Multiagent model simulation : Investigation of self-restraint measures #2

Research and Development by

Yukio OHSAWA, School of Engineering, The University of Tokyo

Corresponding Research Area

Simulation and designing countermeasures against possible COVID-19 resurgence: predicting spreading of infection, estimating and verifying the effectiveness of countermeasures, and predicting deman

Investigation of self-restraint measures

■Transmission model using a scale-free network (SFN) to more closely approximate the spatiotemporal constraints of urban environments

(Results as at September: before this project)
W, m0 constraints equivalent for all nodes (people): Outbreaks occur when W- m0 exceeds m0
Ohsawa Y, Tsubokura M (2020)
 Stay with your community: Bridges between clusters trigger
expansion of COVID-19. in PLOS ONE 15(12)

(Current results: Details are unpublished)
Adding a maximum constraint for individual node W values and a constraint for the ease of encountering people with closer two-dimensional distance, the trend is still the same → Population data at each geographical point can be linked

(Current results: Details are unpublished)
Positing that a decline in infections will result in lower propensity for self-restraint (5% increase across the board in probability of encountering neighbors) → the virus is not controlled and new infections are seen (close to the current transmission pattern; difficult to forecast future trends)

Courtesy of The University of Tokyo

(Project output) Preventing severe outbreaks of infection:
① Stay with Community programs. Limiting the people one encounters to the immediate community (schools, workplaces) prevents the spread of infection even without other self-restraint measures. However, there is the potential for a serious outbreak if there are individuals as a part of the society who, in the course of ordinary life, have opportunities to encounter those from outside the community (people with higher W values). Going forward, the project plans to use large-scale simulations to verify the impact of self-restraint measures.② When self-restraint is abandoned at an early stage, contacts outside the community increase, the virus does not come under control, and infections continue to be seen.

The Risk of Long-Distance Movement (Yukio Ohsawa, UT)

■Probabilistic introduction of links to far-away habitants

MethodUsing synthetic population data [provided by Prof. Tadahiko Murata @ Kansai University; Random sampling of 1% habitants of Tokyo's 23 wards & Yokohama City], peoples’ choice of moving 27 km or more (nearly the distance between Tokyo and Yokohama) by probability pdist is simulated.

What this meansConstruct a network where each person prioritizes one living far away, among contacts with 1/ pdist others. If 1/pdist = 1,000,000 for 70,000 people (randomly selected people from the real data), each person will come into contact with one long-distance person once every 14.3 weeks.

Courtesy of The University of Tokyo

The impact of long-distance travel on an infection spread:
(1) Even a low frequency of commuting is regarded as of sufficiently high risk for explosive infection spread.(2) If the frequency rises to the highest level where everyone always commutes far, the risk above will be mitigated because it will end the role of inter-cluster bridge. However, it will not return to the level of pdist = 0.

Extension to continuous changes and strength values

■Continuous network change and infection spread model

Infection rate a=0.4,N=2000,m0=10,
Contact frequency A (a1: best friend, a2: friend1, a3: friend2, a4: acquaintance, a5: unexpected contact, a6: family ) = [0.9,0.01,0.01,0.01,0,1.00]

After self-restraint, relax, and then re-self-restraint
Average cases per week =393.8 (/2000)
Drastic relaxation after self-restraint
Average cases per week =499.2 (/2000)

W (upper bound of the number of people to touch) is not given here

Courtesy of The University of Tokyo

(The Output) Even if the structural changes of the network and the link strengths are made continuous, the results so far are valid:
(1) It is unlikely that the infection will spread just by increasing the frequency of contact with neighbors.(2) If the number of contacts increases causing the network structure changes, it may develop second, third, etc. waves.