Phd Position F - M Proactive Navigation In Human-Populated Environment. H/F - INRIA

A propos d'Inria Inria est l'institut national de recherche dédié aux sciences et technologies du numérique. Il emploie 2600 personnes. Ses 215 équipes-projets agiles, en général communes avec des partenaires académiques, impliquent plus de 3900 scientifiques pour relever les défis du numérique, souvent à l'interface d'autres disciplines. L'institut fait appel à de nombreux talents dans plus d'une quarantaine de métiers différents. 900 personnels d'appui à la recherche et à l'innovation contribuent à faire émerger et grandir des projets scientifiques ou entrepreneuriaux qui impactent le monde. Inria travaille avec de nombreuses entreprises et a accompagné la création de plus de 200 start-up. L'institut s'eorce ainsi de répondre aux enjeux de la transformation numérique de la science, de la société et de l'économie.
PhD Position F/M Proactive navigation in human-populated environment.
Le descriptif de l'offre ci-dessous est en Anglais
Type de contrat : CDD

Niveau de diplôme exigé : Bac +5 ou équivalent

Autre diplôme apprécié : Master

Fonction : Doctorant

Niveau d'expérience souhaité : Jeune diplômé

A propos du centre ou de la direction fonctionnelle

Inria is the French National Institute for Research in Digital Science, of which the Inria Côte d'Azur University Center is a part. With strong expertise in computer science and applied mathematics, the research projects of the Inria Côte d'Azur University Center cover all aspects of digital science and technology and generate innovation. Based mainly in Sophia Antipolis, but also in Nice and Montpellier, it brings together 47 research teams and nine support services. It is active in the fields of artificial intelligence, data science, IT system security, robotics, network engineering, natural risk prevention, ecological transition, digital biology, computational neuroscience, health data, and more. The Inria Center at Université Côte d'Azur is a major player in terms of scientific excellence, thanks to the results it has achieved and its collaborations at both European and international level.

Contexte et atouts du poste

ACENTAURI and HUCEBOT are involved in the PEPR Robotique through the labelled project HAMMER funded by the ANR. They propose a PhD topic in Proactive navigation in human-populated environment.

ACENTAURI is a robotic team located in Sophia Antipolis that studies and develop intelligent, autonomous and mobile robots that collaborate between them to achieve challenging tasks in dynamic environments. The team tackles perception, decision and control problems for multi-robot collaboration by proposing an original hybrid model-driven/data driven approach to artificial intelligence and by studying efficient optimisation algorithms. The team focus on robotic applications like environment monitoring and transportation of people and goods. In these applications, several robots will share multi-sensor information eventually coming from infrastructure. The effectiveness of the proposed approaches are demonstrated on real robotic systems like cars AGVs and UAVs together with industrial partners. Since, last five years the team members are focusing on autonomous navigation in human populated environment, exploring reactive and cooperative navigation while targeting to develop the concept of proactive navigation.

HUCEBOT is a robotic team based in Nancy dedicated to advancing algorithms for human-centered robots: robots that can autonomously operate while interacting, collaborating, and assisting humans to the best of their capabilities. Our goal is to improve workplace conditions by enhancing ergonomics and safety through robotic technologies, either by physically assisting humans or by replacing them in dangerous and/or remote environments through teleoperated robotic avatars. To achieve this vision, we combine machine learning and modelbased approaches, investigating how to effectively integrate both paradigms to advance motion generation and interaction control for generic robotic platforms, including mobile manipulators and legged systems, operating in complex environments and in close interaction with humans. Experimental campaigns, possibly involving human participants, are a central component of our research approach, aiming at validating and demonstrating our robotic technologies in the wild.

