When to stop and when to skip: Public transport modelling during Covid-19

Uni­ver­si­ty of Twente Pro­fes­sor, Kostas Gkiot­sali­tis, shares his soft­ware mod­els for mod­i­fy­ing the pub­lic trans­port demand dur­ing the Covid-19 pan­dem­ic.

Kostas Gkiot­sali­tis is an Assis­tant Pro­fes­sor with­in the Trans­port Engi­neer­ing and Man­age­ment group at the Uni­ver­si­ty of Twente. His research focus­es on pub­lic trans­port mod­el­ling and traf­fic demand pre­dic­tion. 

After study­ing a Master’s degree in Trans­port and Sus­tain­able Devel­op­ment at Impe­r­i­al Col­lege in Lon­don, Kostas joined the Intel­li­gent Trans­port Sys­tems group as a Research Sci­en­tist at NEC Lab­o­ra­to­ries Europe in Hei­del­berg, Ger­many.

Dur­ing his time at NEC, Kostas was part of three multi­na­tion­al Sev­enth Frame­work Pro­gramme (FP7) projects and a NEC Smart Pub­lic Trans­port project that fund­ed the devel­op­ment of ‘REFLEX’, a deci­sion sup­port tool for pub­lic trans­port plan­ning in Asian cities.

The road to reliable transport

“REFLEX is a soft­ware mod­el that col­lects real-time data about the local traf­fic and pro­vides bus dri­vers with instruc­tions on how to dri­ve accord­ing to the demand for pub­lic trans­port, the ongo­ing traf­fic con­ges­tion, and the loca­tion of oth­er bus­es in the area,” explains Kostas. 

The tool was imple­ment­ed in Sin­ga­pore, Hong Kong and India to pro­vide a reli­able bus ser­vice for pas­sen­gers. The tool inter­face can tell a bus dri­ver when to stop and wait at a con­trol point; when to skip a stop and advance along the route; when to sched­ule a bus to leave the sta­tion; and, when to sus­pend the bus ser­vice.

Kostas believes in the need for com­pu­ta­tion­al mod­els to main­tain the reg­u­lar­i­ty of the bus ser­vice and reduce the wait­ing times for pas­sen­gers.

“Bus­es are a crit­i­cal part of pub­lic trans­port in cities. There are mil­lions of bus rides each day, thou­sands of vehi­cles in oper­a­tion, and often the ser­vice is high-fre­quen­cy with bus­es run­ning along lines every two to three min­utes,” say Kostas.  

He explains that traf­fic con­ges­tion caus­es many prob­lems such as ‘bus bunch­ing’ and over­crowd­ing. 

“Bus bunch­ing occurs if bus­es don’t main­tain even head­ways between con­sec­u­tive trips, result­ing in more than one bus arriv­ing at the same stop at the same time. The first bus is over­crowd­ed with pas­sen­gers which delays the bus; the fol­low­ing bus­es are emp­ti­er, lighter and, there­fore, advance ahead of sched­ule.”

In sev­er­al cities, trans­port author­i­ties pro­vide mon­e­tary incen­tives (i.e. month­ly bonus­es) to bus oper­a­tors that deliv­er a reg­u­lar and reli­able bus ser­vice, as well as penal­ties for those who leave pas­sen­gers wait­ing longer than the esti­mat­ed wait­ing time.

The coor­di­na­tion of pub­lic trans­port is a com­plex busi­ness and com­pu­ta­tion­al mod­el­ling task that has become some­what essen­tial for a smooth-run­ning oper­a­tion. 

Public transport modelling during Covid-19

Kostas explains that pub­lic trans­port mod­el­ling has been par­tic­u­lar­ly impor­tant dur­ing the Covid-19 pan­dem­ic in order to min­imise the spread of the virus and mit­i­gate the eco­nom­ic loss incurred by trans­port com­pa­nies. 

