7.1.1 PSOLVER

• Name:
  Parametric equations solver with IA
• Keywords:
  Numerical solver, Artificial intelligence
• Functional Description:
  This solver mixes supervised deep learning and deterministic methods to obtain exact solutions (i.e. the differences between the provided solutions and the real ones may be arbitrary small). It cannot guarantee to find all solutions but it has a self-learning capacity so that the probability of missing a solution is very low. The solver is based on the predictions of several neural networks that are used as guess for the Newton scheme so that the solving time is low, usually around a few milliseconds. The computer intensive part for establishing the solver is the training of the neural networks using automatically generated learning sets. It shall be used only if the system has to be solved for multiple occurrences of P.
• Contact:
  Jean-Pierre Merlet

8.1.1 Preventive maintenance of cable driven-parallel robots

Participants: Jean-Pierre Merlet [correspondant], Romain Tissot.

The cables of a cable-driven parallel robot (CDPR) follow a path through several mechanical elements that lead to a progressive wear of the cable. As CDPR may be used to assist frail people to walk, for rehabilitation purpose or in industry (e.g. logistic), it is important to estimate the cable wear in order to avoid a situation where a cable may break down. Our objective is therefore predictive maintenance where trends will be estimated for the wear in order to allow to change the cable(s) before a breakdown. Classical cable models involve the Young modulus $E$ and the linear density $\mu$ of the cable material. The Young modulus reflects the elasticity of the cable (i.e its length change when submitted to a tension) and we have estimated that wear will mostly affect it by decreasing its value compared to its value $E_b$ when the cable is brand new.

For a given pose $𝐗$, we can compute the cable lengths $L_0$ under the assumption of known Young modulus for all cables. Wear in cable(s) will decrease $E$ so that the cable lengths will increase which in turn will change the pose ${𝐗}_{𝐦}$ that the robot will reach. As we have means to measure accurately the pose (or part of it) we have simulated the difference $𝐗-{𝐗}_{𝐦}$ for different calibrations poses and we have shown that, if $E$ goes below $E_b/2$, the difference is big enough to be measured. We have then addressed the inverse problem where $𝐗-{𝐗}_{𝐦}$ is measured for different calibrations poses, the objective being to estimate the Young modulus of the cables. We have shown that it is possible to find calibration poses where only a pair of cables are supporting the load (the other being slack) and we have developed two approaches to estimate the Young modulus of this pair of cables:

1. an adapted gradient descent algorithm which has been derived from our work on the inverse kinematics of CDPR (where an optimality criterion has to be defined under the constraints that kinematic equation have to be fulfilled) 19
2. an IA based approach, derived from our work on the inverse kinematics of CDPR 20

The first algorithm is rather slow (about half an hour for a cable pair) but provides almost exact estimation of $E$ while being reasonably robust with respect to measurement errors. The second algorithm is very fast (almost real time) but provides more approximate values for $E$ and is more sensitive to measurement errors. However these results may be improved by creating several neural networks that will be specialized to deal with specific ranges for the values of $E$.

Hence wear estimation is in theory feasible at a low cost but an experimental validation of this result requires very specific equipment for measuring the Young modulus of the cables at different times that are not available to us.

The PhD defense of Romain Tissot presenting some of this work has been held on December 13 22.

8.1.2 Parametric equations solver with IA

Participants: Jean-Pierre Merlet.

Our work in robotics has led us to investigate the use of IA for finding the real root(s) of parametric non linear square system of equations $𝐅(𝐗,\Lambda )=\mathbf{0}$ i.e. systems which have as many unknowns $𝐗$ as equations but with equation coefficients that are functions of a parameters set $\Lambda$ that is assumed to be bounded. Such a system has usually several solutions and their number depends on the parameters values and cannot be determined in advance. The equations must have an algebraic analytic form, not being differential and being at least $C1$.

Our robotics work has shown that as soon as we are able to determine the solutions (not necessarily all of them) for a small set $\Lambda =\{{\Lambda }_1,...{\Lambda }_n\}$ of specific systems, then we may create an hybrid solver mixing Multi-layer Perceptrons (MLPs) and the Newton scheme that provide solutions for any $\Lambda$. This solver requires the training of multiple MLPs and the training time may be large. However, as soon as the training is completed, the solving time is extremely low (typically less than 0.02 seconds). We cannot guarantee to obtain all solutions (although extensive tests have shown that in general we will miss very few solutions). Furthermore an associated self-learning process may allow to reduce the number of missed solution(s) as soon as new system instances are solved. This year we have extended this approach to arbitrary system 15. A first version of the corresponding prototype Psolver is presented in the new softare section 7.

