Estimation pour le développement de systèmes d'aide à la conduite des véhicules à deux-roues motorisés

Abstract : The road accidents investigations are unanimous : Powered Two-Wheeled Vehicles (P2WV) users are the most vulnerable on the road. Although the global number of road accidents tend to decrease for several years, the number of accidents involving motorcycles are still alarming. Unfortunately, this observation concerns many regions in the world and, even more, Asian countries where the use of motorcycle is much higher than elsewhere. Nowadays, safety of P2WV users become a major concern in our society. During the two last decades, the development of Intelligent Transport Systems (ITS) has largely contributed to decrease the number of road accidents while improving safety and comfort in our vehicles. Among the ITS, one can find the well-known Advanced Driver Assistance Systems (ADAS). The most common ADAS are the ABS, the ESP, the ACC, etc. Nevertheless, most of them are only dedicated to Powered Four-Wheeled Vehicles (P4WV). The P2WV dynamics attributes such that the large roll motion or the load transfer phenomenon do not allow to extend directly what were done for P4WV. Hence, such systems are called Advanced Rider Assistance Systems (ARAS) for P2WV. The ARAS development is a real challenge for the actors on the motorcycle market. Indeed, P2WV are generally cheaper than other motorized vehicles. Hence, it requires low cost solution. Moreover, they are more and more compact leading to difficulties for the hardware integration. Furthermore, their dynamics is very complex because of their unstable and highly nonlinear behavior. These three reasons make the ARAS development drastically more complex. This PhD takes place in this context and proposes several works addressing P2WV dynamics estimation. Indeed, the development of ARAS is based on handling indicators computed from some pertinent dynamics variables. Among the latter, some are unmeasurable like the tire forces, whereas some others require expensive sensors such as the lateral speed, etc. For both technical or economic reasons, the use of observation techniques turns out to be an adequate solution in the development of ARAS. Such techniques allow to estimate the vehicle dynamics while reducing the number of sensors and overcoming the unmeasurability of some dynamics states. A part of this PhD is dedicated to model-based observers. It means that the observers are designed using a mathematical model of the P2WV. In this context, three different observers were proposed : an unknown input observer, a nonlinear Luenberger observer and an algebraic observer. Both of them require a simple combination of sensors and take into account realistic assumption like the longitudinal speed variation. They were validated by means of the wellknown simulator BikeSim and even with experimental data. We demonstrated that these three observers have a serious potential for industrial applications. A second part of this PhD deals with vision-based estimation techniques. Indeed, a such approach allows to bypass the use of a P2WV model and all the restrictive assumptions necessary to estimates some dynamics. In a first time, we proposed an estimation and prediction algorithm to assist the rider in turn. It allows to estimate the P2WV relative lateral position and heading angle to the road. Simultaneously, it predicts the road curvature within 30 meters ahead of the vehicle. Then, we used this algorithm to propose a steering behavior characterization technique for over and under-steering detection. In a last work, we proposed an image-based observer to estimate the P2WV roll angle. All these works were validated using BikeSim and we are currently testing them with experimental data.
Document type :
Theses
Complete list of metadatas

https://hal.archives-ouvertes.fr/tel-02024804
Contributor : Frédéric Davesne <>
Submitted on : Tuesday, February 19, 2019 - 1:15:14 PM
Last modification on : Friday, May 17, 2019 - 11:37:57 AM
Long-term archiving on : Monday, May 20, 2019 - 3:53:41 PM

File

2018SACLE034.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-02024804, version 1

Citation

Pierre-Marie Damon. Estimation pour le développement de systèmes d'aide à la conduite des véhicules à deux-roues motorisés. Automatique / Robotique. Université Paris-Saclay; Université d'Evry-Val-d'Essonne, 2018. Français. ⟨NNT : 2018SACLE034⟩. ⟨tel-02024804⟩

Share

Metrics

Record views

61

Files downloads

35