1: \begin{abstract}

2:   In this work, we use iterative Linear Quadratic Gaussian (iLQG) to plan motions for a mobile robot with range sensors in belief space.

3:   We address two limitations that prevent applications of iLQG to the considered robotic system.

4:   First, iLQG assumes a differentiable measurement model, which is not true for range sensors.

5:   We show that iLQG only requires the differentiability of the belief dynamics.

6:   We propose to use a derivative-free filter to approximate the belief dynamics, which does not require explicit differentiability of the measurement model.

7:   Second, informative measurements from a range sensor are sparse.

8:   Uninformative measurements produce trivial gradient information, which prevent iLQG optimization from converging to a local minimum.

9:   We densify the informative measurements by introducing additional parameters in the measurement model.

10:   The parameters are iteratively updated in the optimization to ensure convergence to the true measurement model of a range sensor.

11:   %We show the effectiveness of the proposed modifications through comparisons to variants with either of the modification removed.

12:   We show the effectiveness of the proposed modifications through an ablation study.

13:   We also apply the proposed method in simulations of large scale real world environments, which show superior performance comparing to the state-of-the-art methods that either assume the separation principle or maximum likelihood measurements.

14: \end{abstract}

15: