Learning Social Navigation from Demonstrations with Deep Neural Networks

TL;DR

The paper tackles social navigation by introducing a hierarchical, data-efficient learning framework that uses Conditional Neural Processes (CNPs) to power both global trajectory generation and local velocity control. By training two CNP-based modules on demonstrations generated from a Social Force Model in simulation, the approach can produce multi-modal global paths and reactive local commands that adapt to obstacle changes. Experiments show that CNP-based global planning outperforms a standard feed-forward network in avoiding obstacles, while the local planner imitates socially compliant motion under dynamic conditions. The work demonstrates a promising direction for data-efficient, human-aware navigation with potential extensions to real-robot deployment and group-detection enhancements.

Abstract

Traditional path-planning techniques treat humans as obstacles. This has changed since robots started to enter human environments. On modern robots, social navigation has become an important aspect of navigation systems. To use learning-based techniques to achieve social navigation, a powerful framework that is capable of representing complex functions with as few data as possible is required. In this study, we benefited from recent advances in deep learning at both global and local planning levels to achieve human-aware navigation on a simulated robot. Two distinct deep models are trained with respective objectives: one for global planning and one for local planning. These models are then employed in the simulated robot. In the end, it has been shown that our model can successfully carry out both global and local planning tasks. We have shown that our system could generate paths that successfully reach targets while avoiding obstacles with better performance compared to feed-forward neural networks.

Figures (7)

• Figure 1: Comparison between regular and social navigation.
• Figure 2: General layout of the training phase of our model.
• Figure 3: Data collection on the simulation. The motion trajectory is shown with blue line.
• Figure 4: CNP as the global planner.
• Figure 5: Comparison between our global planner network (CNP) and a 5-layered feed-forward neural network (NN) on global planning in sample environments.