Parameter estimation and wobbling behavior reproduction in lean-operable bicycle simulator

Abstract

Bicycles are regarded as one of the major risk factors for vehicles. In particular, the behavior prediction of bicycle would be a critical issue for design of autonomous driving system. The utilization of simulation is increasingly expected to efficiently and safely evaluate bicycle-related safety functions. For effective simulations incorporating bicycles, it is necessary to construct a model capable of representing cyclist behavior. Construction of accurate cyclist behavior models requires the measurement and analysis of a sufficient amount of riding data. However, acquiring extensive ride data from actual bicycles is very difficult. Consequently, there has been growing interest in employing bicycle simulators for data acquisition. In this paper, to enhance the reproducibility of bicycle behavior in a simulator while employing a simplified bicycle model, first of all, the parameter estimation is performed based on actual riding data captured using a motion capture system. In the bicycle-cyclist system, since the weight ratio of the cyclist is dominant, the parameter estimation is carried out for each cyclist individually. Second, to replicate the behavior that more closely reflects the real behavior, a model designed to reproduce the wobbling phenomenon is developed and implemented. Finally, several evaluations are performed in order to demonstrate the validity of the proposed model.