Impact of take-over control mechanisms on merging operations of conditionally automated vehicles

• In automated driving, take-over control mechanism impacts driver behaviour. • This experimental driving simulation study focused on merging manoeuvres. • Steering wheel mechanism slowed and destabilized merging operations. • Button and pedal-based take-over controls led to smoother and quicker vehicle merging. • Gender influenced merging behaviour after transitions from automated to manual driving. On the path to fully autonomous road transport, Conditionally Automated Driving (CAD) provides a shared driving experience between the human driver and the automated driving system (ADS). Critical moments arise when drivers must take-over control of the vehicle from the ADS to continue the driving task or perform a specific manoeuvre that is beyond the ADS capabilities. This study investigated the effects of different Take-Over Control (TOC) mechanisms on driver behaviour and performance during a merging manoeuvre after resuming control from a CAD system. Three TOC mechanisms, (i) the steering wheel, (ii) pedals, and (iii) button, were evaluated using the driving simulation. Thirty participants completed three driving sessions on a designed test track. For each session, a single TOC mechanism was activated. Repeated-measures simulation results were analysed using Weibull Accelerated Failure Time with Shared Frailty and linear mixed-effects models. The results showed that, during merging manoeuvres, the steering wheel mechanism led to significantly longer lane-change durations than the button and exhibited lower manoeuvre quality than the pedals, likely due to increased cognitive load. Gender differences were also observed: female drivers commenced lane-change manoeuvres earlier than males but took longer to complete them. The findings offer insights into designing more effective and user-centred TOC mechanisms. In addition, the findings highlight the necessity for more comprehensive designs that consider the variabilities stemming from different ADS interfaces and driver characteristics. Future research should include a broader demographic sample and real-world investigations to further validate the findings and refine these mechanisms.