Influence of Vehicle Model Complexity in Autonomous Emergency Manoeuvre Planning
In this paper the effectiveness of an optimal reference manoeuvre is analysed w.r.t. the complexity of the vehicle model used within the optimal control algorithm. The optimal reference manoeuvre is computed by means of a Nonlinear Receding Horizon planning (NRHP) strategy which is based on a simplified vehicle model. The reference manoeuvre is tracked by a controller implemented on a low level faster loop. The system is able to perform autonomously lane change and obstacle avoidance manoeuvres by tracking the computed reference one. The quality of the performed manoeuvres depends on the reference manoeuvre and consequently on the vehicle model used by the NRHP. For manoeuvres with low or mild lateral accelerations reduced order models might yield realistic and reliable reference manoeuvres. However, critical conditions (e.g. evasive manoeuvre) require a manoeuvre planner able to catch highly non-linear vehicle dynamics that characterizes such situations. On the other hand, being the NRHP computational cost generally high and related to the number of equations of the mathematical model, a trade-off between computational efficiency and model complexity is required. The work analyses the reference manoeuvres produced by two vehicle models of increasing complexity used as reference within the NRHP. Optimal planner performance evaluation on evasive manoeuvre in critical conditions will be presented with simulations results.