Final Report Summary - FASTER (Forward Acquisition of Soil and Terrain for Exploration Rover)
Executive Summary:
The FASTER project developed novel and innovative concepts and methods for a reliable in-situ forward evaluation of soil properties and terrain conditions along the planned trajectory of a planetary exploration rover. These methods can be used to make robotic planetary exploration missions more efficient and less risky. Today, the exploration of planetary surfaces by robots is seriously impeded by the lack of a priori detailed information about the soil- and terrain conditions in the vicinity of the exploration rover. In the best case, this lack of information leads to extremely slow and cautious exploration strategies (well below 5 cm/s). In the worst case, rovers may get stuck (temporarily or permanently) due to unforeseen terrain conditions. The recent immobilisation of the NASA MER Spirit rover, as it became trapped in hidden soft sand, dramatically illustrates this scenario.
The FASTER concept foresees a highly mobile scout rover co-operating with a planetary exploration rover (such as EXOMARS). Both are equipped with sensors to acquire soil and terrain information in-situ. The sensor data are fused and used to develop a risk model that indicates potential hazards along the planned trajectory of the exploration rover. The objectives of FASTER were thus to develop key technologies in three different areas:
• Forward-looking characterization of surface properties: FASTER developed methods and instrumentation to characterize the trafficability-properties of planetary surfaces and thus to anticipate potential hazards along the planned trajectory of a planetary exploration rover.
• Innovative locomotion system for a scout type rover: The project developed an innovative locomotion concept to ensure that the scout rover is able to operate in-sync with the exploration rover, with minimal risk of encountering hazardous situations from which it must recover.
• Collaborative operation of a mother/scout pair: FASTER explored the feasibility and performance of a mother/scout rover pair and developed methods to achieve successful collaborative and autonomous robot operation.
FASTER started on November 1st, 2011 with a project duration of 37 months. The project received funding from the European Union’s Seventh Framework Programme for Research, Technological Development and Demonstration under grant agreement no 284419.
A European consortium of five partner from four EU member states contributed to FASTER. The consortium brought together European experts in the areas of planetary science, soil and terrain trafficability analysis, planetary exploration robotics, and robot autonomy and co-operation. It encompassed research centers (DFKI GmbH., Germany), academia (University of Surrey, UK), larger industrial companies (EADS DS, UK and Astri Polska, Poland), and SMEs (Liquifer Systems Group, Austria) with proven competencies in the research areas addressed.
Project Context and Objectives:
The FASTER project was aimed at strengthening the foundation technologies in support of space science and exploration, supporting the transportation of science instrumentation to as yet unexplored parts of other solar system planets and reducing the vulnerability of exploration assets (planetary rovers). In addition, the outcomes of the project may be used to improve the productivity of space exploration whilst expanding the range of environments into which scientific instrumentation can be transported.
The project invoked and encouraged cross-national collaboration in a study that fully supports European Space Policy and exploration initiatives. The FASTER project developed innovative and feasible space robotic technology in support of Europe’s vision for exploration and possible human settlement on Mars or the Moon. The new engineering and technological solutions developed under this project support and complement current and future EU and ESA space exploration projects. The outcome of this study will contribute significantly to ESA’s upcoming space exploration missions, in particular the Mars Sample Return mission.
Strategic objectives: The outcomes of the FASTER project supports the ability of future planetary exploration robots to explore hitherto unexplored parts of solar system planets at lower cost and risk. This raises the possibility of gaining a better understanding of how planets in our Solar System are formed and evolve, thereby raising the value of current planetary modelling and resolving many unanswered questions about our universe. The utility created by FASTER technology enables to an increase in the quality and quantity of scientific returns from interplanetary missions, thereby enriching scientific knowledge and opening up the development of new scientific models and understanding. The project thus enhances Europe’s standing within the space exploration community world-wide and support collaborative and co-operative exploration with other space agencies and nations.
Technical objectives: The overall goal of the FASTER project was to address three key technology developments for planetary exploration:
• Forward looking surface properties characterisation;
• Innovative locomotion system for a scout type rover;
• Collaborative operation of a mother/scout pair.
The first focal point was to develop methods and instrumentation to characterise the properties of planetary surfaces in which robotic rover vehicles operate in order to be able to anticipate hazards in advance of locomotion and navigation over that surface. By assessing the soil trafficability for a rover, decisions affecting rover safety (avoiding becoming stuck) can be reliably taken and hazards avoided.
The second focal point was to look at what innovative locomotion principle on can be applied to a scout rover to ensure that scout is able to conduct its soil characterisation operations with minimal risk of encountering hazardous situations from which it must recover.
The third focal point was to explore the feasibility and performance of a mother / scout rover pair combination to achieve successful collaborative and autonomous operation. Both the forward looking scout rover and the exploration rover were equipped with newly developed sensor technologies to acquire soil and terrain information. This information was used to calculate a risk model that indicated the potential hazards that the terrain represents to the exploration rover.
Reducing the locomotion and traverse risks using the proposed advances in autonomous mother / scout collaboration will allow mission operators to explore planetary surfaces with increased safety. It will make possible new levels of autonomous operations by significantly reducing the greatest uncertainty factor – namely properties of the surface material on which the rover must operate. This will lead to increased operational efficiency and, when coupled with the increased richness of the surface properties data, will lead to much higher scientific returns per capital investment for each mission.
