ABOUT PEOPLE & PARTNERS PROJECTS, RESEARCH
& EDUCATION RESOURCES & EVENTS CONNECT & CONTACT

SMART’s Distinguished speakers

SMART Presentation Summary

It is often perceived that fancy optimization models and techniques are reserved for, and only applicable to mature industries with large research budgets. On the contrary, Johan Joubert argues that the state-of-the-art technologies should actually be the norm for developing countries and situations where each cent counts… where budgets are tightest… where people are most fragile, and where we chase deficit minimization as opposed to profit maximization.

An African Context

Gauteng province is the economic hub of not only South Africa, but also of Africa. It is home to nearly 10-million inhabitants, the majority of which are low-income earners living on the periphery of metropolitan areas. In developed countries the low-income earners typically live close to economic activities where there are denser public transport networks. It is not financially sustainable to have dense public transport networks and transport infrastructure on the periphery of already very large metropolitan areas, yet lack thereof hampers mobility, and excludes the already vulnerable commuters from participating in the economy.

Paratransit, or ‘minibus taxi’ as it is known in South Africa, is a communal taxi mode that is not unique to Africa, and usually evolves in the absence of more formal (usually government-funded) public transport. No more than 50% of all passenger trips are made using public transport, and of those, more than 60% are made using the paratransit mode. Traditional transport planning approaches do not take the unique land use, or paratransit issues into account, yet government relies on such traditional models to decide where to spend their scarce monies.

Better modeling for better decisions

In collaboration with the Swiss Federal Institute of Technology in Zurich, and the Technical University of Berlin, we developed an initial implementation of a large-scale state-of-the-art multi-agent transport simulator (MATSim). The premise of the agent-based approach is that each commuter (represented an agent) makes autonomous decisions. We believe that we cannot accurately describe the transit system, but rather have to rely on the aggregation of the autonomous decisions and activities of agents to observe the phenomena of traffic, and the associated congestion.

The presentation illustrated the traffic dynamics, and the measures and metrics used to validate the model. Although observing the emergence inherent in transport systems is encouraging, it still does not support decision-making.

The multimodal network design presented highlights the process of how to incorporate all modes of transport, including the paratransit minibuses, in a South African environment. The results are achieved, though, with some computational challenges. Once the proposed multimodal network is introduced in the transport simulator, one can observe the changing, and even emerging behavior of commuters. Such decision support allows commuter behavior to be given more prominence when government policies and precious investments in transport infrastructure are evaluated.

• View his PowerPoint presentation (pdf)

Biography

Johan Joubert is an South African whose ancestry dates back to 1688 when Pierre Jaubert landed in the Cape of Good Hope as a French Huguenot. Although Pierre was illiterate, Johan had the privilege of finishing his Masters and PhD in Industrial Engineering at the University of Pretoria, albeit more than 300 years later. Johan’s PhD focused on developing self-regulating and intelligent algorithms embedded in heuristics to solve the routing and scheduling of freight vehicles in constrained problems.

After a few years in both production and consulting environments, Johan returned to the University of Pretoria where he currently heads up the Optimization Group within the Department of Industrial and Systems Engineering. He is also appointed as contract researcher in the Logistics and Quantitative Methods Group of Built Environment at the Council for Scientific and Industrial Research. His objective is to ensure that models, although being wrong most of the time, are useful to decision makers, and reliable in supporting decision-making.

To practice what they preach, Johan’s Optimization Group recently succeeded in using agent-based modeling to develop a large-scale transport simulator for the whole of Gauteng, the economically most prominent province in South Africa with nearly 10-million inhabitants. The implementation of the Multi-Agent Transport Simulator (MATSim) is done in collaboration with the Swiss Federal Institute of Technology in Zurich, and the Technical University of Berlin.