8.1.3 Kinematics of soft robots and AI

Participants: Jean-Pierre Merlet.

There is a growing interest in soft robots for which the links are made of flexible material, many of which have a closed-loop structure (typically like parallel robots). This community has focused on the modeling of the links in various forms and their use for basic kinematic problems such as finding an inverse kinematic solution and starting from this solution finding a new pose for a given set of actuators positions. But these problems are already computer intensive, which prohibits to address more complex issues such as computing the robot workspace, finding all solutions of the direct kinematics or solving design problem (e.g. determining the geometry(ies) of the robot so that all poses in a given workspace are reachable). We have started cooperating with researchers in this domain to see if an IA-based solver may be developed for the inverse and direct kinematics that are the basic tools that may be used to solve these complex problems.

8.1.4 Green robotics

Participants: Jean-Pierre Merlet, Yves Papegay, Clara Thomas [correspondant].

About 40% of the 4 000 000 existing robots are performing manutention operations and about half of them are performing repetitive pick-and-place task where an object has to be moved from A to B. This motion usually involves only 3 translational degrees of freedom and possibly a rotation around the vertical axis.

For this operation, several types of mechanical architecture are used: serial, cartesian and Scara types. These architectures are not energy efficient as their structure imposes to actuate several heavy mechanical elements beside the load (e.g the serial type require energy just to stay in its pose). Furthermore these robots are controlled with a rather powerful computer as they may be used for other purposes and these systems use a lot of mechanical and electronic resources while being not very flexible and difficult to maintain. All in all, this amounts to a large energy consumption and an extensive use of resources for a simple task. Our objective is to propose a specific robot for this task with a much lower energy consumption, less resources and being simple to maintain and offering a larger flexibility.

The first step of our proposal is to adopt a different mechanical architecture based on cable-driven parallel robots. They have multiple advantages:

• they involve only a very limited mobile mass as only the cables are moving
• they are highly flexible: just by moving the winches, it is possible to cover any kind of workspace even very large ones (the FAST telescope covers a circular area of 500 m of diameter),
• their lifting capacity may be very high (our Marionet-crane has a lifting capacity of 2.5 tons),
• they are easy to maintain either by just changing the cables or the winches that are fairly standard,
• the energy efficiency or parallel robots is, in general, 25% higher than any other robot structure.

CDPR may be designed for various degrees of freedom but the simpler one, called the $N$-1 CDPR, has $N$ cables attached at the same point on the platform. This CDPR offers either planar motion with the 2-1 ($x-z$ motion) or spatial motion with the 3-1 and 4-1 (the difference being a larger workspace for the 4-1). The platform may provide an independent rotation motion if necessary.

To check the energy efficiency of a CDPR, we have built a 2-1 CDPR and we have defined a classical pick and place trajectory (a vertical upward motion from A, an horizontal motion toward B and a vertical downward motion at B) and we have measured its energy consumption over 500 trajectories. In parallel colleagues at INRIA Bordeaux are measuring the energy consumption of a serial robot for the same load and the same trajectory. Tests have not yet been fully completed but preliminary results show that the CDPR has an energy consumption which is between 500 and 1000 less than the serial robot. In parallel, we have built two small pieces of software for estimating the power consumption of 2-1 and 3-1 CDPR: one is based on a full theoretical model of the system and the other one on experimental measurements on the motor behavior that has allowed us to have an experimental model of the system. Both software have been used to evaluate the power consumption of our 2-1 prototype and the estimations are in very good agreement with the measured consumption. These models can also be used for the 4-1 with an additional complexity: in general a 4-1 has at most 3 cables simultaneously under tension (and in some special cases only 2) but for a given pose two different sets of three cables, called configurations, may be under tension each of which lead to a different power consumption. Enforcing a configuration for a given pose is easy (we just increase the length of the fourth cable) but at some points, called transition points, a set of 3 selected cables under tension cannot anymore ensure the mechanical equilibrium of the system and we have to move to another configuration hence realizing a configuration change. For a given polygonal trajectory, each segment may be decomposed in different sub-segments for which only 3 specific cables are under tension. This decomposition leads to multiple ways to realize the trajectory with different configurations. Our simulation software allow us to determine the way that leads to the minimal energy consumption (and we may also deal with the same problem but with a minimal number of configuration changes). We have also investigated the influence of the height $h$ of the winches and shown that the energy consumption decreases with $h$. For the 4-1 where the winch locations lies on the corner of a square whose side length is $a$, we have shown that the consumption decreases with $a$ (which has however a minimal value as the trajectory has to be enclosed in the square).

