Numerous activity location recommendation strategies have been developed that recommend activity locations or points of interests to individuals. To the best of our knowledge, these strategies have always been designed from a user-optimal perspective. This paper contrariwise presents a recommendation strategy that aims for a system-optimal distribution of people over a set of locations. The task of generating recommendations is described as a control problem, in which the difference between a locally observed distribution of people over the locations and a desired distribution is minimized. We present a generic approach that uses feedback control laws to tackle this problem. Instead of modelling the effect of single recommendations on human behaviour, we follow a macroscopic approach and aim to determine optimal origin-specific destination choice distributions. The potential of the approach is illustrated by simulation of a simple scenario in a 3-node network.
[1]
Giacomo Albi,et al.
Invisible Control of Self-Organizing Agents Leaving Unknown Environments
,
2015,
SIAM J. Appl. Math..
[2]
Hani S. Mahmassani,et al.
Network performance under system optimal and user equilibrium dynamic assignments: Implications for
,
1993
.
[3]
Francesco Ricci,et al.
Context-Aware Recommender Systems
,
2011,
AI Mag..
[4]
Winnie Daamen,et al.
Optimal crowd evacuation
,
2013
.
[5]
Paul Lukowicz,et al.
Smart crowds in smart cities: real life, city scale deployments of a smartphone based participatory crowd management platform
,
2015,
Journal of Internet Services and Applications.