Machine Learning for dynamics systems

Download job offer in PDF format

Contract type : Fixed-term contract

Level of qualifications required : PhD or equivalent

Other valued qualifications : Master/PhD in robotics

Fonction : Temporary scientific engineer

Level of experience : From 3 to 5 years

About the research centre or Inria department

The Inria centre at Université Côte d'Azur includes 42 research teams and 9 support services. The centre's staff (about 500 people) is made up of scientists of different nationalities, engineers, technicians and administrative staff. The teams are mainly located on the university campuses of Sophia Antipolis and Nice as well as Montpellier, in close collaboration with research and higher education laboratories and establishments (Université Côte d'Azur, CNRS, INRAE, INSERM ...), but also with the regiona economic players.

With a presence in the fields of computational neuroscience and biology, data science and modeling, software engineering and certification, as well as collaborative robotics, the Inria Centre at Université Côte d'Azur  is a major player in terms of scientific excellence through its results and collaborations at both European and international levels.

 

Context

The ACENTAURI research team (https://team.inria.fr/acentauri/), located in the Inria Center of the Côte d'Azur University in Sophia-Antipolis, is offering a robotics engineer position in Machine learning for dynamics systems

ACENTAURI is a robotics team that studies and develops autonomous and intelligent robots that collaborate with each other to perform difficult tasks in complex and dynamic environments.

The team addresses perception, decision and control problems for multi-robot collaboration by proposing an original hybrid approach to artificial intelligence based on models and data and by studying efficient algorithms. The team focuses on applications such as multi-robot patrol systems for environmental monitoring and transporting people and goods. In these applications, several robots share multi-sensor information possibly coming from the infrastructure.

The effectiveness of the proposed approaches is demonstrated on real robotic systems such as cars and drones in collaboration with industrial partners.

Context

Ground vehicle testing today generates vast amounts of data from inertial measurement units, positioning systems, onboard sensors, and dedicated measurement chains. While this data captures the vehicle's actual behavior, fully leveraging it remains challenging; the data is often noisy and incomplete, and it captures only a partial picture of the system's physical state.

Traditional physics-based models remain the benchmark for interpretability, yet they become costly to develop and calibrate when representing phenomena such as tire-ground interactions or highly non-linear regimes. Conversely, purely statistical machine learning models struggle to generalize beyond their training domain and often lack the interpretability required by domain experts.

Inria and DGA Techniques terrestres have established a research partnership to modernize the numerical methods used for military vehicle testing. This position falls within that framework, focusing on Physics-Informed AI—at the intersection of physics-based modeling and machine learning—applied to the analysis and simulation of vehicle dynamics.

The dynamic systems addressed in this postdoctoral project belong to two categories: the general class of dissipative multibody mechanical systems (characterized by friction, damping, and energy dissipation) and the specific systems studied by DGA Techniques terrestres regarding longitudinal vehicle dynamics (acceleration, braking, and force transmission at the tire-ground interface). Experiments conducted during the project must relate to this representative application case, ensuring that the developed methods directly address DGA Techniques terrestres' operational needs regarding vehicle testing and modeling.

 

Potential Scientific Directions

This position is part of a broader program dedicated to utilizing test data and developing hybrid models for vehicle dynamics.

Depending on the recruit's profile, available data, and priorities established with the DGA Land Systems (DGA Techniques terrestres) teams, the work will focus primarily on one main area—potentially complemented by a second—chosen from the following:
 
  • Test data processing and qualification.** Work may involve analyzing multi-sensor recordings, signal synchronization, detecting outliers or drift, and automatically identifying the various dynamic regimes present in the tests.
  • Hybrid physics-based/learning-based modeling.
  • This may entail combining existing physical models with learned components or parameters identified from data to better represent phenomena that are poorly understood or difficult to model. Particular attention will be paid to stability, adherence to physical constraints, and the interpretability of results.
  • Virtual sensors and uncertainty quantification
  • Research may aim to estimate quantities that are difficult to measure directly using available sensors, while also assessing the confidence level associated with these estimates. Approaches considered may include Bayesian methods, model ensembles, or GPU-accelerated Monte Carlo methods.