A drawback of suspended 2-1, 3-1 and 4-1 CDPR is the load oscillation. This problem may be limited by having a pair of cables on each winch exiting at two output points $A_1,A_2$ and attached at two points on the platform $B_1,B_2$ such that $\left|\right|{𝐀}_{\mathbf{1}}{𝐀}_{\mathbf{2}}\left|\right|=\left|\right|{𝐁}_{\mathbf{1}}{𝐁}_{\mathbf{2}}\left|\right|$ and consequently the quadrilateral $A_1{B}_1{B}_2{A}_2$ is a parallelogram whose side $B_1{B}_2$ is always parallel to $A_1{A}_2$. For the 2-1 this prohibits oscillation around the perpendicular to the CDPR plane while for the 3-1 and 4-1 all oscillations are avoided.

Regarding energy consumption we are still using a computer to control our CDPR although a low power computer may be used (a Raspberry Pi is largely sufficient) and electronic resources to control the robot motion and the objective is to get rid of this computer. For a $N$-1 CDPR, it is easy to determine the cable lengths time functions $\dot{\rho }\left(t\right)$ to execute a trajectory at a given speed. At any time we may convert $\dot{\rho }\left(t\right)$ to a distance $d\left(t\right)$ from a point $M$ to a fixed point $U$ as $d\left(t\right)=\dot{\rho }\left(t\right)/k+d_0$ where $k$ is an arbitrary amplification factor. Assume now that we have created a rotating came and have a slider that measure the distance between the contour and a fixed point $N$ located on the vertical: this cam may designed so this distance is exactly equal to $d$ at any time. Consequently, if the slider actuates a potentiometer, its tension may be used as velocity input for a motor controller that will actuate the winch and, with a cam for each cable, the CDPR will perform the trajectory at the desired speed up to a fixed gain. Consequently we have gotten rid of the computer and some electronic resources and have just added the energy consumption of a small motor that actuate the cable cams. We have not lost the possibility of programming the trajectory as new cams may easily be 3D-printed. Note that the cam may be designed to offer the possibility of moving from A to B and then back to A so that the cam will just rotate continuously.

We are also considering the case where the sliders will directly actuate the winches, using a single motor: we will be back to a mechanically programmable robot using a single motor !

This work on using CDPR geometrical modularity to minimize energy consumption will play a role for the second part of C. Thomas thesis that will be managing the modularity of a 4-1 CDPR devoted for facilitating the mobility of frail people. A fixed geometry CDPR allows only for mobility in a single room but we plan to use the geometrical modularity of CDPR to allow to use a CDPR with only four fixed winches in several rooms.

8.1.5 Underconstrained CDPR

Participants: Sara Gabaldo, Jean-Pierre Merlet [correspondant].

Underconstrained CDPR have less cables than the platform degrees of freedom (dofs). Typically we can consider a CDPR with 3 cables that are attached to three different points of the platform: they allow to control only three of the platform six dof but any combination of 3 dof among the 6 may be controlled and this may be interesting for some industrial tasks with the advantage of reducing the cost.

In a joint work with the laboratory IRMA of Bologna University, we are addressing the following problem: being given a translational trajectory for the platform that has to be performed at a given velocity, what should be the geometry of the CDPR so that the angular motion of the platform remains in a given bounded domain while the maximal distance between the desired trajectory and the followed one should be lower than a fixed threshold ?

We have to consider that a given CDPR, which is velocity controlled with a discrete-time control law, cannot follow exactly the trajectory and hence both the maximal positioning error with respect to the trajectory and the maximal angular motion are functions of the controller sampling time, of the desired speed and of the CDPR geometry but it is very difficult to establish analytic values for these maximum as functions of the parameters. However, being given a nominal trajectory, we have been able to design a continuation software to fully simulate the CDPR motion in this context and are working on a fast method to provide an upper bound on the positioning errors and on the amplitude of the angular motions. This software could be extended to deal with uncertainties on the CDPR geometry with the objective of providing worst case upper bounds. A brute force approach may then be used to determine good geometries.