Project Results:
Objectives and Requirements
The main results of the FASTER project are concepts and technologies for an in-situ and on-the-ground evaluation the soil-properties in the vicinity of a planetary exploration rover. Although the technologies developed in FASTER can also be used for scientific exploration of planetary surfaces, the primary objective in the project was to establish a trafficability map that indicates a safe trajectory for the planetary exploration rover. The primary use-case was a mission to Mars, such as the planned ESA Exomars mission.
The study was based on a requirements evaluation that covers all aspects of the potential Martian mission, including solar radiation environment, Mars atmosphere conditions and surface regolith material, properties and features (eg. rocks, outcrops, dunes, etc.). These were used to derive system requirements for the planned robotic and sensor sub-systems. Although it was beyond the scope of the FASTER project to develop space-qualified technologies for a specific mission, the overall goal was to stay as close as possible to the requirements of a planned real ESA mission to Mars in order to ensure basic compatibility of the FASTER concept with ESA planning.
Technological developments
The main components of the system consist of two robotic platforms, a main planetary exploration rover (such as ExoMars) and a light-weight scout rover (developed in FASTER). The main sensory and software components that were developed in FASTER were distributed among these two platforms.
Mobile robotic platforms
Primary Rover: The term “Primary Rover” denotes the main robotic rover used for a planetary exploration mission. Initially, the objective was to prepare FASTER system as a direct support for ExoMars rover planned by ESA. However, after redefinition of the mission, FASTER system is now rather a technology demonstrator prepared with regards to ExoMars requirements (mass, volume, energy requirements, etc.) but in fact capable to be used as an universal safe terrain traversal supporting system for any planned rover of similar size. In the project, as the Primary Rover, existing ExoMars mockup Bridget, developed in Airbus DS, was used.
Coyote II Scout Rover: Coyote II was developed by the Robotics Innovation Lab of project partner and coordinator DFKI. It is a micro rover with high mobility performance in various terrains. Equipped with its own power source, on-board sensor suite and computer it is able to perform autonomous exploration tasks. The communication subsystem allows to co-operate with other systems and provides a link for remote control. Due to the robust structural design and powerful actuators, Coyote II is able to carry several kilograms (> 6 kg) of payload. A special characteristic of Coyote II is its novel locomotion concept. It combines the high mobility performance of hybrid legged-wheels (in the front) with the smooth wheel movement of spherical helical wheels (in the rear). Therefore, the scout rover is able to move on soft soil as well as on unstructured terrain and can perform side-to-side steering movements. It is however, possible to mechanically tilt the rear axis in a horizontal position allowing to operate Coyote II with four equally shaped wheels in the front and rear.
Within the FASTER project Coyote II acts as scout rover with the aim to improve the mission safety and the effective traverse speed for planetary rover exploration. To avoid uncertain estimations concerning the trafficability of the areas to be explored, the scout rover provides suitable information on the terrain ahead of a primary exploration rover. To handle this task, Coyote II is equipped with an additional soil sensor payload. This consists of a Wheel-Leg Soil Interaction Observation (WLSIO) system and second, a motorized Dynamic Cone Penetrometer (mDCP). Another potential sensor payload is a Ground Penetrating Radar (GPR).
Innovative sensors
In the project, a number of sensors that can be used to assess the condition and trafficability of Martina soil were modified (miniaturized) or newly developed. The sensors were mounted either on the Scout rover or on the main exploration rover.
Wheel-Leg-Soil Interaction Observation (WLSIO) System: The WLSIO System uses the specific design of the scout rover’s front wheels in combination with visual data collected by two cameras installed a the underside of the scout rover, which observe the interaction of each wheel with the soil. From the collected images, sinkage of the wheels in the soil can be easily determined and when combined with other wheel-leg operating parameters (e.g. motor current, IMU data), soil parameters and trafficability can be calculated. The most appealing characteristics of this soil sensing sub-system are the facts that it inherently runs while the scout rover in motion and that it can operate continuously, making it ideal for producing a fast on-line trafficability diagnosis with minimum impact on the mission performance in terms of speed and power. The WLSIO was developed by project partner Surrey Space Center (University of Surrey, UK).
Motorised Dynamic Cone Penetrometer (mDCP): The mDCP is a simple dynamic hammering device, designed to be self-hammered to measure the penetration depth per blow. Such data enable general discrimination of the soil type and give the overall characterization of the geotechnical parameters of the soils, based on experience with a similar and well-known terrestrial geotechnical tool. Based on the available penetrometer designs, only those modelled on the principles of the motorised Dynamic Cone Penetrometer (mDCP) for their mechanical operation appear feasible for the FASTER application. Other designs are either too massive to be carried by the Scout or require a significant reaction force to be applied during operation, which cannot be provided by a small vehicle operating in reduced gravity. The data generated by the mDCP is arguably somewhat crude – a simple index (DCPi) based on penetration/blow. Significant variability exists in the methods of determining DCPi. However, the correlation between the DCPi and useful soil parametric data is well established, e.g. log-log correlation between DCPi and CBR (California Bearing Ratio), and should be sufficient to provide the baseline device and to enable Go/NoGo decisions to be made with good accuracy. The small, self-contained design of the mDCP makes it a good soil sensor for both space and terrestrial applications where terrain analysis of the upper soil layer is required. Its design can also be scaled up for deeper penetrations or for use with larger rovers. The mDCP was also developed by the Surrey Space Center.
