.png){kind=icon}
.png){kind=icon}
Return
Efficient Controller Design Using Bayesian Optimization for Low-Speed Motorcycles Stabilization
.png){kind=icon}
PRE
AI
DOI：10.1541/ieejjia.20250101.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper explores the control of a motorcycle using automatic tuning based on Bayesian optimization. Motorcycles exhibit high stability at high speeds owing to the gyroscopic effect. However, this effect diminishes at low speeds, forcing riders to rely on weight shifting or steering inputs to maintain balance. To improve safety in low-speed scenarios, autonomous regulation of motorcycle stability becomes essential. Existing self-balancing methods include applying a linear quadratic regulator (LQR). This approach involves creating a specialized low-speed region model using SPACAR, a finite element method-based program, and applying LQR control to achieve stabilization. However, when this approach is implemented on an actual motorcycle, unmodeled dynamics, such as tire friction, can hinder control. Therefore, gains must be fine-tuned through trial and error on actual motorcycles. This paper presents a method using Bayesian optimization for adjusting controller gains more efficiently. Through simulations and experiments, we demonstrate that Bayesian optimization can explore a wider range of parameters and find better parameters than manual tunning.
Keywords:
Autonomous vehicle
Bayesian optimization
Linear Quadratic Regulator
Motorcycle
Journal
I
IF:
1.1
Papers:
141
Citations:
679