The aim of this thesis is to investigate the variables that influence the behaviour of an autonomous vehicle in order to determine acceptable and safe ranges of variation in various typical use cases. It seeks to identify rules that enable safe, acceptable and efficient behaviour. Strict application of traffic rules with large safety margins can lead to conflicts and degradation of performance, possibly leading to deadlock situations. Experiments have shown that driving style has a strong influence on acceptability, especially in conflict situations. Prior communication about safety also improves acceptability. The results will make it possible to propose algorithmic adaptations and recommendations to manufacturers in order to improve the acceptability of autonomous vehicles.