    The specific scientific program will be defined at the start of the contract to establish a coherent and feasible project within the postdoctoral timeframe. The goal is to explore a limited number of scientific questions in depth while validating them against real experimental data and a representative use case.

    The work is expected to lead to publishable scientific contributions as well as a documented, reproducible implementation. Depending on the maturity of the results, the work may also contribute to a demonstrator for the DGA Land Systems teams.

Resources provided

The successful candidate will have access to models and tools developed within the framework of the partnership, as well as computing resources suited to the requirements of machine learning and numerical simulation.

Professional licenses for development assistants—specifically OpenAI Codex and Claude Code—will also be provided to facilitate prototyping, test writing, the exploration of different approaches, and code documentation.

These tools will be used under the supervision of the successful candidate, who must be capable of evaluating and validating the generated results from scientific, mathematical, and software perspectives.

The work will benefit from the support of Inria researchers specializing in artificial intelligence, control theory, and robotics; engineers specializing in simulation and data processing; and teams from DGA Land Systems involved in the testing.

Travel: Travel to the DGA-TT site in Angers is required.

Keywords: Physics-informed AI, hybrid physics-learning models, dynamical systems, dissipative multibody mechanical systems, longitudinal vehicle dynamics, system identification, virtual sensors, uncertainty quantification, multi-sensor data processing, military vehicle testing, scientific computing, machine learning.

Assignment

The missions entrusted to the research engineer will mainly be on hybrid AI, in particular:

  • Develop multi-scale and multi-frequency learning methods
  • Develop Multi-step learning methods with hidden states
  • Adapt learning methods by implementing a Lyapunov approach
  • Develop physics-informed Bayesian learning methods
  • Extend work in state estimation (DL-MHE)
  • Code and test algorithms
  • Write publications  

Collaboration : The candidate will work in close collaboration with the ACENTAURI team, the DGA-TT and one research collaborator at LERIA.

Responsibilities: The candidate will have to integrate into the ACENTAURI team and participate in animation of the team. In addition, he (she) will have to carry out the important tasks of communication, publication writing, and methodology implemented in ACENTAURI (project monitoring and project management under Git and Gitlab). He will participate in the supervision of masters

Main activities

  • Bibliographic research
  • Proposal and coding of AI Hybrid solutions
  • Test on Datasets and in real conditions
  • Comparison of solutions
  • Writing publications

Complementary activities:

  • Write the reports and deliverables for the DGA-TT contract
  • Write meeting minutes
  • Test, modify until validated

Skills

Technical skills and level required:

The candidate should preferably have obtained a PhD in AI for Robotics (PhD in artificial intelligence, control systems, robotics, computational mechanics, applied mathematics, signal processing, scientific simulation, or a related field).

. The candidate must have a solid foundation in software development (Matlab, C/C++, Python, Git, OpenCL, CMAKE, ...), machine learning methods (learning and inference) and modelling and control of robots.

Languages:

a good level in English read/written/spoken is expected.

Interpersonal skills:

The candidate will be in contact with the members of the team and will have to integrate into the ACENTAURI team. He/she must have the appropriate relational qualities.

Additional skills appreciated: He/she must also be highly motivated for multidisciplinary studies and all aspects of R&D ranging from fundamental to experimental work.

Required skills and experience

Strong skills in at least two of the following areas:

    Machine learning / artificial intelligence
    Dynamic modeling (differential equations, system identification, numerical solvers)
    Scientific computing and numerical method development (Python, PyTorch, NumPy)
    Physics and mechanics, particularly multibody dynamics

Candidates are expected to be able to develop, test, and document scientific code, as well as rapidly acquire proficiency in other aspects of the subject matter.

Desired experience (not all boxes need to be checked)

    Physics-informed AI or hybrid physics-AI models
    System identification
    State estimation under partial observation
    Uncertainty quantification
    Multi-sensor data processing

Asset (not a prerequisite)

    Experience in automotive dynamics

Benefits package

  • 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)

Remuneration

From 2692 € gross monthly (according to degree and experience)