1: \begin{abstract}

2: In mathematical epidemiology, epidemic control often aims at driving

3: the number of infected individuals to zero, asymptotically.

4: However, during the transitory phase, the 

5: number of infected can peak at high values. 

6: In this paper, we consider mosquito vector control in 

7: the Ross-Macdonald epidemiological model,

8: with the goal of capping the proportion of infected by dengue at the peak. 

9: % This approach differs from the widespread stationary vaccination control

10: % strategies, based upon having control reproductive number stricly less

11: % than one to ensure convergence, 

12: % and also from cost minimization optimal control ones.

13: % Indeed, instead of aiming at an equilibrium or optimizing, we look for

14: % policies able to maintain the proportion of infected individuals below to

15: % threshold for all times. Thus doing, we focus both on transitories and

16: % on asymptotics, in to robust way. 

17: We formulate this problem as one of control of a dynamical system

18: under state constraint. 

19: We allow for time-dependent fumigation rates to reduce the population

20: of mosquito vector, in order to maintain the proportion of infected 

21: individuals by dengue below a threshold for all times.

22: The so-called viability kernel is the set of initial states 

23: (mosquitoes and infected individuals) for which  

24: such a fumigation control trajectory exists.

25: Depending on whether the cap on the 

26: proportion of infected is low, high or medium,

27: we provide different expressions of the viability kernel.

28: We also characterize so-called viable policies that produce, 

29: at each time, a fumigation rate as a function of 

30: current proportions of infected humans and mosquitoes,

31: such that the proportion of infected humans remains 

32: below a threshold for all times. 

33: We provide a numerical application in the case of control

34: of a dengue outbreak in 2013 in Cali, Colombia.  

35: 

36: \end{abstract}

37: