Deploying robots in human-shared spaces requires understanding interactions among nearby agents and objects. Modelling cause-and-effect relations through causal inference aids in predicting human behaviours and anticipating robot interventions. However, a critical challenge arises as existing causal discovery methods currently lack an implementation inside the ROS ecosystem, the standard de facto in robotics, hindering effective utilisation in robotics. To address this gap, this paper introduces ROS-Causal, a ROS-based framework for onboard data collection and causal discovery in human-robot spatial interactions. An ad-hoc simulator, integrated with ROS, illustrates the approach's effectiveness, showcasing the robot onboard generation of causal models during data collection.
ROS-Causal is a ROS-Based Causal Analysis Framework that facilitates robot onboard data collection and causal discovery in human-robot spatial interactions. It is composed by four main ROS nodes:
roscausal_robot: collects data from several rostopics related to the robot
(e.g., position, velocity, target position, etc.), and merge them into a single rostopic roscausal/robot
roscausal_human: collects data from several rostopics related to the human
(e.g., position, velocity, target position, etc.), and merge them into a single rostopic roscausal/human
roscausal_data: subscribes to the topics /roscausal/robot and /roscausal/human
and begins collecting data in a CSV file. Once the desired time-series length (rosparam) is reached, the node provides the
option to post-process the data and finally saves the CSV file into a designated folder
roscausal_discovery: performs causal discovery analysis on the collected data and publishes the result on
the roscausal/causal_model rostopic.
So far, it incorporates two causal discovery methods: PCMCI and F-PCMCI.
roscausal_data and roscausal_discovery ROS nodes operate asynchronously,
allowing the simultaneous execution of causal analysis on one dataset while continuing the collection of another.
In this paper,
we provide a detailed technical explanation of ROS-Causal,
accompanied by an experimental evaluation in a simulated environment: ROS-Causal_HRISim.
ROS-Causal_HRISim is a Gazebo-based human-robot interaction simulator that accurately replicates HRI scenarios
involving a TIAGo robot and multiple pedestrians. The pedestrians are controlled either by social forces or user
teleoperation (via keyboard). The simulator has been designed to facilitate the setup of real-life HRI scenarios
and the execution of causal analysis within them. For the latter, ROS-Causal has been integrated into the ROS-Causal_HRISim simulator.
To evaluate ROS-Causal, we designed an HRI scenario involving a TIAGo robot and a teleoperated individual, represented by the red manikin. The green dot indicates the target position for the person. For the causal analysis, we considered a predefined set of variables, comprising:
@inproceedings{castri2024ros,
title={ROS-Causal: A ROS-based Causal Analysis Framework for Human-Robot Interaction Applications},
author={Castri, Luca and Beraldo, Gloria and Mghames, Sariah and Hanheide, Marc and Bellotto, Nicola},
booktitle={Workshop on Causal Learning for Human-Robot Interaction (Causal-HRI), ACM/IEEE International Conference on Human-Robot Interaction (HRI)},
year={2024}
}
Deploying robots in human-shared environments requires a deep understanding of how nearby agents and objects interact. Employing causal inference to model cause-and-effect relationships facilitates the prediction of human behaviours and enables the anticipation of robot interventions. However, a significant challenge arises due to the absence of implementation of existing causal discovery methods within the ROS ecosystem, the standard de-facto framework in robotics, hindering effective utilisation on real robots. To bridge this gap, in our previous work we proposed ROS-Causal, a ROS-based framework designed for onboard data collection and causal discovery in human-robot spatial interactions. In this work, we present an experimental evaluation of ROS-Causal both in simulation and on a new dataset of real-world human-robot spatial interactions to assess its performance and effectiveness. Our analysis demonstrates the efficacy of this approach, showcasing how causal models can be extracted directly onboard by robots during data collection. The online causal models generated from the simulation are consistent with those from real-world experiments. These findings can help researchers to enhance the performance of robotic systems in shared environments, firstly by studying the causal relations between variables in simulation without real people, and then facilitating the actual robot deployment on real human environments.
In this paper, we extend the previous work by
adding an experimental evaluation of ROS-Causal by using HRSI data from a real-world scenario.
The experiment and data collection occurred in a laboratory room of the University of Lincoln (UK), measuring 5 x 8.2m.
Fifteen participants (6 females), aged between 25 and 55 took part in the experiment. Seven of them were used to work with a robot.
They were required to walk between four goal positions (G1, G2, G3, G4)
and avoid the robot if a cross occurs.
A predefined rectangular path was set for the TIAGo robot to navigate along the room and generate frequent interactions with the participants.
The robot was equipped by a 3D Velodyne VLP-16 LiDAR (1) to track the person taking part at the experiment.
|
|
|
The following figures present the execution time and data requirement analyses.
In particular, (left) and (centre) depict the execution time and the Structural Hamming Distance (SHD) metrics
for various time horizon (time-series length). In this analysis, we execute the causal analysis using both PCMCI
and F-PCMCI causal discovery methods
included in ROS-Causal, considering different percentages of the time-series length, ranging from 10% up to 100% of the full
length, corresponding to an average of 5 minutes per participant. For each time-horizon, we measure the execution time needed for
the causal discovery analysis and the SHD of the reconstructed causal model.
On the other hand, (right) shows an analysis of the sampling frequency. Specifically, considering the full length time-series,
we performed the causal discovery analysis for various sampling frequencies, ranging from 0.5Hz to 10Hz that is the original
sampling frequency.
These analyses helped us identify the sampling frequency and time-series length required to generate accurate causal models for this specific
HRSI scenario, while also evaluating the execution time for the causal analysis. More details and considerations can be found in the associated paper.
The dataset captures a Human-Robot Spatial Interaction (HRSI) scenario between a person and the TIAGo robot. It focuses specifically on human-goal and human-robot spatial interaction in an indoor environment, captured from the perspective of a 3D Velodyne VLP-16 LiDAR mounted on the TIAGo robot. It includes:
@inproceedings{castri2024exp,
title={Experimental Evaluation of ROS-Causal in Real-World Human-Robot Spatial Interaction Scenarios},
author={Castri, Luca and Beraldo, Gloria and Mghames, Sariah and Hanheide, Marc and Bellotto, Nicola},
booktitle={2024 33nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)},
pages={},
year={2024},
organization={IEEE}
}