To achieve autonomous navigation in human-populated environment (such as in historical centers, museums, fairs, etc.), it is necessary to estimate the behavior of actors (including humans, moving obstacles, and other robots), to compute real-time control while ensuring safety, and to perform proactive navigation (i.e., compute actions that are compatible and acceptable to humans while avoiding the Freezing Robot Problem [1]). Understanding human behavior is likely the most complex issue that must be addressed. Firstly, understanding Human-to-Human interactions is essential. In [2], the author presents a categorization of the surrounding space into different types (Intimate, Personal, Social, Public) from which proxemics were defined for Human-to-Human interactions (H2H). In [3], the Social Force Model (SFM) is described. Pedestrian motion is modeled as a result of the application of a set of forces called social forces, which are applied to the pedestrian from different sources. Several versions of this model have been developed, including the Extended Social Force Model (ESFM) [4], the Headed Social Force Model (HSFM) [5], and the Collision Prediction Extension of Social Force Model (CPESFM) [6]. All these extensions consider interactions with Humans (H2H), Obstacles (H2O), Robots (H2R), and, more generally, everything (H2X). Based on the H2X Social Force model, the SPACiSS simulator [7] was developed and recently used to simulate pedestrian behavior in the presence of an autonomous vehicle operating in a shared and open space proactively [8]. The accumulated knowledge in H2X interaction modeling can be beneficial to the modeling of Robot to everything (R2X). However, predicting human intentions in particular environments remains challenging. For example, in [9], the authors propose a Vehicle to Human (V2H) interaction model where the pedestrian's cooperation with the vehicle is estimated online. Recently, ACENTAURI has started to work on a further extension of the SFM by defining the Universal Social Force
Model (USFM) that can be used by the robot controller for autonomous navigation.

In this PhD subject, we intend to complete and finalize this model and deploy it in different environments. The USFM can be used either for H2X interaction modeling, but also for R2X interaction modeling, making the robot behavior (mainly robot motion) more sociable and liable in human and other autonomous, or human-operated devices, populated environments. Computing real-time control while ensuring safety over a fixed time horizon is a challenging task. One approach to address this challenge is Model Predictive Control (MPC), which involves computing optimal control while anticipating future events within a time horizon using actual state measurements. However, state-of-the-art works often fail to mention the computational efficiency of MPC in real-time applications. A major drawback of using an MPC controller is that it requires considerable time and computational resources to perform an online optimization problem at each time step, which limits its application in realworld scenarios [10]. Control parametrization is an effective solution [11][12], which significantly reduces the number of control variables in the optimization problem while minimizing performance loss. Recent works have applied control parametrization on real-time simple systems [13] and dynamic vehicles [10]. Recently, ACENTAURI has worked on a new control input parametrization [14] that outperforms previous ones and allows for an efficient real-time implementation on a real robot in complex scenarios. In this PhD subject, we will explore this idea. Another interesting work was proposed in [17]. The concept of dynamic channel is introduced as a way to define an open channel for navigation between
the different agents of the situation. The main question here is how to select the dynamic anchors of the selected channel.

The main goals of the PhD will be to complete and extend the concept of Universal Social Force Model to be used for H2X or R2X interaction modeling, to develop a module to estimate the cooperability of the human and other present autonomous and human-operated devices, to put in place a proactive navigation strategy based on a parametrized nonlinear MPC framework. The outcome of the work will be validated in a physical environment using a ROS2-based robot system. To accurately estimate an actor's behavior, learning the model automatically from demonstrated real-world data using Inverse Reinforcement Learning is one option to be investigated. An example is given in [15] for autonomous driving application. Another area of investigation is tuning parameter values using reinforcement learning [19] and incorporating human action prediction into the MPC [18]. In ACENTAURI and HUCEBOT, we strongly believe that developing hybrid techniques (mixing data-based and knowledgebased techniques) is an important track to be followed. An example is given in[16].

