Résumé : Wastes have a real potential as being players in the energy mix of tomorrow. They can have a high heating value depending on their composition, which makes them good candidates to be converted into liquid fuel via pyrolysis. Among the different types of wastes, automotive residues are expected to rocket due to the increasing number of cars and the tendency to build cars with more and more polymers. Moreover, the existing regulations concerning the recycling of end-of-life vehicles become more and more stringent. Unconventional fuels such as those derived from automotive shredder residues (ASR) have a particular composition which tends to increase the amount of pollutants comparing with conventional fuels. Relying on alternative combustion modes, such as reactivity controlled compression ignition (RCCI), is a solution to cope with these pollutants. In RCCI, two types of fuels are burned simultaneously, namely a light fraction with a low reactivity, and a heavy fraction with a high reactivity. The heavy fraction governs the ignition as it is injected directly in the cylinder close to the end of compression. A variation of its ignition delay could impact the quality of the combustion. Nevertheless, this issue can be tackled by adjusting the injection timing. As long as the low reactivity fuel is concerned, such a solution cannot be adopted as its reactivity depends on the initial parameters (equivalence ratio, inlet temperature, exhaust gas recirculation ratio). However, if the fuel is too reactive, it could create knock that have a dramatic impact on the engine, leading to damages. Thus, being able to predict its features is a key aspect for a safe usage. Predicting methods exist but had never been tested yet with fuels derived from automotive residues. With petroleum products, usual prediction methods stand at three different levels: the chemical composition, the properties, and the reactivity in an appliance. The fuel is studied at these three levels. First, the structure gives a good overview of the fuel auto-ignition. For instance, aromatics tend to have higher ignition delay time (IDT) than paraffins. Second, the octane numbers are good indicators of the fuel IDT and of the resistance toward knock. Precisely, the octane numbers depict the resistance of a fuel towards an end-gas auto-ignition. Last, the IDT was studied in a rapid compression machine and a surrogate fuel was formulated. Surrogate fuels substitute real fuels during simulations because real fuels cannot be modelled by kinetic mechanisms due to their complexity.The existing methods to estimate the composition were updated to predict the n-paraffin, iso-paraffin, olefin, napthene, aromatic and oxygenate(PIONAOx) fractions. A good accuracy was achieved compared with the literature. This new method requires the measurement of the specific gravity, of the distillation cut points, of the CHO atom fractions, of the kinematic viscosity and of the refractive index.Two methods to predict the octane numbers were developed based on Bayesian inference, principal component analysis (PCA) and artificial neural network (ANN). The first is a Bayesian method which modifies the pseudocomponent (PC) method. It introduces a correcting factor which corrects the existing formulation of the PC method to increase its accuracy. A precision of more than 2% is achieved. The second method is based on PCA and ANN. 41 properties are studied among which reduced set of principal variables are selected to predict the octane numbers. 10 properties calculated only with the distillation cut points, the CHO atom fraction and the specific gravity were selected to accurately predict the octane numbers.Measurements of the IDT in a rapid compression machine (RCM) of a fuel produced from ASR were realized. They are the first measurements insuch a machine ever made. This provide experimental data to the literature. Moreover, these experimental data were used to formulate a surrogate fuel. Surrogate fuels can be used to realize simulations under specific conditions. The current thesis investigates fuels derived from ASR. It was showed that this fuel can be burnt in engines as long as their properties are carefully monitored. Among others, the IDT is particularly important. Nevertheless, additional experimental campaigns and simulations in engine are required in order to correctly assess all of the combustion features of such a fuel in an engine.