• experimental validation
• human-machine interaction
• motorcycle dynamics
• rider behaviour
• road safety
• vehicle modelling

The worldwide diffusion of Powered Two-Wheelers (PTWs) poses relevant road safety problems, as they carry a higher risk than other modes of transportation despite technological advancements. Riders are more susceptible to severe injuries; tilting vehicles have complex dynamics and are generally unstable; consequently, human errors are the leading cause of PTW crashes. Improved training and re-training procedures could prevent inappropriate actions by the rider, advanced assistance systems could rectify their mistakes, and improving the vehicle design process could mitigate their impact on the dynamics.

Understanding PTWs dynamics, rider behaviour, and rider-vehicle interaction better would benefit these strategies. To this end, this research mainly aimed to investigate the dynamics of motorcycle riding, focusing on the rider-vehicle interaction particularly when negotiating a corner, and identify the most relevant limits to develop countermeasures for lack of vehicle control. The thesis employed experimental tests conducted up to ∼70 kmh−1 with sports naked and sports touring instrumented motorcycles as the primary means to gather data to be analysed to derive conclusions. Quasi-static and transient manoeuvres, concerning longitudinal, lateral, and combined dynamics, were employed.

A novel tyre model formulation was proposed, whose characterisation does not require a dedicated test bench. The model could correctly reproduce the primary behaviour of a Magic Formula model, including tyre moments and steering torque. Characterising the model through real riding data proved feasible, and its robust formulation limited the propagation of estimation errors.

Next, a simplified steering assembly model was defined to estimate the steering torque signal. The experimental data relative to three different PTWs showed good estimation accuracy. Transfer functions were derived from the model and used to predict the impact of several design parameters on the vehicle’s handling.

Riding simulators are precious to studying rider behaviour; therefore, the thesis defined an unambiguous procedure to tune the underlying model of a low-complexity simulator, obtaining realistic steering feedback. It allowed reproducing the handling of three PTWs in steady-state and transient manoeuvres.

Interpreting riding data requires an expert; additionally, the process is subjective and time-consuming. An unsupervised segmentation algorithm was defined to facilitate research employing experimental data. The tool used the statistics of the roll angle and its derivatives to subdivide the dataset into manoeuvres. Analysing the segmented trial, with a special focus on a corner entry manoeuvre, revealed the effectiveness and usefulness of the approach.

Riding behaviour was then investigated objectively, defining methods and metrics. Motorcycle dynamics and rider inputs were measured during an extensive experiment involving seven participants. Their statistics correlated well with the experience level, the effect of familiarisation, and the instruction given. Experienced riders used more intense dynamics, especially concerning the braking jerk. Analysing the rider inputs revealed a variety of riding practices.

Lastly, a preliminary assessment of emergency steering assistance systems for PTWs was performed. A pool of experts used various approaches and crash databases to evaluate three safety functions concerning their safety-improvement potential. Simulated crashes revealed the injury-reduction capability of Motorcycle Autonomous Emergency Steering (MAES). An external action was applied to a real vehicle to successfully avoid
an obstacle by swerving.

The insights, models, and methodologies developed and validated in this thesis should aid the research in the field. The tyre model formulation offers a cost-effective alternative to traditional approaches. The method to estimate the steering torque allows recreating this signal without a dedicated sensor. The tuning process for low-complexity simulators should make them easier to design and calibrate. The unsupervised segmentation algorithm can assist researchers in interpreting experimental data, and the metrics and methodologies to describe and categorise riding preferences should aid them in behavioural analysis. Lastly, the promising results concerning assessing the usefulness and feasibility of steering assistance for PTWs advise further exploration of the topic.