A list of my projects developed during my PhD at Aston University, Birmingham.
A dataset for social navigation with robots. It captures the positions of humans, objects, and the robot within a room environment. As labels, it includes a discomfort score that quantifies how the robot's current location disrupts people's comfort.
SNGNN is a Graph Neural Network to model adherence to social-navigation conventions for robots. Given a particular scenario composed of a room with any number of walls, objects and people (who can be interacting with each other) the network provides a social adherence ratio from 0 to 1. This network is trained with SocNav1 dataset.
SONATA is a tool to generate datasets for social navigation with robots and Human Robots Interaction (HRI) scenarios. It provides a set of tools to generate synthetic data, including 3D scenes, human trajectories and their interactions.
SocNav2 is the second version of SocNav1. This new datasets includes several improvements such as dynamic scenarios and an additional score to evaluate how well the robot navigates towards the goal.
This is the second version of SNGNN. This Graph Neural Network also models adherence to social-navigation conventions for robots. However, this new version provides several improvements over its predecessor.
This projects consitis of the generation of disruption maps for Human-Aware Navigation (HAN) with robots using graph neural networks. This maps can be used for real-time HAN as demonstrated in the experiments.
SNGNN is a Graph Neural Network to model adherence to social-navigation conventions for robots.
A combination of GNNs and the well-known GAN arquitecture called SPADE for traffic image generation.
A combination of GNNs and the well-known GAN arquitecture called SPADE for traffic image generation.
