This paper deals with the development of a computational efficient approach to approximate the solution to the Hamilton Jacobi Bellman equation. The primary focus is to generate optimal feedback controllers for nonlinear systems in higher dimensions. Solving the Hamilton Jacobi Bellman partial differential equation is known to be a computationally challenging problem due to the curse of dimensionality with the increase in dimension. A collocation based approach is adopted, where the collocation points are chosen as the recently developed Conjugate Unscented Transform points to avoid the curse of dimensionality. Further a l1-norm based optimization problem is proposed to optimally select the basis that is suitable for the given dynamical system.