In this work, we research and evaluate end-to-end learning of monocular semantic-metric occupancy grid mapping from weak binocular ground truth. The network learns to predict four classes, as well as a camera to bird's eye view mapping. At the core, it utilizes a variational encoder-decoder network that encodes the front-view visual information of the driving scene and subsequently decodes it into a 2-D top-view Cartesian coordinate system. The evaluations on Cityscapes show that the end-toend learning of semantic-metric occupancy grids outperforms the deterministic mapping approach with flat-plane assumption by more than 12% mean IoU. Furthermore, we show that the variational sampling with a relatively small embedding vector brings robustness against vehicle dynamic perturbations, and generalizability for unseen KITTI data. Our network achieves real-time inference rates of approx. 35 Hz for an input image with a resolution of 256×512 pixels and an output map with 64×64 occupancy grid cells using a Titan V GPU.

Optical incremental encoders are extensively used for position measurements in motion systems. The position measurements suffer from quantization errors. Velocity and acceleration estimations obtained by numerical differentiation largely amplify the quantization errors. In this paper, the time stamping concept is used to obtain more accurate position, velocity and acceleration estimations. Time stamping makes use of stored events, consisting of the encoder counts and their time instants, captured at a high resolution clock. Encoder imperfections and the limited resolution of the capturing rate of the encoder events result in errors in the estimations. In this paper, we propose a method to extend the observation interval of the stored encoder events using a skip operation. Experiments on a motion system show that the velocity estimation is improved by 54% and the acceleration estimation by 92%.

No abstract

No abstract

No abstract