Developing a representative driving cycle for paratransit that reflects measured data transients: Case study in Stellenbosch, South Africa

Paratransit plays a critical role in meeting transportation needs in many cities in sub-Saharan Africa (SSA). However, it faces deep issues related to pollution, congestion, and safety. Understanding the driving patterns of paratransit in SSA can provide valuable insights into the transportation needs in the region, which is particularly relevant nowadays given the increasing focus on sustainable transportation solutions in Africa. Representative driving cycles, which provide a realistic simulation of the driving conditions a vehicle is likely to encounter, are key to framing policies for effective transportation management, vehicle design, and urban and regional planning. However, cycle development has been limited in SSA due to a lack of data and standardized testing procedures. This study develops a representative driving cycle using GPS data gathered on paratransit vehicles traveling around Stellenbosch, South Africa, providing a benchmark for evaluation and a platform for further research and testing in SSA's dominant transport industry. A novel time series shape-based clustering methodology is employed that combines dynamic time warping and mixed integer programming to cluster micro-trips of varying length based on their time series shapes. Representative micro-trips from each cluster are stitched together with a maximum likelihood approach to curate the final cycle. By including transients from the measured data in cycle development, this novel approach to cycle development is particularly suited for capturing the notoriously unconventional and aggressive driving style of paratransit. The constructed cycle and several international cycles are assessed against the measured database on the basis of eight characteristic kinematic parameters. The constructed cycle emerges as the most fitting choice to represent paratransit operating conditions, with an average deviance of 3.65% across the parameters, compared to deviations of 23%-34% for the international cycles.