Design of pattern reconfigurable sparse arrays based on multi-task learning

The pattern reconfigurable antenna array can dynamically alter the array pattern in need and the sparse arrays benefit the array design in expense and complexity. A novel method based on multi-task learning is proposed for the optimal design of pattern reconfigurable sparse array antennas in view of the minimum number of elements and pattern matching as perfect as possible. The design of sparse and reconfigurable antenna array is reformulated as an equivalent problem of multi-matrices linear regression, and the iterative shrinkage threshold method for multi-task learning is exploited to achieve the compromise between the array sparseness and pattern matching. Simulation results demonstrate that multi-pattern reconfigurations can be realized with the sparse layout deduced from the proposed method.