“Crowd­ed pub­lic trans­port is con­sid­ered one of the virus trans­mis­sion fac­tors. Since the pan­dem­ic hit many trans­port com­pa­nies have closed sta­tions, can­celled cer­tain lines, lim­it­ed their ser­vice hours, altered the fre­quen­cy and routes of trans­port, and imposed vehi­cle capac­i­ty lim­its, leav­ing a sig­nif­i­cant num­ber of unserved pas­sen­gers.”

He con­tin­ues, “The reduced demand for pub­lic trans­port dur­ing the pan­dem­ic, in con­junc­tion with the finan­cial penal­ties, has jeop­ar­dized the future of many trans­port busi­ness­es and many are con­cerned for their eco­nom­ic loss­es.”

Kostas received a fund from The Nether­lands Organ­i­sa­tion for Health Research and Devel­op­ment (Zon­Mw) to devel­op a mod­el for mod­i­fy­ing pub­lic trans­port ser­vice pat­terns to account for the imposed Covid-19 capac­i­ty. 

His Python-based mod­el, that’s based on bus line 9 con­nect­ing the Uni­ver­si­ty of Twente with Hen­ge­lo and Enschede, aims to improve crowd­ing lev­els inside vehi­cles and reduce the pas­sen­ger wait time. 

Using OV-Chip­kaart data from Dutch pub­lic trans­port ser­vice provider, Keo­lis, the mod­el esti­mates the num­ber of pas­sen­gers wait­ing for a bus and the fre­quen­cy of bus­es trav­el­ling along the line so that adjust­ments can be made to ensure that bus­es are not over­crowd­ed. 

“In the morn­ing, when the demand for pub­lic trans­port is high­er, trans­port com­pa­nies can sched­ule more bus­es to leave the ter­mi­nal. At times when the demand is low­er, com­pa­nies can sus­pend bus­es so that they do not lose mon­ey,” he remarks. 

Sentiments about sharing software 

Kostas uses the ver­sion con­trol soft­ware man­age­ment sys­tem, GitHub, for man­ag­ing his soft­ware code dur­ing active research projects and val­ues the oppor­tu­ni­ty of being able to make it avail­able for reuse in the 4TU.ResearchData repos­i­to­ry. 

“I’ve pub­lished the soft­ware code that under­lies my peer-reviewed pub­li­ca­tion as well as doc­u­men­ta­tion about the mod­el so that oth­ers can under­stand and reuse it,” says Kostas.

“I’ve worked for com­mer­cial enter­pris­es in the past where it wasn’t pos­si­ble to share my code. Now I’m work­ing for the Uni­ver­si­ty of Twente, I realise how impor­tant these mod­els are for wider soci­ety and I want oth­ers to be able to build upon my work and improve it.”

He also explains that hav­ing a sta­ble URL for his research out­puts is impor­tant to him. “I like that my data and code are assigned with a DOI and can be pre­served long-term in a repos­i­to­ry that has a rep­utable iden­ti­ty with­in the Nether­lands.”

Planning ahead

Kostas’s next project involves work­ing on a Joint Pro­gram­ming Ini­tia­tives col­lab­o­ra­tion with six uni­ver­si­ties in the Nether­lands, Aus­tria, Italy, Turkey, Bel­gium and Ger­many to study the impact of mobil­i­ty hubs on pub­lic trans­port. For exam­ple, this could be a train sta­tion that inte­grates trains, bus­es, shared cars, shared bicycles/e‑scooters and oth­er modes of trans­port. 

Kostas is cur­rent­ly work­ing with Uni­ver­si­ty of Twente data stew­ard, Simone Fricke, to coor­di­nate the data man­age­ment of this inter­na­tion­al Euro­pean con­sor­tium, and to plan for data pub­li­ca­tion at the end of the project.  

We thank Kostas for shar­ing his research sto­ry with us and for mak­ing his research soft­ware pub­licly avail­able online. 

Please con­sult this guide for more infor­ma­tion about con­nect­ing your GitHub and 4TU.ResearchData accounts or con­tact us for assis­tance.

Writ­ten by Con­nie Clare (4TU.ResearchData)
Cov­er image illus­trat­ed by Con­nie Clare (4TU.ResearchData)
Fea­tured image by trav­elpho­tog­ra­ph­er from Pix­abay