8.1.6 Networks of conversational agents and robots

Participants: Yves Papegay.

The use of Large Language Models (LLMs) in robotics is an emerging trend. These new natural language processing capabilities aim to improve the induction of high-level robot behavior. The proposed approach relies on conversational agents organized in graphs, a powerful, modular, and flexible method that leverages LLMs to interpret and contextualize commands while integrating classical components for motion planning, dynamic environment control, or object manipulation. The first challenge is to reliably transform descriptions of the environment, instructions, or constraints expressed in natural language into a functional decomposition of robotic tasks. This requires combining the contextualization provided by LLM embedding mechanisms with access to factual information about the environment, whether dynamic (from sensors) or static. A second challenge lies in the increasing complexity of agent organization, which grows with the complexity of the tasks to be managed. This organization must dynamically adapt to changes in the robots’ behavior or capabilities while ensuring the correct execution of tasks. With specialized agents possessing domain expertise, these technologies promise to enhance robots' ability to adapt to complex situations and provide robust solutions to the application problems encountered.

This new explorating work is done in collaboration with David Daney (Auctus team, centre Inria de l'Université de Bordeaux).

8.2.1 Movements characterization with smart barriers

As human activities often rely on mobility, a central issue is the monitoring and characterization of the movements of a person or a group of people within a building. In the previous years, we have chosen to develop and use virtual barriers that record the date, direction, and an estimate of the speed of their crossings.

Thanks to available funding for equipment, we have been able to make progress simultaneously on experimentation (both hardware and algorithmic aspects) and on system design methods and information processing. We have set up two experiments, the first in a corridor serving rooms, offices, and consultation rooms at the Valrose EHPAD in Nice, and the second in a day hospital treating patients suffering from Alzheimer's disease, in Nice as well. In these past experiments (conducted over several months, 24 hours a day), we have recorded a significant number of barrier crossings.

Currently, the information extracted by merging sensor data allows for the generation of barrier-crossing instances and provides a certain characterization by indicating the direction and a rough estimate of the person's speed of movement. However, visual inspection of the various data allows for the identification of additional characteristics such as crossings by a wheelchair, the use of a cart, crossings by a group of people, and their proximity. Furthermore, the use of contextual data associated with the measurements allows for further refinement of the information regarding people's behavior.

This leads us to explore how AI methods can be used to more reliably and automatically characterize crossings and to leverage contextual data. These methods could also enable real-time monitoring and analysis, which our current work does not allow.

8.2.2 Smart digital liaison notebook

For seniors, aging well means staying as healthy, active, and independent as possible in their own home. However, when the loss of autonomy and dependency begins, if the person is isolated, remaining at home is complex: it requires the establishment of an ecosystem organization around them, including family, close friends, caregivers, and healthcare providers. Within this heterogeneous group, communication and sharing of daily information are crucial for the proper functioning and effectiveness of this organization, as well as for ensuring the peace of mind of both the individual and their family by demonstrating this proper functioning. Continuing developments made during the technological maturation of the startup creation project LOCANAC in 2022, we are working on a Web/Mobile application that serves as a communication and information-sharing tool, an 'intelligent' digital liaison notebook. This application has three main functionalities:

• Collect: Offer an ergonomic and accessible human-machine interface on a smartphone or tablet for the input of impressions, testimonies, and facts by caregivers, healthcare providers, visiting relatives, and possibly the senior themselves. Provide software interfaces and an automated connection mechanism to collect potential environmental data from servers (fluid consumption, weather, etc.), data from possible home automation sensors, or wearable sensors.
• Analyze and Synthesize: Reliably produce, from the collected data and its temporal tracking, a set of standard indicators of autonomy, activity, physical and mental well-being, and quality of life related to health,
• Share: Provide the individual and their family with a dashboard of the organization in place for aging well, showing its proper functioning, the state of the community, and the individual. Make available to each member of this community the information they need to be effective and that the elderly person is willing to share with them.

8.3.1 Impact of parasitoids on an invasion process

Participants: Yves Papegay, Eric Wajnberg [correspondant].

