navigation across changing human contexts by proposing a two-layer architecture that couples a DRL-based navigation foundation (SARL) with an online, tracklet-driven social module. The online social context learning mechanism updates only the social module in real time, using a GRU-MLP social value network and a labeling strategy based on tracklet dynamics to adapt to new environments without disrupting the core navigation policy. The approach achieves improvements over state-of-the-art methods in simulations and real-robot experiments, including enhanced social efficiency and robustness across diverse scenarios, and is complemented by open-source code and data. Collectively, the method enables fast online adaptation to unseen social contexts, supporting more robust and socially aware long-term robot deployment in human-rich spaces.

Abstract

Robot social navigation needs to adapt to different human factors and environmental contexts. However, since these factors and contexts are difficult to predict and cannot be exhaustively enumerated, traditional learning-based methods have difficulty in ensuring the social attributes of robots in long-term and cross-environment deployments. This letter introduces an online context learning method that aims to empower robots to adapt to new social environments online. The proposed method adopts a two-layer structure. The bottom layer is built using a deep reinforcement learning-based method to ensure the output of basic robot navigation commands. The upper layer is implemented using an online robot learning-based method to socialize the control commands suggested by the bottom layer. Experiments using a community-wide simulator show that our method outperforms the state-of-the-art ones. Experimental results in the most challenging scenarios show that our method improves the performance of the state-of-the-art by 8%. The source code of the proposed method, the data used, and the tools for the per-training step are publicly available at https://github.com/Nedzhaken/SOCSARL-OL.

Online Context Learning for Socially Compliant Navigation

TL;DR

This work addresses the challenge of socially compliant robot navigation across changing human contexts by proposing a two-layer architecture that couples a DRL-based navigation foundation (SARL) with an online, tracklet-driven social module. The online social context learning mechanism updates only the social module in real time, using a GRU-MLP social value network and a labeling strategy based on tracklet dynamics to adapt to new environments without disrupting the core navigation policy. The approach achieves improvements over state-of-the-art methods in simulations and real-robot experiments, including enhanced social efficiency and robustness across diverse scenarios, and is complemented by open-source code and data. Collectively, the method enables fast online adaptation to unseen social contexts, supporting more robust and socially aware long-term robot deployment in human-rich spaces.

Abstract

Robot social navigation needs to adapt to different human factors and environmental contexts. However, since these factors and contexts are difficult to predict and cannot be exhaustively enumerated, traditional learning-based methods have difficulty in ensuring the social attributes of robots in long-term and cross-environment deployments. This letter introduces an online context learning method that aims to empower robots to adapt to new social environments online. The proposed method adopts a two-layer structure. The bottom layer is built using a deep reinforcement learning-based method to ensure the output of basic robot navigation commands. The upper layer is implemented using an online robot learning-based method to socialize the control commands suggested by the bottom layer. Experiments using a community-wide simulator show that our method outperforms the state-of-the-art ones. Experimental results in the most challenging scenarios show that our method improves the performance of the state-of-the-art by 8%. The source code of the proposed method, the data used, and the tools for the per-training step are publicly available at https://github.com/Nedzhaken/SOCSARL-OL.
Paper Structure (15 sections, 12 equations, 4 figures, 1 table, 1 algorithm)

This paper contains 15 sections, 12 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Socially Compliant Navigation
  4. Online Context Learning
  5. Discussion
  6. Method
  7. Problem Formulation
  8. SARL
  9. Social Module
  10. Online Social Context Learning
  11. Evaluation
  12. Social Module Initialization
  13. Simulation Results
  14. Real Robot Experiments
  15. Conclusions

Figures (4)

  • Figure 1: Conceptual diagram of our proposed approach. The navigation module selects the robot's next action. The social module adds social value to the proposed action. The online social context learning method updates the social module to adapt it to the new social environment, which is represented by human trajectories.
  • Figure 2: The overview diagram of our method. Based on human and robot states, the control algorithm identifies the optimal action from the action space. The human information is applied to update the social module. The 'human' and 'robot' tracklets creator blocks are identical. The orange and green blocks are the ORL-based social module and the DRL robot navigation elements respectively. The yellow block is the combination of ORL and DRL results.
  • Figure 3: A example of robot tracklet and the proposed social value network.
  • Figure 4: Real robot experiment and its results.