Integrated modelling of crowdsourcing for autonomous vehicles in urban networks

The upcoming era of autonomous vehicles (AVs) promises profound changes in travel behaviour and urban transportation systems. A critical new scenario involves private autonomous vehicle (PAV) owners renting their vehicles to ride-hailing platforms during idle periods, i.e., AV crowdsourcing. This pattern may significantly influence vehicle flows in ride-hailing systems, the travel behaviour of AV owners, and consequently the overall performance of the urban network. To analyse these effects, this study develops a novel network assignment model that incorporates PAV travellers’ crowdsourcing decisions. The framework accounts for both PAV travellers and shared autonomous vehicle (SAV) riders, where PAV travellers choose between self-parking or crowdsourcing their vehicles to satisfy SAV demand after reaching their destinations. The interaction between traveller decisions and the ride-hailing system, combined with routing choices across the network, constitutes a complex game-theoretical problem. We propose an integrated model, formulate the equilibrium conditions, prove its existence, and develop an algorithm based on the method of successive averages (MSA) to solve it. Numerical analyses are conducted on the Sioux Falls and Chicago networks to reveal the potential network-wide impacts of AV crowdsourcing. The results show that AV crowdsourcing offers a promising alternative for travellers and stimulates travel demand, with only a slight increase in congestion in the system, as it partially reduces the flow of self-parking PAVs and relocation of SAVs. In addition, AV crowdsourcing shows a spatially heterogeneous influence on congestion. It primarily increases congestion on links originating from regions with high net departure rates while reducing flow in the opposite direction.