Ground Penetrating Radar (GPR): The feasibility of using GPR on the Scout rover was evaluated in the FASTER project. However, the GPR was not included in the demonstrator as the financial framework of the project did not support the development of a miniaturized version of a GPR system. In general, GPR is based on the reflections and attenuation of a beam of electromagnetic waves sent directly into the prospected soil, where it is reflected and finally received by the device. GPR is a device well-known in terrestrial applications. It is widely used by geologists, engineers and archaeologists and can deliver valuable data of the subsurface soil properties. Applied to the FASTER scenario, forward looking GPR could be particularly valuable in detecting duricrusts (hard but fragile layers of encrusted soil that cover loose granular material and/ subsurface voids) without having to stop the Scout and without the risk of the Scout breaking through the duricrust (which could happen when the WLSIO or the mDCP sensors are used). It could also be used as a tool for scientific investigation of the planetary surface.
Wheeled Bevameter (WB): The wheeled bevameter is a device mounted to the primary rover to obtain the ‘last chance’ data related to the geotechnical properties of the soil directly before the rover. The method of the measurements is simple: in its normal mode, the sinkage, motion resistance and rotation rate of the free rolling wheel are measured while the rover is driving. This enables to calculate %Tr of the soil using empirical equations and, using Bekker’s equations, permits to solve for the soil geotechnical properties. The WB as conceived by LSG uses a dedicated test wheel placed on the terrain as the loading device to enable both bearing strength and shear strength measurements while the host vehicle is driving.
The method behind the WB is according to the terrain properties estimation method used on the NASA MER rovers with the primary difference being that for FASTER it would follow a real-time approach (and use a dedicated test wheel), rather than being an off- line method as done on MER. A measurement wheel (‘test wheel’) is used to load the terrain (from natural weight of the deployed test wheel assembly) for acquiring the needed terrain and vehicle-terrain interaction parameters. The test wheel is arranged such that it protrudes in the rover driving direction. It is not a driving wheel, i.e. it is not powered. The WB includes a placement mechanism for the test wheel and would be stowed until after landing on Mars. It is expected that the test wheel would remain lowered onto the ground during nominal rover motion, including when climbing and descending slopes.
During normal operations, the placement mechanism assumes the function of a passive suspension of the test wheel, allowing the wheel to follow the terrain contour (including rolling over rocks and climbing as well as descending slopes of up to 25° as per the FASTER Primary rover require-ments). The system is capable of autonomous-ly detecting test wheel ‘stalls’ against rocks exceeding the wheel obstacle climbing capa-bility, permitting an autonomous raising of the test wheel off the surface, using the placement mechanism active joints, and its subsequent re- placement.
It’s worth to note, the substitute device, called PathBeater, was analysed during the prelimi- nary development phase of the project. After a trade-off , it was decided to do not develop this device as a final part of the soil sensing system for FASTER project. Overall, the WB would also be suited as a soil physical properties scientific instrument on planetary rovers as key parameters of the soil along the vehicle traverse path are continually measured and thus documented.
Remote Sensing (RS): The Remote Sensing sensor was also installed on the primary rover. It enables the detection and localisation of rocks in the field of view of the camera mounted on the Primary rover. Detection and tracking of rocks on the surface of planets can be achieved using unsupervised modelling techniques that can identify and de- scribe ‘regions of importance’ (ROIs) in terms of quasi-thematic features, including colour, texture, intensity, shape and so on. In the con- text of this project, this involves detection and tracking based on visual saliency. Specific surface characteristics of objects on planetary surfaces provide sufficient information for them to be distinguished in the visual scene, and computer vision paradigms that use descriptions of objects in terms of their visual saliency to segregate them from their periphery are em- ployed. Thus rock detection is performed via visual saliency based semantic description of objects using a modified saliency model based on the ‘Rudinac’ algorithm which incorporates colour information. This has the advantage over purely intensity based techniques of im proving rejection of false detections arising from textures in the sand, despite the relatively limited colourspace of planetary surfaces. Once rocks have been identifi ed using these saliency techniques, the actual dimensions and location of the rocks in 3D space is estimated using disparity information from the main rover stereo camera. The rock locations and dimen- sions identifi ed are sent to the common Data Fusion subsystem in the same way as for the rest of the soil sensors.
Hardware and software for co-operative robots
Based on the project goal of improving average traversal speeds of planetary rovers through the forward acquisition of trafficability information by the use of a Scout rover, the operational concept developed concentrates on the ‘traverse phase’ of surface exploration missions, that is the segment of operations requiring the mission rover to traverse large distances with minimal or no science in that period.
The key capabilities enabling primary rover autonomy in the context of the FASTER operational scenario are:
- Traverse graph operations: Graph search and maintenance operations allowing the addition/ deletion/modification of vertices and edges in the traverse graph.
- Mapping: Build Digital Elevation Maps (DEMs) based on stereo images. This should include the capability to merge data from multiple stereo image pairs, including those from the Scout rover. Additionally, images and/or data from the Primary rover should be filtered to ensure that the Scout rover is not included as part of the elevation map.
