Unplanned carpooling with service guarantee in low-density areas

In France, low-density areas remain heavily car-dependent, with roughly 80 % of trips made by private car, while the cost of a bus network or of demand-responsive transport is often prohibitive for local authorities. Carpooling is a credible alternative for decarbonising these mobilities, but it faces a well-known vicious circle: few passengers show up because there are few drivers, and few drivers commit because there are few passengers. This thesis addresses an unplanned carpooling system with a service guarantee, inspired by the scheme operated by the company ECOV on major corridors leading into Lyon and Grenoble. A passenger arrives at a dedicated stop, signals the request through a mobile application or a roadside terminal, and waits for a driver to stop. The driver is paid, while the ride is free for the passenger, the cost being covered by the local authority. Neither booking nor prior planning is required. The innovative feature under study is the service guarantee: if the waiting time at the stop exceeds a threshold, a taxi is automatically dispatched to pick up the passenger at no additional cost. This guarantee is precisely intended to break the vicious circle by making the service reliable from the outset. The aim is to assess the conditions under which such a system constitutes a viable mobility solution across a low-density territory. The chosen tool is agent-based simulation, coupled with optimisation methods. The work is divided into two parts. The first concerns the simulation of the dynamic transport system, involving passengers and vehicles, an itinerary engine under uncertainty (stochastic shortest path problem), and user behaviour profiles. The second part addresses the optimisation of the taxi fleet ensuring the service guarantee, formulated as a Dynamic Dial-a-Ride Problem, whose combinatorial nature makes it particularly hard to solve. The system is analysed over three time horizons. At the operational level, agents optimise their decisions locally according to the information available in real time. At the tactical level, the thesis addresses the sizing and positioning of the taxi fleet. At the strategic level, it considers the number and location of the network's stops. The main scientific challenge lies in embedding stochastic optimisation of passenger routes and taxi tours within an agent-based simulation. This work extends research carried out at the DISP laboratory on demand-responsive transport and on the coupling of optimisation with agent-based simulation, adding the service guarantee dimension.