Parameterized Sensor Model and an Approach for Measuring Goodness of Robotic Maps

Map building is a classical problem in mobile and autonomous robotics, and sensor models is a way to interpret raw sensory information, especially for building maps. In this paper we propose a parameterized sensor model, and optimize map goodness with respect to these parameters. A new approach, measuring the goodness of maps without a handcrafted map of the actual environment is introduced and evaluated. Three different techniques; statistical analysis, derivative of images, and comparison of binary maps have been used as estimates of map goodness. The results show that the proposed sensor model generates better maps than a standard sensor model. However, the proposed approach of measuring goodness of maps does not improve the results as much as expected.