Exchanges between partners in 2014

During the third year of the project we had the following research visit exchanges:

1. Nikolaos Bekiaris-Liberis, Post-Doc at UC Berkeley, visited Inria from May 9 to June 6, 2014.

Advances of the work program

During this third year we focused on the development and the solution of the discrete-time system optimal dynamic traffic assignment (SO-DTA) problem with partial control using the discrete adjoint method. The objective being to optimally control any subset of the agents in the network such that the total congestion of all agents in the network is minimized. We adopt a flow dynamics model that is a Godunov discretization of the Lighthill-Williams-Richards (LWR) partial differential equation with a triangular flux function and a corresponding multi-commodity junction solver. Full Lagrangian paths are assumed to be known for the controllable agents, while we only assume knowledge of the aggregate split ratios for the non-controllable (selfish) agents. We solve the resulting finite horizon non-linear optimal control problem using the discrete adjoint method. Moreover, we showed how the complexity of the gradient computation in nonlinear optimal control problems can be greatly decreased by using the discrete adjoint method and exploiting the decoupled nature of the problem’s network structure, leading to efficient gradient computation methods. We demonstrated the efficiency of this method by using it to reroute vehicles in real-time in response to an accident on a 8 mile freeway stretch in California with a parallel arterial route.
Go to Demos for some examples of our results.

Joint papers resulting from the collaboration

During this year we submitted the paper

S.Samaranayake, W.Krichene, J. Reilly, M.L. Delle Monache, J.B. Lespiau, P.Goatin, A. Bayen, Discrete-time system optimal dynamic traffic assignment (SO-DTA) with partial control for horizontal queuing networks.