As this was explained in the report written last year, an international collaboration was developed with Israeli scientists located at the Ben-Gurion University of the Negev - published this year 16.Although the currently political situation in Israel renders this scientific connection more difficult, the work developed with these Israeli colleagues was developed further.

Rapidly, we developed a probabilistic model to understand the invasion process in brown widow spiders (Latrodectus geometricus) invading populations and habitats of black widow spiders (Latrodectus indistinctus). Both species are parasitized by a parasitoids species and it is expected that the parasitoid attack can play a significant role in the ability of the brown widow spiders to invade black widow spiders. A probabilistic model was developed to understand this more formally and to identify situations in which the parasitoid can “help” brown widow spiders to invade black widow spiders. Results obtained look promising and a manuscript is currently under preparation to be submitted in an international journal.

8.3.2 Monte-Carlo simulation models

Participants: Eric Wajnberg.

• During a 8-months sabbatical leave in Brazil (September 1, 2023 till April 30, 2024), invited by the FAPESP (i.e., « Fundação de Amparo à Pesquisa do Estado de São Paulo » ; The São Paulo Research Foundation), in the ESALQ (the biggest agronomic school all over Latin America; belonging to the University of São Paulo), an important Monte Carlo simulation model was developed to understand the spatial and temporal dynamics of the interactions between insect pests (essentially aphids) and their symbiotic bacteria that protect them, at least partially, from attacks by their parasitoids. This topic currently represents an active research area in the literature since the evolutionary mechanisms at play are still not fully understood, and the fact that hosts hosting symbionts can be protected from parasitoid attacks can lead to jeopardize potentially the efficacy of biological control programs using parasitoids as biocontrol agents. The model took into account all the main ingredients on the ecological relationships between hosts, their symbionts and their parasitoids. More specifically, both the vertical (from parents to their progeny) and horizontal (infection to neighboring hosts) transmission of the symbionts were considered, along with the spatial distribution of the aphids in a 2D environment. The cost (in terms of reproductive ability) of hosting symbionts was also considered. This model simulates the interaction between the three protagonists (the hosts, their symbionts and their parasitoids) during 100 generations and also quantifies the efficacy of the parasitoids to control their hosts (biological control efficacy). All computations were done over a computer grid distributed worldwide. The results obtained appeared to be interesting and quite innovative. The dynamics of the association between hosts and their symbionts was influenced by both the longevity of the parasitoid females and the host spatial aggregation pattern. These effects were never considered in the literature before, and can have important consequences in developing efficient pest control strategy based on parasitoids releases. These results have been published in an international journal 18.
• Another topic developed concerns the fact that pollinating bees are foraging for two types of flowers in their environments. Simple flowers (e.g., with one tubular corolla) can produce nectar and pollen that can be found easily by the bees, but in small quantities. On the other hand, complex flowers (e.g., with several inflorescences), are providing more reward, but the bees need time to learn how to exploit them efficiently. For decades, ecologists are wondering about how these two types of flower can persist in the environment, and are suspecting that the pollinating bees play an important role in this. Through a cooperative work with Israeli colleagues (from the Haifa University in the north of Israel), a detailed Monte Carlo model was developed. The goal was to simulate the bees foraging behavior in an environment in which the two types of flower can be found, over several generations, in order to see if these two types of flower can coexist in a steady-state frequency. In all the simulations, with different values of the parameters defining the features of the environment, it was indeed found that the two types of flower can coexist in a stable manner, and the pollinating bees are indeed playing a central role in this. A manuscript has been submitted to an international journal (BMC Ecology and Evolution) and it currently under evaluation.
• A simulation model was developed within the framework of the SPARCBio (São Paulo Advance Research for Biological Control) program, in Piracicaba. The goal was to develop another simulation model to optimize the releasing strategy of the egg parasitoid Trichogramma galloi (Hymenoptera: Trichogrammatidae) against the sugarcane borer Diatraea saccharalis (Lepidoptera: Crambidae). The model produced interesting results demonstrating that the pest population should be more efficiently controlled by increasing the number of weekly releases than by increasing the total number of parasitoids released 14.
• Eric Wajnberg participated to the organization of the 7th International Entomophagous Insects Conference (IEIC7), in Buenos Aires, Argentina, in 2023. Entomophagous insects, which prey on or parasitize other insects, play a pivotal role in ecosystems and are widely utilized as biocontrol agents. All the papers presented during this conference were published in a Special Issue in an international journal in the field of entomology. This Special Issue includes two review articles and nine research papers covering diverse subjects such as ecology, physiology, behavior, genetics, chemical ecology, and biological control, which reflects the diversity of topics presented and discussed during the conference. An introductory publication was produced presenting all these papers 17.

