MUlti-Sensor monitoring System for gesture Analysis of Kayakists

Publié le
Equipe
Date de début de thèse (si connue)
01/10/2022
Lieu
ENSSAT Lannion
Unité de recherche
IRISA - UMR 6074
Description du sujet de la thèse

Hypotheses, questions asked, identification of scientific locks

There is a growing trend in many sports to monitor human physiologic function, technical skills and performance during in-situ training. Indeed, current kayaking training programs make monitoring of the athlete’s training load and performance a key concept. To do so, several performance devices and sensors are becoming more readily available that allow performance to be quantified. Recent technological advancements have enabled athletes, coaches, and researchers to track movements, workload and biomechanical indicators using sensors or cameras to monitor, optimize performance and prevent injuries (Camomilla et al., 2018). Among all biomechanical variables, kinematical and dynamical parameters are of crucial interest. Usually, such evaluations are performed  indoor on ergometers. Previous works showed that 3D human kinematics including both upper lower limbs & paddle’s motion but also dynamics including forces acting on the paddle and footrest may contribute to the overall performance (Begon et al., 2009; Begon et al., 2010). Moreover, these variables may significantly  influence the boat’s movement (e.g.  displacement, velocity, acceleration, roll, yaw, and pitch angles).

However, despite the major contribution of these previous works, assessment in laboratory conditions suffers from major drawbacks that make it difficult to apply for the field. Indeed, the measurement devices generally used in lab conditions (e.g.  marker based optoelectronic systems, force plates) makes it   impractical for mobile, real-field daily training monitoring. Scarce on-water measurements considered only  dynamical evaluations  focusing on the pressure on the foot, shaft and paddle strains (Gomes et al., 2011; Sturm et al., 2010) that were not  combined with kinematics. However, recent improvement regarding precision of Inertial Motion Units (IMU) dramatically increased (while their cost on the contrary decreased), which makes them serious candidates for gesture monitoring, even for water sports such as canoe and kayak.

An interesting monitoring system was developed in (Bonaiuto et al., 2020), including pressure sensor for feet, IMU embedded in the paddle, Bluetooth transmission within the kayak and Wi-Fi to communicate data to the trainer at a remote place. However this system, neglecting 3D athlete’s kinematics, suffers a limited range of communications with riverside, poor energy efficiency (Wi-Fi is really energy consuming) and large dimensions for main access points. Some enhancements have to be brought to have an efficient system that could achieve real-time monitoring possibilities for the coaches. Thanks to a complete kinematics chain, the latter will then be able to give direct feedbacks and advices to the kayakists to change their gesture or adapt the paddling cadence.  

Some sensors could be added to estimate the angle of the paddle at water penetration. This feature is of prime importance and directly linked to the gesture efficiency and the kayakist's developed power. Moreover, to obtain the complete kinematics chain, not only the paddle has to be equipped but also the kayakist himself with at least one sensor node on each segment. The GRANIT team of IRISA has recently developed the Zyggie platform (Courtay et al., 2018), a compact, energy efficient wireless node that could serve as a basis for this gesture monitoring. 

To be able to monitor a complete training session, a special attention has to be paid to energy. Energy managers have to be specially designed to find the best trade-off between accuracy of motion estimation and energy consumption. To deal with ranges as long as 500m or more, LPWAN (Low Power Wide Area Networks) wireless standards have to be envisioned, and as radio is the most energy consuming task in a wireless sensor node, data processing as close to the sensors as possible will be achieved.

Moreover, biomechanical analysis coupled with dynamic and kinematic measurements of the boat has never been carried out in-situ. This coupling would allow the musculoskeletal analysis of the kayaker in a real situation. The development of a motion analysis methodology based on a musculoskeletal model from embedded data would provide scientific and practical added value by providing feedbacks to the field.


Methodological and technical approaches foreseen

The first part of this PhD will be logically dedicated to the state of the art of existing monitoring systems in sport in general and more specifically in canoe and kayak. A particular attention has to be paid to the sensors used (IMU, strain gauges, GNSS) to find the best trade-off between price, accuracy and energy consumption. The communication part is of course of prime importance since it will definitely impact the lifetime of the whole system and the size of the battery. Most of current systems are limited either by their short range (e.g. Wi-Fi) or by the limited data rate, but the LPWAN family could be the adequate solution. The student will explore different standards (LoRa, 802.15.4g, 802.11ah) to find the best suited transceivers.

The global system will then be designed and optimizations will be conducted at the protocol and application levels. The PhD student will have to design a dedicated energy manager that will schedule different local communications (Bluetooth) and long range communications with the LPWAN transceiver embedded in a central node in the kayak. To avoid a large amount of data to be transmitted to the riverside, part of the processing could be conducted locally in the central node, according to the application requirements in terms of accuracy of each sensor data (data will also be stored locally during the whole training session for a posteriori analysis if needed).

