This paper considers joint linear processing at multi-antenna sources and one multiple-input multiple-output (MIMO) relay station for both one-way and two-way relay-assisted wireless communications. The one-way relaying is applicable in the scenario of downlink transmission by a multi-antenna base station to multiple single-antenna users with the help of one MIMO relay. In such a scenario, the objective of join linear processing is to maximize the information throughput to users. The design problem is equivalently formulated as the maximization of the worst signal-to-interference-plus-noise ratio (SINR) among all users subject to various transmission power constraints. Such a program of nonconvex objective minimization under nonconvex constraints is transformed to a canonical d.c. (difference of convex functions/sets) program of d.c. function optimization under convex constraints through nonconvex duality with zero duality gap. An efficient iterative algorithm is then applied to solve this canonical d.c program. For the scenario of using one MIMO relay to assist two sources exchanging their information in two-way relying manner, the joint linear processing aims at either minimizing the maximum mean square error (MSE) or maximizing the total information throughput of the two sources. By applying tractable optimization for the linear minimum MSE estimator and d.c. programming, an iterative algorithm is developed to solve these two optimization problems. Extensive simulation results demonstrate that the proposed methods substantially outperform previously-known joint optimization methods.

Beamforming design for multi-user wireless relay networks under the criterion of maximin information throughput is an important but also very hard optimization problem due to its nonconvex nature. The existing approach to reformulate the design as a matrix rank-one constrained optimization problem is highly inefficient. This paper exploits the d.c. (difference of two convex functions) structure of the objective function and the convex structure of the constraints in such a global optimization problem to develop efficient iterative algorithms of very low complexity to find the solutions. Both cases of concurrent and orthogonal transmissions from sources to relays are considered. Numerical results indicate that the proposed algorithms provide solutions that are very close to the upper bound on the solution of the non-orthogonal source transmissions case and are almost equal to the optimal solution of the orthogonal source transmissions case. This demonstrates the ability of the developed algorithms to locate approximations close to the global optimal solutions in a few iterations. Moreover, the proposed methods are superior to other methods in both performance and computation complexity.

This paper considers beamforming problem in relay-assisted multi-user communication and presents an efficient algorithm to solve Min-max optimization of information throughput under limited power resources. We use DC programming and penalty function method to handle the inherent non-convexity of the associated objective function and follow an iterative procedure to solve for the global optimal beamforming vector. Extensive simulations are performed to establish the effectiveness and superiority of the proposed algorithm over conventional techniques based on randomization.

This paper discusses dynamic state estimation for nonlinear measurement model through distributed multisensor network under power constraints. For this scenario, we propose an optimized power allocation strategy based on semidefinite programming, that achieves minimum mean-squared error for the estimate subject to constraints on total transmit power. System nonlinearity is handled effectively with the help of distributed unscented Kalman filtering and linear fractional transformation. Furthermore, advantage of using multiple sensors over a single independent sensor is established through simulation results for tracking a maneuvering target.

This paper presents an efficient technique for power allocation to the sensor nodes in a nonlinear sensor network (NSN). We minimize mean square error of the estimation of a random scalar parameter subject to a constraint on total amount of power consumed by the sensor nodes. This estimation is carried out at fusion center (FC) which receives the local observations from the sensors located at different positions. We convert the optimization problem into a convex one, and then use semidefinite programming to find the global optimal solution. The simulation results show that our approach outperforms the previous work both for the channel with white noise and the one with colored noise. The proposed strategy also gives better results in case of nonlinear model when compared to the strategy of assigning equal power to sensor nodes.