A research team at Japan’s Shibaura Institute of Technology (SIT) has developed a driver-intention-aware control system for two-wheeled vehicles such as bicycles that can distinguish between cornering and possible tipping. Depending on the driving condition, stabilization support can be activated if necessary to prevent a fall.
Read more after the ad
Conventional stability control systems can only respond to vehicle movement. In the case of two-wheeled vehicles, however, there is hardly any difference between an inclination when cornering and an initial inclination when slipping. It is correspondingly difficult to recognize when a stability system should intervene and when not.
From steering movement to driving stabilization
The SIT scientists have chosen a different approach that recognizes the driver’s intention and does not rely on force feedback from vehicle movement. The researchers have developed a steer-by-wire bicycle that electronically records the driver’s steering movements and converts them into control signals for electric motor steering. The measured control signals can be used to record and evaluate the steering behavior and the interaction between the rider and the bike, as the SIT scientists write in the study “Rider-Intent-Aware Scenario-Adaptive Stabilization Control for a Steer-by-Wire Bike,” which was published in IEEE/ASME Transactions on Mechatronics. The natural steering feel is retained.
The researchers coupled the steer-by-wire system with a machine learning system to classify driver intentions. Using a “Long Short-Term Memory” network (LSTM), a machine learning model, patterns in the time-dependent control data are recognized and evaluated. Before training the system, the researchers used K-means clustering, an unsupervised learning technique, that allows them to divide the driving data into three categories: straight-ahead driving, cornering and instability.
During driving tests, the LSTM model analyzed various variables such as steering angle, vehicle speed, roll angle, lateral acceleration and reaction torque. This allows both the condition of the bicycle and the interaction between the driver and the vehicle to be recorded. The system learns to recognize driving situations in real time.
The researchers tested the system in driving tests in which they intentionally cornered and caused unstable driving conditions. The system was able to reliably distinguish between the two driving conditions, even though there were lean angles in both cases. With this information, the researchers controlled automatic stabilization support, which only intervened in unstable driving conditions to restore the vehicle’s balance and ensure control of the bike. The scientists emphasize that the natural driving behavior is not affected by the system.
Driving stability for more safety
Read more after the ad
The SIT researchers are of the opinion that such an overall system consisting of stability detection and stabilization control can ensure greater safety for bio-bikes, e-bikes and electric motorcycles. Older and less experienced two-wheeler riders in particular could benefit from such an assistance system.
The scientists want to further improve the driving assistance system that works with humans and expand it to recognize different driving situations. Different road surfaces should also be taken into account.
(olb)