- Path planning: Planning a path to a local goal based on a DEM. Apart from geometric obstacles and rover capabilities, this should take into account the detected rocks (as reported by the visual soil sensing component of the FASTER Soil Sensing System (SSS)) and any known trafficability information.
- Path traversal (rover control): Following a planned path safely, while interacting with components of the FASTER SSS that have been deployed on the Primary rover. This should implicitly allow synchronization with the path traversal by the Scout rover, with the Scout rover acting as an artificial obstacle.
- Self-localization: Accurate self-localization is required for successful, long range traversal, both in the context of following a specific planned path that avoids nearby hazards, as well as being able to reach the target location specified by operators.
- Scout localization: Recognition and localization of the Scout rover using camera images from the Primary rover (potentially using the navigation stereo bench). This is to be used to support scout rover traversal by providing bounds for the localization error.
- Communication with the Scout: Transmission of commands to the Scout rover, and recep-tion of trafficability information and other data (such as stereo images and status).
Key Scout rover capabilities enabling the planned operation of the team are:
- Path traversal: Safely moving along a path planned by the Primary rover, provided to the scout along with the local DEM, while conducting ‘forward sensing’ that is operating the components of the FASTER SSS that are part of the Scout rover.
- Self-localization: While it is expected that the Primary rover will provide periodic localization updates to the Scout, the Scout rover will be able to localize itself using the external localization updates from the Primary rover as a means of correction.
- Return to Primary rover: In certain cases of hazard detection during traversal, as well as some emergency scenarios (identified in D2.1) the Scout rover should be able to return to the Primary rover using the local DEM and the last path followed (or an updated path received from the Primary rover).
- Communication with Primary rover: Sending telemetry (especially trafficability) information to the Primary rover, and reception of tasks (primarily traversal commands).
The FASTER project successfully developed requirements and operational schemes needed for two planetary rover common operations. The final functionality of the system can be found as a significant enhancement of the mobility, which is the important goal for the planetary engineering. The planetary perspective and success metrics will be completely different from typical industrial standards presented by the terrestrial systems.
Software for data processing and data fusion
Data management and data processing for the FASTER system are realized by complex subsystem joining hardware and software components across the system. The data management for the FASTER can be simply divided into hardware component, including communication subsystem on both rovers and enabling the transmission of the data between them, on-board computer (OBC), enabling data processing and calculation and software of the FASTER system plus separated devices. Data Fusion (DF) process in the systems integrating data from various sensors is crucial for overall system effectiveness and correct operations. The Data Fusion subsystem is located in between the soil sensors interface nodes, additional data (e.g. time, position, rover attitude etc.) and, on the other side, the NAV MAP node (as a part of the Navigation Subsystem).
Different types of data from sensors and additional data sources are used as input data on the same way. The specialized Data Fusion input interface simply read all data. The interface defined in this document is the base for all data incoming to DF module through input interface. Data Fusion is prepared as one ROS node, with additional possibility to prepare additional node before the DF module in case the need of further pre- processing data (e.g. problems with compatibility or timing of input data). The common definition of input interface makes the subsystem highly flexible. The DF module is ready to acquire data from additional sources, like now discussed, possible simple addition of GPR sensor which can be a good example of the potential benefits of such DF module development strategy.
The DF subsystem is capable to use both synchronous and asynchronous data as the input. Asynchronous data are handled by publisher-subscriber architecture and each sensor publishes messages to common channel, which is subscribed by DF module. In case of synchronous data (client-server architecture) the prioritising can be used to processed time-restricted data sources e.g. sensors deployed on demand. Asynchronous interface is based on ROS messages. On this channel, input data are buffered, by internal ROS mechanism, and processed in continuous loop. A synchronous channel is implemented as a ROS service with request of the same type as asynchronous messages and empty response. Data passed by this interface are processed with highest priority and are not buffered.
Data exchanged by ROS are always packed in proper messages, and there is not necessary to check its consistency. However, part of fields has defined constraints and this formal correctness is checked on the interface level and rejected when any problem occurred. Mainly, the trafficability percentage (%Tr) value is finally in 0 to 100 range. Pre-processing and analysis steps are provided for statistical algorithms and other methods, which pro- duce sensor related model parameters for each data source and to generalise data and check it for inconsistency with developed core fusion method.
While the input interface makes the Data Fusion module independent from the data format, this steps provides versatility with respect to measurement values, both over time and over value level. The fusion of the measurements should be done in two subsequent steps. The first step is called Recursive Update, in which the current measurement is fused with an estimation computed from previous readings from the same sensor. In this step, the algorithm fuses measurements of the same sensor over time. A Bayes Filter is a standard algorithm to compute that estimation. In a second step, the algorithm merges information from each sensor estimation over current cell. Because of different sensors measurement methods, which could cause different properties/obstacles detection over data from different sensors, it is not appropriate to use Bayes Filtering during this step. In this case, a simple solution called Worst Case Scenario Merging is implemented.
