doctoral contract (M/F) : Long-Distance Visual Familiarity Navigation: Application to a Quadruped Robot

Job Description

Thesis Subject

-Problem Statement

Autonomous navigation is a major challenge, especially in the absence of satellite positioning systems (GNSS) or communication infrastructure. Current solutions, such as lidars, are often energy-intensive, expensive, and active, making them less discreet.

Recent research has shown that panoramic images of just a few thousand pixels can be used to memorize and retrace paths. This basic visual approach is currently paired with minimalist memorization, enabling mobile robots and drones to follow paths of tens of meters (Gattaux et al., 2025, 2026).

The idea is to modify visual processing to learn and retrace longer and more varied visual paths, particularly those accessible to legged robots.

Approach
This PhD project draws inspiration from the visual strategies of invertebrates (e.g., ants, bees), which navigate with remarkable efficiency despite limited cognitive resources.

Many animals also appear to use visual familiarity-based processing to find their home and to memorize familiar routes.

The visual familiarity principle (ultra-frugal AI), inspired by invertebrate cognitive mechanisms, relies on a minimalist encoding of visual information. Unlike classical computer vision approaches — which require significant computational resources and memory — this method captures and memorizes only the massive contrast combinations present in an image or sequence of images (panoramic or otherwise). These contrast combinations are encoded and stored in a vector of just 15,000 bits (Gattaux et al., 2025, 2026), called a MBON (Mushroom Body Output Neurons, referencing insect neural structures). Each MBON acts as a compact visual memory, encoding the essence of a scene without storing unnecessary details.

During the recall phase, a new image is encoded using the same principle and compared to the stored MBON(s). A familiarity index is then calculated, quantifying the similarity between the current image's signature and the stored visual signatures. This mechanism enables a robot to recognize previously explored locations with remarkable precision while using minimal hardware and energy resources. This approach — both frugal and robust — paves the way for autonomous navigation systems suited to constrained environments, where efficiency and sobriety are critical.
The idea will be to adapt the minimalist visual processing pipeline to memorize longer and more varied visual paths accessible by a legged robot.

Job Profile

This multidisciplinary thesis requires strong expertise in several of the following areas:

Robotics, computer vision, control systems, dynamic modeling, signal processing, or machine learning.
Programming experience (Python, MATLAB) and handling of experimental data.
Interest in bio-inspiration, animal cognition, mobile robotics, or autonomous systems.

The candidate must hold one of the following degrees:

A Master's or engineering degree in robotics, AI, computer science, embedded systems, computer vision, or mechatronics.

Your Work Environment

Work Context

The selected candidate will be employed by the CNRS on a 3-year fixed-term contract at the Lab-STICC UMR6285 laboratory (labsticc.fr) on the ENSTA Brest campus (ensta.fr) and will be enrolled in the SPIN Doctoral School (ed-spin.doctorat-bretagne.fr).

This thesis is part of the ANR Ant'noid project (https://anr.fr/Projet-ANR-24-CE33-0218 ).


Additional Information

Applications must be submitted via the CNRS portal (PDF format only) and must include:

A detailed CV highlighting relevant knowledge and experience.
A cover letter explaining your motivation and interest in the position.
Detailed academic transcripts (Bachelor's, Master's, or engineering degrees).
Full contact details (name, position, email, phone) of two references (e.g., internship or academic supervisors) who can be contacted directly.

Applications will be reviewed on a rolling basis until the position is filled. The expected start date is October 2026, with some flexibility.

For any questions, please contact F. Ruffier: franck.ruffier@cnrs.fr.

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions