Interactive multi-objective vehicle routing via GA-based dynamic programming

This research is focused on dealing with multiple conflicting objectives in vehicle scheduling problem in an urban delivery context. The distinctive feature of this research is that an interactive reference point approach is applied to support the trade-off analysis of multiple conflicting objectives, including the minimization of total time, distance and CO2 emissions in the context of time-varying congestion data, derived as time-dependent historical average travel speed data from the UK road network. Due to congestion, average travel speeds on different roads change throughout the day and optimal routes may differ across time slots. Because of the conflict among the objectives, a solution that is optimal for one objective may or may not be optimal for other objectives. A hybrid algorithm mixing dynamic programming with an evolutionary algorithm is first developed to generate sets of efficient solutions. The proposed interactive approach is then described, which alternates between solution generation and preference elicitation from a decision maker. The effectiveness of the approach is illustrated using a case study that combines synthetic demand data for a company with the actual road and congestion information in the UK. The results obtained using the proposed interactive approach are compared to those obtained when a single objective is optimised.