The main task for Data Fusion is to update the trafficability layer of the global map. This information and the terrain-cost map on a second layer are used as a base for path planning process. Because whole control system is managed by the scheduler, the DF module should update the global map on demand. For that interface, a ROS service (or ROS action) will be used. The general idea is to exchange the local part of the map. The map for update will be as large as the range of robots system perception and the distance to the next planned waypoint. We should take into account an additional buffer, which will be needed because of the Primary rover traverse. The work on the Data Fusion algorithm was focused not only on the direct management and processing of the incoming data, but also on the wide discussion on the uncertainty of the acquired data. Although the simple algorithm was finally chosen by the partners, the other possible algorithms and ideas on data comparison and validation were also discussed. Additionally, the further activities in the topic of fusion of soil were recognized and discussed.
Potential Impact:
In addition to a scientific contribution to the current state-of-the-art in sensor technology and co-operative robotics, the overall impact of the FASTER project relates mainly to three areas:
Creating a European network of experts in space exploration technology - The FASTER system for soil and terrain data acquisition for exploration rovers is complex and requires the cooperation of highly qualified specialists in multiple sectors, such as robotic rovers, sensor development, software development, planning of space exploration missions and others. Thus the cooperation of research groups from industry and academia in several European member states achieved in FASTER helped to form lasting alliances between key players in this sector that are likely to play a role in the development of key technologies for Space exploration in the future. This will help to maintain and improve the competitive edge of European research and development on space robotics towards the other space-faring nations.
Contribution to non-space robotics research and industry
Investment in space research has always had a significant impact on non-space markets and applications. The notorious Teflon pan is one example of a product that was developed for space, but has had a huge economic impact on Earth. The great economic potential of space technology for terrestrial applications is even more obvious with respect to space robotics. “The technologies applicable to space robotics will enable a wide range of Earth-based exploration and material processing activities from automated undersea inspection to mining and mineral extraction under hazardous conditions” (from: EUROP Strategic Research Agenda 2008).
The methods and technologies developed in FASTER do not only have the potential to make future planetary exploration missions more (cost) efficient, but also be of great value for terrestrial exploration robots. This applies in particular to the solutions for co-operative robots that were developed in FASTER. Here, spin-out opportunities, for example the co-operative monitoring of dikes and other structures, applications in humanitarian de-mining etc., were already defined and discussed during the FASTER dissemination workshops.
Market potential
The concepts and technologies developed and demonstrated in FASTER do have a significant potential for Space exploration. However, as this market is driven by government funding, i.e. missions planned by ESA, NASA and other national space agencies, it is very difficult to gauge the market potential here. The FASTER concept was explicitly developed to match as much as possible the requirements of a future ESA mission to Mars. If ESA is going ahead with this mission and if ESA decides to try new and innovative robotic and sensor concepts on this mission, then the FASTER project can play a key role in setting the stage for the development and space-qualification of such concepts. The consortium is convinced that concepts similar to the FASTER concept will eventually be used for planetary exploration. However, it is not possible at the moment to give a serious estimate about the time frame of such developments.
On the other hand, the terrestrial market for mobile robotics is growing steadily and analysts even predict a significant acceleration of growth in the years to come. This applies to industrial robotics, where flexible and mobile systems play an increasing role, but also to the sector of professional service robotics and even consumer robotics. In all these application domains, the co-operation of two or even multiple mobile robots to achieve a higher level of functionality is a key component. Here, the algorithms and technical solutions for robot co-operation developed in FASTER can be readily applied.
List of Websites:
https://www.faster-fp7-space.eu/
Contact: Dr. Thomas Vögele, DFKI Robotics Innovation Center, Robert-Hooke Str 1, 28359 Bremen, Germany
thomas.voegele@dfki.de, +49 421 17845 4130
The FASTER project developed novel and innovative concepts and methods for a reliable in-situ forward evaluation of soil properties and terrain conditions along the planned trajectory of a planetary exploration rover. These methods can be used to make robotic planetary exploration missions more efficient and less risky. Today, the exploration of planetary surfaces by robots is seriously impeded by the lack of a priori detailed information about the soil- and terrain conditions in the vicinity of the exploration rover. In the best case, this lack of information leads to extremely slow and cautious exploration strategies (well below 5 cm/s). In the worst case, rovers may get stuck (temporarily or permanently) due to unforeseen terrain conditions. The recent immobilisation of the NASA MER Spirit rover, as it became trapped in hidden soft sand, dramatically illustrates this scenario.
The FASTER concept foresees a highly mobile scout rover co-operating with a planetary exploration rover (such as EXOMARS). Both are equipped with sensors to acquire soil and terrain information in-situ. The sensor data are fused and used to develop a risk model that indicates potential hazards along the planned trajectory of the exploration rover. The objectives of FASTER were thus to develop key technologies in three different areas:
• Forward-looking characterization of surface properties: FASTER developed methods and instrumentation to characterize the trafficability-properties of planetary surfaces and thus to anticipate potential hazards along the planned trajectory of a planetary exploration rover.
• Innovative locomotion system for a scout type rover: The project developed an innovative locomotion concept to ensure that the scout rover is able to operate in-sync with the exploration rover, with minimal risk of encountering hazardous situations from which it must recover.
• Collaborative operation of a mother/scout pair: FASTER explored the feasibility and performance of a mother/scout rover pair and developed methods to achieve successful collaborative and autonomous robot operation.
FASTER started on November 1st, 2011 with a project duration of 37 months. The project received funding from the European Union’s Seventh Framework Programme for Research, Technological Development and Demonstration under grant agreement no 284419.
