Time Use and Travel Behaviour with Automated Vehicles

Abstract

Automated vehicles (AVs) have been a dream for a long time. From science fiction in the 1930s to countless prototypes, extensive road testing, and first use cases at present, the technology has clearly come a long way. So too has the vision of practitioners and academics matured to recognise the various potential benefits (e.g., accessibility, traffic safety, productivity, well-being) as well as threats (e.g., safety and security risks, induced travel demand, urban sprawl) of automation. The task at hand is to comprehensively assess these impacts in preparation for the AV future. In order to perform such assessment, the analyst needs to anticipate the travel behaviour and aggregate travel patterns of the future AV users. This is not a trivial task: letting go of the steering wheel may mean more than making travel more pleasant for some travellers (or perhaps less so for others who prefer to stay in control). For current car drivers, this may mean gained time and energy in a day that could let them re-optimise their activity schedule. For instance, they may choose to perform work tasks during commute, and spend less time at work as a result. That would let them increase the time spent – and potentially, trips made – for leisure. The schedule changes may be even larger for those who may become new car users with the introduction of AVs. In the aggregate, such individual-level transitions will likely form complex and significant trends in the transport system, in terms of, for example, changing person- and vehicle-kilometres, modal split, spatial and temporal distribution of travel demand and land-use patterns. How can the policy makers anticipate such complex developments? The answer to such queries has, for a long time, been provided by the coupling of travel behaviour and (large-scale) transport models. However, these models have so far been developed, successfully applied and fine-tuned for predicting travel patterns with the current, non-AV travel modes. The question that needs to be answered before using them to predict transport system developments with AVs is evident: can they reliably describe the travel behaviour of future AV users? This PhD is, for the largest part, inspired by my conviction that the answer to this question is ‘no’. In particular, I argue that the time-use dimension of travel demand models – that is, the effects of time-use in AVs on daily time-use – has not been sufficiently developed. Even state-of-the-art models commonly assume that on-board activities in AVs will lower the so-called travel time penalty or the value of travel time. In the prediction context, this inevitably leads to a prediction of more person- (and vehicle-) travel. In the evaluation context, this approach gives an illusion that the benefits from travel time savings will accrue gradually and not step-wise, due to, for example, discrete schedule re-arrangements. A simplified modelling approach such as this can bias the predictions of aggregate travel patterns, which can lead to misguided policy decisions. This thesis aims to narrow the gap between the expected travel and time-use behaviour of AV users on the one hand and the models that describe it on the other. Throughout the chapters, it, first, provides intuition that such gap indeed exists. Second, it analyses empirical evidence that partially supports this intuition. Third, it develops three time-use and travel behaviour models that incorporate some of the missing behavioural elements. Lastly, this thesis provides first insights into how these model updates make a difference for the predictions of aggregate travel patterns – a crucial input for transport policy making for the AV era.