8.3.3 Other research activities developed this year

Participants: Eric Wajnberg.

Another work developed is related to the fact that, globally speaking, the world has become progressively less liberal and democratic in recent years, as can be seen from different metrics publically available. In this respect, in several democracies (e.g., Hungary, Israel, etc.) such decline in democracy had several consequences on ecological research, since it is usually associated with a loss of freedom of speech and free thinking for scientists to conduct independent researches. This is especially true for ecology, conservation, evolution and environmental sciences, disciplines that are progressively in danger. With several colleagues at the international level from Israel, USA and India, a perspective article was written to raise an alarm for the scientific community and the corresponding manuscript has been submitted to an international journal (Conservation Science and Practice). It is now at the last “minor revision” step and should be accepted for publication shortly.

9.1.1 Visits of international scientists

9.1.2 Visits to international teams

9.2.1 Other european programs/initiatives

• Hephaistos is part of the euROBIN, the Network of Excellence on AI and robotics that was launched in 2021

10.1.1 Scientific events: organisation

10.1.2 Journal

10.1.3 Editorial activities

• During 2024, Eric Wajnberg did an important editorial work that led to edit and finalize a book that has been released by December 2024 and published by the publisher Wiley-Blackwell 21. The topic of this volume is on life history evolution, and it addresses traits like body size, timing of maturation, lifespan, ageing, offspring size and number, parental care, sex allocation, social living, dispersal, etc. These different traits are discussed in plants, insects, mammals, and even humans. Also, interactions between predators and their prey, parasites and their hosts, symbionts and their hosts, plants and their herbivores, plants and their pollinators, etc. are also discussed. The book contains 24 chapter, written by a total of 48 authors coming from 17 different countries.

10.1.4 Invited talks

• Jean-Pierre Merlet gave a talk about ethics in experiment to the IEEE Young Professionals during the conference IEEE Humanoids in Nancy and presented his work on solving parametric equations to the PhD students of 3IA.
• Eric Wajnberg has given invited seminars entitled "Stochastic Individual-Based Model (IBM) in parasitoid ecology and Genetic Algorithm" at Universidade Federal de São Carlos, at Universidade de São Paulo in Ribeirao Prato, and at FCAV/UNESP, Jaboticabal in Brazil

10.1.5 Leadership within the scientific community

• Jean-Pierre Merlet is a member of the IFToMM (International Federation for the Promotion of Mechanism and Machine Science) technical Committees on History and on Computational Kinematics. He is a member of the IFToMM Executive Council Publication Advisory Board and an IEEE Fellow.
• Jean-Pierre Merlet has got an Emeritus research director position at INRIA and an emeritus senior chair of 3IA Côte d’Azur since December 2023.
• Jean-Pierre Merlet is a member of the advisory "Conseil des Sages" of GDR Robotique and participates to the GDR working group on “Frugality and sobriety” whose purpose is to reduce the ecological impact of robotics.

10.1.6 Scientific expertise

• Jean-Pierre Merlet is a nominator for the Japan’s Prize and is active in the "Robotics in Provence" industrial cluster.
• Yves Papegay is a member of the OpenMath Society, building an extensible standard for representing the semantics of mathematical objects.
• Eric Wajnberg is an appointed member of the Academic Committee of the Hebrew University of Jerusalem, an appointed member of the International Advisory Board of the “International Center for Excellence in Biological Control”.

10.1.7 Research administration

• Yves Papegay is the head of local CUMI (Committee of users of the numerical resources and tools).

10.2.1 Teaching

• Jean-Pierre Merlet was one of the animators in the Special Semester on Rigidity and Flexibility 2024 Workshop 5 on “Kinematics Aspects of Robotics”, organized by the Radon Institute for Computational and Applied Mathematics (RICAM) of the Austrian Academy of Sciences.
• Clara Thomas has taught 15 hours on parallel robots to Master ISC (M2) at University of Toulon and 64 hours on Mechanics (L2) at Polytech Nice

10.2.2 Supervision

• Jean-Pierre Merlet is supervisor of the PhD of Romain Tissot . Together with Yves Papegay , he is supervising the PhD of Clara Thomas .