A European consortium of five partner from four EU member states contributed to FASTER. The consortium brought together European experts in the areas of planetary science, soil and terrain trafficability analysis, planetary exploration robotics, and robot autonomy and co-operation. It encompassed research centers (DFKI GmbH., Germany), academia (University of Surrey, UK), larger industrial companies (EADS DS, UK and Astri Polska, Poland), and SMEs (Liquifer Systems Group, Austria) with proven competencies in the research areas addressed.
Project Context and Objectives:
The FASTER project was aimed at strengthening the foundation technologies in support of space science and exploration, supporting the transportation of science instrumentation to as yet unexplored parts of other solar system planets and reducing the vulnerability of exploration assets (planetary rovers). In addition, the outcomes of the project may be used to improve the productivity of space exploration whilst expanding the range of environments into which scientific instrumentation can be transported.
The project invoked and encouraged cross-national collaboration in a study that fully supports European Space Policy and exploration initiatives. The FASTER project developed innovative and feasible space robotic technology in support of Europe’s vision for exploration and possible human settlement on Mars or the Moon. The new engineering and technological solutions developed under this project support and complement current and future EU and ESA space exploration projects. The outcome of this study will contribute significantly to ESA’s upcoming space exploration missions, in particular the Mars Sample Return mission.
Strategic objectives: The outcomes of the FASTER project supports the ability of future planetary exploration robots to explore hitherto unexplored parts of solar system planets at lower cost and risk. This raises the possibility of gaining a better understanding of how planets in our Solar System are formed and evolve, thereby raising the value of current planetary modelling and resolving many unanswered questions about our universe. The utility created by FASTER technology enables to an increase in the quality and quantity of scientific returns from interplanetary missions, thereby enriching scientific knowledge and opening up the development of new scientific models and understanding. The project thus enhances Europe’s standing within the space exploration community world-wide and support collaborative and co-operative exploration with other space agencies and nations.
Technical objectives: The overall goal of the FASTER project was to address three key technology developments for planetary exploration:
• Forward looking surface properties characterisation;
• Innovative locomotion system for a scout type rover;
• Collaborative operation of a mother/scout pair.
The first focal point was to develop methods and instrumentation to characterise the properties of planetary surfaces in which robotic rover vehicles operate in order to be able to anticipate hazards in advance of locomotion and navigation over that surface. By assessing the soil trafficability for a rover, decisions affecting rover safety (avoiding becoming stuck) can be reliably taken and hazards avoided.
The second focal point was to look at what innovative locomotion principle on can be applied to a scout rover to ensure that scout is able to conduct its soil characterisation operations with minimal risk of encountering hazardous situations from which it must recover.
The third focal point was to explore the feasibility and performance of a mother / scout rover pair combination to achieve successful collaborative and autonomous operation. Both the forward looking scout rover and the exploration rover were equipped with newly developed sensor technologies to acquire soil and terrain information. This information was used to calculate a risk model that indicated the potential hazards that the terrain represents to the exploration rover.
Reducing the locomotion and traverse risks using the proposed advances in autonomous mother / scout collaboration will allow mission operators to explore planetary surfaces with increased safety. It will make possible new levels of autonomous operations by significantly reducing the greatest uncertainty factor – namely properties of the surface material on which the rover must operate. This will lead to increased operational efficiency and, when coupled with the increased richness of the surface properties data, will lead to much higher scientific returns per capital investment for each mission.
Project Results:
Objectives and Requirements
The main results of the FASTER project are concepts and technologies for an in-situ and on-the-ground evaluation the soil-properties in the vicinity of a planetary exploration rover. Although the technologies developed in FASTER can also be used for scientific exploration of planetary surfaces, the primary objective in the project was to establish a trafficability map that indicates a safe trajectory for the planetary exploration rover. The primary use-case was a mission to Mars, such as the planned ESA Exomars mission.
The study was based on a requirements evaluation that covers all aspects of the potential Martian mission, including solar radiation environment, Mars atmosphere conditions and surface regolith material, properties and features (eg. rocks, outcrops, dunes, etc.). These were used to derive system requirements for the planned robotic and sensor sub-systems. Although it was beyond the scope of the FASTER project to develop space-qualified technologies for a specific mission, the overall goal was to stay as close as possible to the requirements of a planned real ESA mission to Mars in order to ensure basic compatibility of the FASTER concept with ESA planning.
Technological developments
The main components of the system consist of two robotic platforms, a main planetary exploration rover (such as ExoMars) and a light-weight scout rover (developed in FASTER). The main sensory and software components that were developed in FASTER were distributed among these two platforms.
Mobile robotic platforms
Primary Rover: The term “Primary Rover” denotes the main robotic rover used for a planetary exploration mission. Initially, the objective was to prepare FASTER system as a direct support for ExoMars rover planned by ESA. However, after redefinition of the mission, FASTER system is now rather a technology demonstrator prepared with regards to ExoMars requirements (mass, volume, energy requirements, etc.) but in fact capable to be used as an universal safe terrain traversal supporting system for any planned rover of similar size. In the project, as the Primary Rover, existing ExoMars mockup Bridget, developed in Airbus DS, was used.
