Ground-View Event Camera-Based Velocity Estimation Enabled by Spiking Neural Networks for Ground Robots

Accurate velocity estimation plays a key role in ensuring precise motion control and safe navigation of ground robots. Ground-view event cameras provide a reliable sensing modality for this task. They exploit the static and stable nature of the ground, together with the sensor’s high temporal resolution, to extract motion cues robustly even under challenging environments such as scene dynamics. Building on this observation, this paper proposes a novel velocity estimation method for ground robots using a ground-view event camera, whose application in a ground-looking configuration has received limited attention, and leverages a spiking neural network to process event-based visual information. The neural network exploits its intrinsic temporal dynamics to infer motion directly from ground texture event streams, eliminating the need for long-term event accumulation. The recurrence of output spikes is also integrated into the network to aid in preserving past sequential information. To address the challenge of costly hyperparameter tuning associated with the conventional loss function, we propose a novel geometric loss function that automatically balances translation and rotation by utilizing the differences between the coordinates of events, warped by the predicted and ground-truth motion, as the loss, thereby leading to more balanced learning across outputs and more accurate motion estimation. Experiment results on both synthetic and real-world data demonstrate that the proposed method not only can effectively handle diverse ground surface textures and variations in the camera’s pose relative to the ground, but also generalizes well to real-world scenarios, even when the network is trained exclusively on synthetic data.