Belief Space Planning for Mobile Robots With Range Sensors Using iLQG

In this work, we use iterative Linear Quadratic Gaussian (iLQG) to plan motions for a mobile robot with range sensors in belief space. We address two limitations that prevent applications of iLQG the considered robotic system. First, assumes differentiable measurement model, which is not true sensors. show only requires differentiability dynamics. propose derivative-free filter approximate dynamics, does require explicit model. Second, informative measurements from sensor are sparse. Uninformative produce trivial gradient information, optimization converging local minimum. densify by introducing additional parameters The iteratively updated ensure convergence model sensor. effectiveness proposed modifications through an ablation study. also apply method simulations large scale real world environments, superior performance comparing state-of-the-art methods either assume separation principle or maximum likelihood measurements.