Coyote II Scout Rover: Coyote II was developed by the Robotics Innovation Lab of project partner and coordinator DFKI. It is a micro rover with high mobility performance in various terrains. Equipped with its own power source, on-board sensor suite and computer it is able to perform autonomous exploration tasks. The communication subsystem allows to co-operate with other systems and provides a link for remote control. Due to the robust structural design and powerful actuators, Coyote II is able to carry several kilograms (> 6 kg) of payload. A special characteristic of Coyote II is its novel locomotion concept. It combines the high mobility performance of hybrid legged-wheels (in the front) with the smooth wheel movement of spherical helical wheels (in the rear). Therefore, the scout rover is able to move on soft soil as well as on unstructured terrain and can perform side-to-side steering movements. It is however, possible to mechanically tilt the rear axis in a horizontal position allowing to operate Coyote II with four equally shaped wheels in the front and rear.
Within the FASTER project Coyote II acts as scout rover with the aim to improve the mission safety and the effective traverse speed for planetary rover exploration. To avoid uncertain estimations concerning the trafficability of the areas to be explored, the scout rover provides suitable information on the terrain ahead of a primary exploration rover. To handle this task, Coyote II is equipped with an additional soil sensor payload. This consists of a Wheel-Leg Soil Interaction Observation (WLSIO) system and second, a motorized Dynamic Cone Penetrometer (mDCP). Another potential sensor payload is a Ground Penetrating Radar (GPR).
Innovative sensors
In the project, a number of sensors that can be used to assess the condition and trafficability of Martina soil were modified (miniaturized) or newly developed. The sensors were mounted either on the Scout rover or on the main exploration rover.
Wheel-Leg-Soil Interaction Observation (WLSIO) System: The WLSIO System uses the specific design of the scout rover’s front wheels in combination with visual data collected by two cameras installed a the underside of the scout rover, which observe the interaction of each wheel with the soil. From the collected images, sinkage of the wheels in the soil can be easily determined and when combined with other wheel-leg operating parameters (e.g. motor current, IMU data), soil parameters and trafficability can be calculated. The most appealing characteristics of this soil sensing sub-system are the facts that it inherently runs while the scout rover in motion and that it can operate continuously, making it ideal for producing a fast on-line trafficability diagnosis with minimum impact on the mission performance in terms of speed and power. The WLSIO was developed by project partner Surrey Space Center (University of Surrey, UK).
Motorised Dynamic Cone Penetrometer (mDCP): The mDCP is a simple dynamic hammering device, designed to be self-hammered to measure the penetration depth per blow. Such data enable general discrimination of the soil type and give the overall characterization of the geotechnical parameters of the soils, based on experience with a similar and well-known terrestrial geotechnical tool. Based on the available penetrometer designs, only those modelled on the principles of the motorised Dynamic Cone Penetrometer (mDCP) for their mechanical operation appear feasible for the FASTER application. Other designs are either too massive to be carried by the Scout or require a significant reaction force to be applied during operation, which cannot be provided by a small vehicle operating in reduced gravity. The data generated by the mDCP is arguably somewhat crude – a simple index (DCPi) based on penetration/blow. Significant variability exists in the methods of determining DCPi. However, the correlation between the DCPi and useful soil parametric data is well established, e.g. log-log correlation between DCPi and CBR (California Bearing Ratio), and should be sufficient to provide the baseline device and to enable Go/NoGo decisions to be made with good accuracy. The small, self-contained design of the mDCP makes it a good soil sensor for both space and terrestrial applications where terrain analysis of the upper soil layer is required. Its design can also be scaled up for deeper penetrations or for use with larger rovers. The mDCP was also developed by the Surrey Space Center.
Ground Penetrating Radar (GPR): The feasibility of using GPR on the Scout rover was evaluated in the FASTER project. However, the GPR was not included in the demonstrator as the financial framework of the project did not support the development of a miniaturized version of a GPR system. In general, GPR is based on the reflections and attenuation of a beam of electromagnetic waves sent directly into the prospected soil, where it is reflected and finally received by the device. GPR is a device well-known in terrestrial applications. It is widely used by geologists, engineers and archaeologists and can deliver valuable data of the subsurface soil properties. Applied to the FASTER scenario, forward looking GPR could be particularly valuable in detecting duricrusts (hard but fragile layers of encrusted soil that cover loose granular material and/ subsurface voids) without having to stop the Scout and without the risk of the Scout breaking through the duricrust (which could happen when the WLSIO or the mDCP sensors are used). It could also be used as a tool for scientific investigation of the planetary surface.
Wheeled Bevameter (WB): The wheeled bevameter is a device mounted to the primary rover to obtain the ‘last chance’ data related to the geotechnical properties of the soil directly before the rover. The method of the measurements is simple: in its normal mode, the sinkage, motion resistance and rotation rate of the free rolling wheel are measured while the rover is driving. This enables to calculate %Tr of the soil using empirical equations and, using Bekker’s equations, permits to solve for the soil geotechnical properties. The WB as conceived by LSG uses a dedicated test wheel placed on the terrain as the loading device to enable both bearing strength and shear strength measurements while the host vehicle is driving.
