Deep Terrain Estimation for Planetary Rovers
Fabio Vulpi, Annalisa Milella, Florian Cordes, Raúl Domínguez, Giulio Reina
In 15th International Symposium on Artificial Intelligence, Robotics and Automation in Space (i-SAIRAS'20), (iSAIRAS-2020), 19.10.2020, Online-Conference, o.A., Oct/2020.
This research is developed within the ADE (Autono-mous DEcision making in Very Long Traverses) pro-ject funded by the European Union to develop novel technologies for future space robotics missions. ADE’s objective is to increase the range of traveled distance of a planetary exploration rover up to 1 km/sol, while ensuring at the same time optimal scien-tific data return. In this context, the ability to sense and classify the type of traversed surface plays a critical role. The paper presents a terrain classifier that is based on the measurements of motion states and wheel forces and torques to predict characteristics relevant for locomotion using machine and deep learning algo-rithms.
The proposed approach is tested and demonstrated in the field using the SherpaTT rover, that uses an active suspension system to adapt to terrain unevenness.