[1] P. Trautman and A. Krause, Unfreezing the robot: Navigation in dense, interacting crowds, in 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2010, pp. 797-803.
[2] E. Hall, The Hidden Dimension. Garden City N.Y: Doubleday, 1966.
[3] D. Helbing and P. Molnar, Social force model for pedestrian dynamics, Physical Review E (PRE), vol. 51, no. 5, p. 4282, 1995.
[4] G. Ferrer, A. Zulueta, , F. Cotarel, and A. Sanfeliu, Robot socialaware navigation framework to accompany people walking side-byside, Autonomous Robots, vol. 41, no. 4, pp. 1573-7527, 2017.
[5] F. Farina, D. Fontanelli, A. Garulli, A. Giannitrapani, and D. Prattichizzo, Walking ahead: The headed social force model, PLOS ONE, vol. 12, no. 1, pp. 1-23, 2017.
[6] F. Zanlungo, T. Ikeda, and T. Kanda, Social force model with explicit collision prediction, Europhysics Letters, vol. 93, no. 6, p. 68005, 2011.
[7] M. Predhumeau, L. Mancheva, J. Dugdale, A. Spalanzani, Simulating Realistic Pedestrian Behaviors in the Context of Autonomous Vehicles in Shared Spaces., in Proceedings of the 20th International Conference on Autonomous Agents and MultiAgent Systems (AAMAS '21), pp. 1010-1018, May 2021,
[8] M. Kabtoul, A. Spalanzani, and P. Martinet, Proactive and smooth maneuvering for navigation around pedestrians, in IEEE International Conference on Robotics and Automation (ICRA), 2022, pp. 4723- 4729.
[9] M. Kabtoul, A. Spalanzani, and P. Martinet, Towards proactive navigation: A pedestrian-vehicle cooperation based behavioral model, in IEEE International Conference on Robotics and Automation (ICRA), 2020, pp. 6958-6964.
[10] Z. R. M. Junior, A. M. De Almeida, and R. V. Lopes, Vehicle stability upper-level-controller based on parameterized model predictive control, IEEE Access, vol. 10, pp. 21 048-21 065, 2022.
[11] M. Alamir, Stabilization of nonlinear systems using receding-horizon control schemes: a parametrized approach for fast systems. Springer, 2006, vol. 339.
[12] M. Muehlebach and R. D'Andrea, Parametrized infinite-horizon model predictive control for linear time-invariant systems with input and state constraints, in American Control Conference (ACC), 2016, pp. 2669-2674.
[13] F. Fusco, G. Allibert, O. Kermorgant, and P. Martinet, Benchmarking nonlinear model predictive control with input parameterizations, in International Conference on Methods and Models in Automation and Robotics
(MMAR22), 2022.
[14] E. Fiasch´e, P. Martinet, E. Malis Towards autonomous robot navigation in human populated environments using an Universal SFM and parametrized MPC, 2023 IEEE/RSJ International Conference on Intelligent Robots and
Systems (IROS23), Detroit, USA, October 1-5th, 2023
[15] David Sierra Gonz´alez, ¨Ozg¨ur Erkent, V´ictor Romero-Cano, Jilles Dibangoye, Christian Laugier, Modeling Driver Behavior From Demonstrations in Dynamic Environments Using Spatiotemporal Lattices, in ICRA 2018 - Proceedings of the 2018 IEEE International Conference on Robotics and Automation, Brisbane, Australia, May 2018, pp. 3384-3390
[16] Z. Liu, E. Malis, P. Martinet, A New Dense Hybrid Stereo Visual Odometry Approach, 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS22), Kyoto, Japan, October 23-27th, 2022
[17] C. Cao, P. Trautman, S. Iba, Dynamic Channel: A Planning Framework for Crowd Navigation, International Conference on Robotics and Automation (ICRA), Montreal, Canada, May 20-24, 2019
[18] L. Vianello, J.B. Mouret, E. Dalin, A. Aubry, and S. Ivaldi, Human Posture Prediction During Physical Human-Robot Interaction, in IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 6046-6053, July 2021
[19] L. Penco, E. M. Hoffman, V. Modugno, W. Gomes, J. -B. Mouret and S. Ivaldi, Learning Robust Task Priorities and Gains for Control of Redundant Robots, in IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 2626-2633, April 2020

Mission confiée

Work program
- State of the art on social and autonomous navigation
- State of the art on human behavior understanding and modeling.
- State of the art on RL and MPC
- Development and evaluation of a human behavior model
- Development and evaluation of a Human behavior prediction module based
on a combination of MPC and RL.
- Development and evaluation of a proactive navigation framework
- Real implementation and experimental test on a real mobile robot in
Sophia-Antipolis.

Principales activités
Key activities: Conduct a state-of-the-art review Implement simulation tools Develop methods and devise solutions Implement the proposed solutions in a real-world setting Showcase contributions through high-quality publications Additional activities: Provide weekly progress updates Set objectives Establish a work plan
Compétences
Languages: English is essential for disseminating knowledge Interpersonal skills: comfortable and confident in a scientific environment
The ideal candidate has a strong background in planning, control, and robotics.

The candidate must be a proficient user of C/C++, Python, and ROS 2, and any relevant computer vision library (e.g., ViSP, OpenCV, PCL). Previous experience with frameworks for MPC and RL is a plus. Scientific curiosity,
large autonomy, and the ability to work independently are also expected.

Avantages

- Subsidized meals
- Partial reimbursement of public transport costs
- Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
- Possibility of teleworking and flexible organization of working hours
- Professional equipment available (videoconferencing, loan of computer equipment, etc.)
- Social, cultural and sports events and activities
- Access to vocational training
- Contribution to mutual insurance (subject to conditions)

Rémunération

Duration: 36 months

Location: Sophia Antipolis, France

Gross Salary per month: 2300 €

• Python
• Anglais
• Computer vision
• OpenCV
• Langage C
• Pro-activité