This paper addresses power management of wireless sensor nodes which receive their energy from solar cells. In an outdoor environment, the future available energy is estimated and used as input to a receding horizon controller. We want to maximize the utility of the sensor application given the time-varying amount of solar energy. In order to avoid real-time optimization, we precompute off-line an explicit state feedback solution. However, it is a well-known problem of the optimal feedback solution that the computational complexity grows very quickly, which is particularly unfavourable for sensor nodes. A new method to derive approximate solutions to a multiparametric linear programming problem is presented. The resulting control laws substantially reduce the on-line complexity in terms of computational and storage demand. We show that a sensor node's performance is not necessary decreased due to suboptimality of the control design.
展开▼
