ActiSurTT

The increasing requirements as well as the growing pressure on production activities has led the machine used in off-road conditions (such as public works or agriculture) more and more complex. In addition, the velocity of such vehicles, acting on irregular grounds increases irremediably. Moreover, bad grip conditions encountered in the considered context, appear to be variable (due to ground nature), as well as vehicle parameters (i.e. on loaded implements or ballasting masses). All these specificities tend to make risks related to loss of control substantially increase. As a consequence, frequency and seriousness of accidents in this field of area are always too high. If the driver education and responsibility is an important factor for risk reduction, the development of assistance devices constitutes a promising alternative, still poorly investigate in off-road applications. Such devices have been largely and successfully designed and applied in on-road conditions. Off-road applications could hardly take benefit of these innovations. Nevertheless, the numerous encountered situations and conditions as well as their variability decrease the efficiency of systems such as ABS or ESP. Moreover, situations leading to accidents are also quite different (mainly loss of trajectory control in on-road conditions against loss of stability in off-road). The proposed ActiSurTT project aims at designing new driver assistance systems dedicated to off-road vehicles, enabling to anticipate and then reduce the risk of accident. Situations of accident particularly addressed in the frame of that project will be the rollover (longitudinal and lateral) and the control loss due to bad grip condition (leading to vehicle drift, especially in slope). Because of uncertainty and variability of dynamics encountered in the framework of that project, a first objective lies in the accurate description and knowledge of the machine motion in real time. Based on both reconstruction algorithms and the design of a low cost perception system using innovative sensors, adaptive models to characterize uncertain dynamics will be developed. This will enable the estimation of behaviour in real time with respect to the context and interaction with the environment. Thanks to this knowledge, the design of algorithms for risk estimation and anticipation will be designed. Such a prediction, gathered with the study of the driver behaviour will allow the definition of warning systems easily understandable. In order to be fully applicable, such devices have also to warranty their integrity (accuracy, fault detection...). This will be investigated by a sensitivity study coupled with probabilistic reliability algorithms. Above driver warnings, the risks detection and anticipation will also enable the synthesis of autonomous correction law, acting in relationship with the driver on actuators, enabling to maintain vehicle in a safe domain. This project, scheduled on 36 months, will gather the scientific and technical contributions around an experimental vehicle, enabling to test and evaluate the benefits of the device developed. This experimental vehicle will be derived from an actual agriculture marketed machine particularly facing accident risks. The synergy between laboratories and manufacturers involved in that project will favoured the efficiency and applicability in middle term of the project contributions on "off road" vehicles.