Search

Las futuras misiones de exploración planetaria exigen cada vez más autonomía, ya que estas misiones son cada vez más complejas. Un ejemplo claro es la misión de retorno de muestras a Marte, en la que el Sample Fetch Rover debe recoger tubos de muestras en un lugar remoto y llevarlos de vuelta a la estación base para lanzarlos a la Tierra. Esta misión requiere ampliar las capacidades autónomas a bordo. En primer lugar, el componente de navegación debe ser capaz de detectar y localizar los tubos de muestra, y en segundo lugar, los de guiado y control deben situar el rover cerca de los tubos de muestra y mover el manipulador para recogerlos. Estas son las principales aportaciones de este trabajo. La primera cuestión se ha resuelto mediante el uso de Redes Neuronales Profundas, que permiten identificar los tubos de muestra previamente entrenados en imágenes, y la segunda se ha resuelto ampliando el algoritmo de planificación de trayectorias dentro del componente de Guiado. Para demostrar y validar los métodos propuestos, se han realizado dos experimentos. Una primera prueba de campo en el terreno experimental de Búsqueda y Rescate de la Universidad de Málaga, y una segunda prueba de laboratorio en el Laboratorio de Robótica Planetaria de la Agencia Espacial Europea. Ambos experimentos se llevaron a cabo utilizando el Rover de Pruebas ExoMars, propiedad de esta última institución. Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech.

Future planetary exploration missions are demanding more and more autonomy since these missions are getting more complex. A clear example is the Mars Sample Return mission, where the Sample Fetch Rover needs to collect sample tubes on a remote location, and bring them back to the base station to be launched to Earth. This mission requires to extend the autonomous capabilities onboard. First, the Navigation component needs to be able to detect and locate the sample tubes, and second, the Guidance and Control ones require to place the rover close the sample tubes and move the manipulator to pick them up. These are the main contributions of this paper. The first issue has been solved by the use of Deep Neural Networks, which allow to identify the previously trained sample tubes on images, and the second one has been solved by extending the path planning algorithm within the Guidance component. To demonstrate and validate the proposed methods, two experiments were carried out. A first field test in the Search and Rescue experimental terrain at the University of Malaga, and a second lab test in the Planetary Robotics Lab at the European Space Agency. Both experiments were carried out using the ExoMars Testing Rover owned by the last institution. Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech

Accurate and complete terrain maps enhance the awareness of autonomous robots and enable safe and optimal path planning. Rocks and topography often create occlusions and lead to missing elevation information in Digital Elevation Maps (DEMs). Currently, mostly traditional inpainting techniques based on diffusion or patch-matching are used by autonomous mobile robots to fill-in incomplete DEMs. These methods cannot leverage the high-level terrain characteristics and the geometric constraints of line of sight we humans use intuitively to predict occluded areas. We propose to use neural networks to reconstruct the occluded areas in DEMs. We introduce a self-supervised learning approach capable of training on real-world data without a need for ground-truth information. We accomplish this by adding artificial occlusion to the incomplete elevation maps constructed on a real robot by performing ray casting. We evaluate our self-supervised learning approach on several real-world datasets which were recorded during autonomous exploration of both structured and unstructured terrain with a legged robot, and additionally in a planetary scenario on Lunar analogue terrain. We state a significant improvement compared to the Telea and Navier-Stokes baseline.

International audience

Rovers operating on Mars are in need of more and more autonomous features to ful ll their challenging mission requirements. However, the inherent constraints of space systems make the implementation of complex algorithms an expensive and difficult task. In this paper we propose a control architecture for autonomous navigation. Efficient implementations of autonomous features are built on top of the current ExoMars navigation method, enhancing the safety and traversing capabilities of the rover. These features allow the rover to detect and avoid hazards and perform long traverses by following a roughly safe path planned by operators on ground. The control architecture implementing the proposed navigation mode has been tested during a field test campaign on a planetary analogue terrain. The experiments evaluated the proposed approach, autonomously completing two long traverses while avoiding hazards. The approach only relies on the optical Localization Cameras stereobench, a sensor that is found in all rovers launched so far, and potentially allows for computationally inexpensive long-range autonomous navigation in terrains of medium difficulty.

This paper introduces a motion planning strategy aimed at the coordination of a rover and manipulator. The main purpose is to fetch samples of scientific interest that could be placed on difficult locations, requiring to maximize the workspace of the combined system. In order to validate this strategy, a simulation environment has been built, based on the VORTEX Studio platform. A virtual model of the ExoTer rover prototype, owned by the European Space Agency, has been used together with the same robot control software. Finally, we show in this paper the benefits of validating the proposed strategy on simulation, prior to its future use on the real experimental rover. Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech

Autonomy on rovers is desirable in order to extend the traversed distance, and therefore, maximize the number of places visited during the mission. It depends heavily on the information that is available for the traversed surface on other planet. This information may come from the vehicle’s sensors as well as from orbital images. Besides, future exploration missions may consider the use of reconfigurable rovers, which are able to execute multiple locomotion modes to better adapt to different terrains. With these considerations, a path planning algorithm based on the Fast Marching Method is proposed. Environment information coming from different sources is used on a grid formed by two layers. First, the Global Layer with a low resolution, but high extension is used to compute the overall path connecting the rover and the desired goal, using a cost function that takes advantage of the rover locomotion modes. Second, the Local Layer with higher resolution but limited distance is used where the path is dynamically repaired because of obstacle presence. Finally, we show simulation and field test results based on several reconfigurable and non-reconfigurable rover prototypes and a experimental terrain.

Planetary exploration vehicles (rovers) can encounter with a great variety of situations. Most of them are related to the terrain, which can cause the end of the mission if these vehicles are not able to traverse it. It was the case of Spirit rover, which got stuck in loose sand, making it impossible to continue advancing. A solution to this is to make rovers capable of modifying their locomotion to traverse terrains with particular terramechanic parameters. Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech.

The paper in hand presents a mobile testbed –namely the Heavy Duty Planetary Rover (HDPR)– that was designed and constructed at the Automation and Robotics Laboratories (ARL) of the European Space Agency to fulfill the lab’s internal needs in the context of long range rover exploration as well as in order to provide the means to perform in situ testing of novel algorithms. We designed a rover that: a) is able to reliably perform long range routes, and b) carries an abundant of sensors (both current rover technology and futuristic ones). The testbed includes all the additional hardware and software (i.e. ground control station, UAV, networking, mobile power) to allow the prompt deployment on the field. The reader can find in the paper the description of the system as well as a report on our experiences during our first experiments with the testbed.

This paper presents the first publicly available dataset for Close Range On-Orbit Servicing Computer Vision (CROOS-CV) intended for testing and benchmarking of computer vision algorithms. It is an representative image dataset for CROOS operations with distances of 2 m between servicer and client satellite that was recorded under illumination conditions similar to a Low Earth Orbit. A training set with 180 trajectories and a test set with 810 trajectories are provided. Both were recorded at three different sun incidence angles and with multiple different shutter times. Each trajectory consist of stereo image pairs along with the ground truth pose of the cameras. Additionally, a 3D model of the client and all calibration data is provided with the dataset. The paper provides details about the recording setup, the calibration and recording procedure. Results from tests with a visual tracking algorithm are provided. The dataset is available online at http://rmc.dlr.de/rm/en/staff/martin.lingenauber/crooscv-dataset.