One of the main advantages of MUSSAKA is that it can provide in real-time both kinematic and dynamic data. The kinematic data will be collected through a wireless body area network (WBAN) made up of IMU positioned on the lower and upper limbs. The dynamic interaction data will be obtained from force sensors positioned at the foot / boat and hand / paddle interfaces (+ seat?). 3D kinematics including body segment orientations and joint angles will be estimated using inverse kinematics and a biomechanical model. The calculation  will be conducted using the OpenSense toolbox that uses the IMU quaternions as inputs to the model and applies a sensor-to-segment calibration.  In the aim of assessing the accuracy of kinematic parameters (and more particularly joint angles) will be calculated by comparing IMU-based and optoelectronic-based motion capture in laboratory conditions. A second step will consider the 3D joint torques and forces estimation using an inverse dynamics approach. A special attention will be paid to the time synchronization between the force applied to the boat, joint movements and paddling force.  Indeed, the contribution of foot-bar force from lower-limb action significantly contributes to kayakers' paddling performance since the minimization of time delay between maximal foot-bar force occurence and  maximal paddling force  is a key factor to improve efficiency.  

All experiments will be conducted on high level kayakers whose recruitment will be facilitated by a collaboration with federal partners and local high-level staffs (Pôle France Kayak Rennes).


Positioning and scientific environment

The project will be conducted within the GRANIT and M2S laboratories. For more than ten years, the GRANIT team of IRISA has developed several lightweight IMU-based wireless platforms dedicated to indoor positioning and gesture monitoring. In Brittany, some other teams are working on this subject, but rather than IMU, they are using low-cost cameras to analyze gestures, for example in swimming.

The M2S lab, internationally renowned in the field of sports sciences including movement analysis on both experimental data and digital human models. M2S lab benefits from an expertise due to previous academic and industrial projects and numerous publications in the field of analysis of swimming. 

The collaboration between the GRANIT team of IRISA and M2S is not novel, since they were already implied in common projects around sport monitoring. A first system was developed a few years ago in collaboration with Ericsson for tennis monitoring (both gesture analysis and positioning on the tennis court) and a demo was done during international tennis open in Lannion in 2015. More recently, both teams participated in the Labex Cominlabs Moonlight project that aimed to develop a mobile, low cost and energy efficient monitoring system for cycling. A common publication on this prototype should be submitted before the end of 2021.

This project will be conducted at the interface between scientists, coaches and sport performance analysts within high-level kayaking structures. More specifically, it will benefit from the support of the French Canoe-Kayak Federation (FFCK) through its high performance analysis unit (Head: Rémi Gaspard) and the  local network including the  “Pôle France Kayak” in Rennes and the Research Innovation department in Sport and Health at Campus Sport Bretagne (Head: Rémy Masson).

This project could also imply Michele Magno, from ETH Zurich, whose research expertise totally falls within the scope of this project and who is a regular collaborator of GRANIT members.

Bibliographie

Bonaiuto V., Gatta G., Romagnoli C., Boatto P., Lanotte N. & Annino G. A pilot study on the e-Kayak system: a wireless DAQ suited for performance analysis in flatwater sprint kayaks. Sensors 20.2 (2020): 542

Courtay A., Le Gentil M., Berder O. & Carer A. Zyggie : a Wireless Body Area Network Platform for Indoor Positioning and Motion Tracking, in : IEEE International Symposium on Circuits and Systems (ISCAS), Florence, Italy, (2018).

Begon M., Colloud F. & Lacouture P. Measurement of contact forces on a kayak ergometer with a sliding footrest–seat complex. Sports Eng 11, 67–73 (2009).

Begon M., Colloud F. & Sardain P. Lower limb contribution in kayak performance: modelling, simulation and analysis. Multibody Syst Dyn 23, 387–400 (2010). 

Gomes B., Viriato N., Sanders R., Conceição F., Vilas-Boas J.P., Vaz, M. Analysis of the on-water paddling force profile of an elite kayaker. In Proceedings of the Biomechanics in Sports 29, Porto, Portugal, 27 June–1 July 2011; Volume 1.

Sturm D., Yousaf K., Eriksson M. A wireless, unobtrusive Kayak Sensor Network enabling Feedback Solutions. In Proceedings of the 2010 International Conference on Body Sensor Networks, Singapore, 7–9 June 2010.

 

Liste des encadrants et encadrantes de thèse

Nom, Prénom
BERDER Olivier
Type d'encadrement
Directeur.trice de thèse
Unité de recherche
IRISA UMR 6074
Département
Equipe

Nom, Prénom
BIDEAU Nicolas
Type d'encadrement
Co-encadrant.e
Unité de recherche
M2S

Nom, Prénom
COURTAY Antoine
Type d'encadrement
Co-encadrant.e
Unité de recherche
IRISA UMR 6074
Département
Equipe

Nom, Prénom
NICOLAS Guillaume
Type d'encadrement
Co-encadrant.e
Unité de recherche
M2S
Contact·s
Nom
BERDER Olivier
Email
olivier.berder@irisa.fr
Téléphone
+33296469345
Mots-clés
Wireless sensor networks, sport monitoring, energy efficiency, data fusion