10.2.3 Juries

• Eric Wajnberg has been member of a PhD defense committee at the University of Palermo (Trippa D.A., 2024/03/08)

10.3.1 Participation in Live events

• Jean-Pierre Merlet presented twice HEPHAISTOS activities to 3eme interns. He presented the HEPHAISTOS activities on handicap to INRIA general public in several occasions. He gave also two talks about the ecological drawbacks and advantages of electrical mobility (for cars and bikes).
• Clara Thomas and Jean-Pierre Merlet have participated to the Fête de la Science, exhibiting the green robot which has attracted a lot of attention.
• Eric Wajnberg has given eight general public talks in the scope of Science pour Tous 06, about "Pourquoi la sexualité ? Le regard de la biologie", about "Lutter contre les ravageurs de culture - De l'usage des pesticides vers des alternatives plus respectueuses de l'environnement" and about "Les procédures de vote de nos démocraties sont-elles objectives pour connaître la volonté du peuple ? Introduction à la théorie du choix social.".
• In the scope of Inform@thiques.fr, Yves Papegay animated a Summer School on Experimental Mathematics in July in Oxford, for high-school students. He also animated a Summer School on tools for STEAM education, for high-school French and Romanian teachers in Cluj-Napoca and an Autumn School on Data Analysis for French and Romanian high school students in Cluj-Napoca.

International journals

• 14 articleA.Adriano Gomes Garcia, E.Eric Wajnberg and J. R.José Roberto Postali Parra. Optimizing the releasing strategy used for the biological control of the sugarcane borer Diatraea saccharalis by Trichogramma galloi with computer modeling and simulation.Scientific Reports141April 2024, 9535HALDOIback to text
• 15 articleJ.-P.Jean-Pierre Merlet. Mixing neural networks, continuation and symbolic computation to solve parametric systems of non linear equations.Neural Networks1762024, 106316HALback to text
• 16 articleM.Michal Segoli, Y.Yves Papegay, T.Tamir Rozenberg and E.Eric Wajnberg. Why do predators attack parasitized prey ? Insights from a probabilistic model and a literature survey.Behavioural Processes216March 2024, 105002HALDOIback to text
• 17 articleD.Diego Segura, M. F.María Fernanda Cingolani, E.Eric Wajnberg and L.Leo Beukeboom. Entomophagous insects: Predators and parasitoids that shape insect communities and offer valuable tools for insect pest management.Entomologia Experimentalis et Applicata1726April 2024, 455-459HALDOIback to text
• 18 articleE.Eric Wajnberg and F.Fernando Cônsoli. Dynamics of Insects and Their Facultative Defensive Endosymbiotic Bacteria: A Simulation Model.Ecology and Evolution1412December 2024HALDOIback to text

International peer-reviewed conferences

• 19 inproceedingsJ.-P.Jean-Pierre Merlet. The inverse kinematics of cable-driven parallel robot with more than 6 sagging cables Part 1: from ideal to sagging cables.ARKARK 2024 - 19th International Symposium on advances in robot kinematicsLjulbjana, SloveniaSpringerJune 2024HALback to text
• 20 inproceedingsJ.-P.Jean-Pierre Merlet. The inverse kinematics of cable-driven parallel robot with more than 6 sagging cables Part 2: using neural networks.ARKARK 2024 - 19th International Symposium on advances in robot kinematicsLjlubljana, SloveniaSpringerJune 2024HALback to text

Scientific book chapters

• 21 inbookE.Eric Wajnberg. Monte Carlo Simulations to Model the Behaviour of Agricultural Pests and Their Natural Enemies.8Modelling Insect Populations in Agricultural LandscapesEntomology in FocusSpringer International PublishingFebruary 2024, 29-47HALDOIback to text

Doctoral dissertations and habilitation theses

• 22 thesisR.Romain Tissot. Use of AI for the analysis of a cable-driven parallel robot designed to assist vulnerable individuals.Universite Cote d'AzurDecember 2024HALback to text

Scientific popularization

• 23 articleJ.-P.Jean-Pierre Merlet. Une histoire de la robotique d’aujourd’hui.IntersticesNovember 2024HAL