A game theoretical model to examine pedestrian behaviour and safety on unsignalised slip lanes using AI-based video analytics.
Journal:
Accident; analysis and prevention
PMID:
40250019
Abstract
Left-turn slip lanes, also known as channelised right-turn lanes in right-hand driving countries, are widely implemented to facilitate left-turning at signalised intersections. However, pedestrian safety on slip lanes is not well known. At unsignalised crosswalks, the joint decision-making process of both pedestrians and motorists is complex, involving joint communication dynamics, yet current research primarily focuses on examining individual decisions. This study proposes a game theory-based approach, formulating the interaction as a two-player, non-cooperative, simultaneous game to examine those joint decision-making dynamics, their resulting behaviours, and associated crash risks. The approach compares two well-known equilibriums of game theory, namely Quantal Response Equilibrium (QRE) and Nash Equilibrium (NE), based on real-world pedestrian-motorist interaction video data collected over two days, each day spanning 12 h, from two slip lanes with zebra crossings at signalised intersections in Brisbane, Australia. Artificial intelligence-based video analytics extracted interaction data, which was modelled using binary logit models to understand the crossing and yielding decisions of pedestrians. Results demonstrate that the QRE outperforms the NE in predicting the crossing intention of pedestrians and the yielding intention of motorists. Results indicate that a) motorists are less likely to yield to pedestrians when the vehicle speed is higher, b) pedestrians are more likely to let the motorists go first if they start crossing from the curbside of the road, and c) motorists are more likely to yield to pedestrians when the distance between pedestrians and vehicles is longer. According to the QRE model, the probability of conflict is 5.8%, indicating that 5.8% of pedestrian interactions with vehicles in these slip lanes result in conflicts. Similarly, the confusion probability indicates that about 5.2% of pedestrians were confused about initiating crossing in the presence of zebra crossing even though drivers yielded them on slip lanes at signalised intersections. This study highlights the significance of using game theory-based approaches to understand road users' behaviour on slip lanes. These findings can help to identify pedestrian crossing intentions and support connected automated vehicles in making stopping decisions to enhance pedestrian safety and reduce potential conflicts with other road users.