The method behind the WB is according to the terrain properties estimation method used on the NASA MER rovers with the primary difference being that for FASTER it would follow a real-time approach (and use a dedicated test wheel), rather than being an off- line method as done on MER. A measurement wheel (‘test wheel’) is used to load the terrain (from natural weight of the deployed test wheel assembly) for acquiring the needed terrain and vehicle-terrain interaction parameters. The test wheel is arranged such that it protrudes in the rover driving direction. It is not a driving wheel, i.e. it is not powered. The WB includes a placement mechanism for the test wheel and would be stowed until after landing on Mars. It is expected that the test wheel would remain lowered onto the ground during nominal rover motion, including when climbing and descending slopes.
During normal operations, the placement mechanism assumes the function of a passive suspension of the test wheel, allowing the wheel to follow the terrain contour (including rolling over rocks and climbing as well as descending slopes of up to 25° as per the FASTER Primary rover require-ments). The system is capable of autonomous-ly detecting test wheel ‘stalls’ against rocks exceeding the wheel obstacle climbing capa-bility, permitting an autonomous raising of the test wheel off the surface, using the placement mechanism active joints, and its subsequent re- placement.
It’s worth to note, the substitute device, called PathBeater, was analysed during the prelimi- nary development phase of the project. After a trade-off , it was decided to do not develop this device as a final part of the soil sensing system for FASTER project. Overall, the WB would also be suited as a soil physical properties scientific instrument on planetary rovers as key parameters of the soil along the vehicle traverse path are continually measured and thus documented.
Remote Sensing (RS): The Remote Sensing sensor was also installed on the primary rover. It enables the detection and localisation of rocks in the field of view of the camera mounted on the Primary rover. Detection and tracking of rocks on the surface of planets can be achieved using unsupervised modelling techniques that can identify and de- scribe ‘regions of importance’ (ROIs) in terms of quasi-thematic features, including colour, texture, intensity, shape and so on. In the con- text of this project, this involves detection and tracking based on visual saliency. Specific surface characteristics of objects on planetary surfaces provide sufficient information for them to be distinguished in the visual scene, and computer vision paradigms that use descriptions of objects in terms of their visual saliency to segregate them from their periphery are em- ployed. Thus rock detection is performed via visual saliency based semantic description of objects using a modified saliency model based on the ‘Rudinac’ algorithm which incorporates colour information. This has the advantage over purely intensity based techniques of im proving rejection of false detections arising from textures in the sand, despite the relatively limited colourspace of planetary surfaces. Once rocks have been identifi ed using these saliency techniques, the actual dimensions and location of the rocks in 3D space is estimated using disparity information from the main rover stereo camera. The rock locations and dimen- sions identifi ed are sent to the common Data Fusion subsystem in the same way as for the rest of the soil sensors.
Hardware and software for co-operative robots
Based on the project goal of improving average traversal speeds of planetary rovers through the forward acquisition of trafficability information by the use of a Scout rover, the operational concept developed concentrates on the ‘traverse phase’ of surface exploration missions, that is the segment of operations requiring the mission rover to traverse large distances with minimal or no science in that period.
The key capabilities enabling primary rover autonomy in the context of the FASTER operational scenario are:
- Traverse graph operations: Graph search and maintenance operations allowing the addition/ deletion/modification of vertices and edges in the traverse graph.
- Mapping: Build Digital Elevation Maps (DEMs) based on stereo images. This should include the capability to merge data from multiple stereo image pairs, including those from the Scout rover. Additionally, images and/or data from the Primary rover should be filtered to ensure that the Scout rover is not included as part of the elevation map.
- Path planning: Planning a path to a local goal based on a DEM. Apart from geometric obstacles and rover capabilities, this should take into account the detected rocks (as reported by the visual soil sensing component of the FASTER Soil Sensing System (SSS)) and any known trafficability information.
- Path traversal (rover control): Following a planned path safely, while interacting with components of the FASTER SSS that have been deployed on the Primary rover. This should implicitly allow synchronization with the path traversal by the Scout rover, with the Scout rover acting as an artificial obstacle.
- Self-localization: Accurate self-localization is required for successful, long range traversal, both in the context of following a specific planned path that avoids nearby hazards, as well as being able to reach the target location specified by operators.
- Scout localization: Recognition and localization of the Scout rover using camera images from the Primary rover (potentially using the navigation stereo bench). This is to be used to support scout rover traversal by providing bounds for the localization error.
- Communication with the Scout: Transmission of commands to the Scout rover, and recep-tion of trafficability information and other data (such as stereo images and status).
Key Scout rover capabilities enabling the planned operation of the team are:
- Path traversal: Safely moving along a path planned by the Primary rover, provided to the scout along with the local DEM, while conducting ‘forward sensing’ that is operating the components of the FASTER SSS that are part of the Scout rover.
- Self-localization: While it is expected that the Primary rover will provide periodic localization updates to the Scout, the Scout rover will be able to localize itself using the external localization updates from the Primary rover as a means of correction.
- Return to Primary rover: In certain cases of hazard detection during traversal, as well as some emergency scenarios (identified in D2.1) the Scout rover should be able to return to the Primary rover using the local DEM and the last path followed (or an updated path received from the Primary rover).
- Communication with Primary rover: Sending telemetry (especially trafficability) information to the Primary rover, and reception of tasks (primarily traversal commands).
The FASTER project successfully developed requirements and operational schemes needed for two planetary rover common operations. The final functionality of the system can be found as a significant enhancement of the mobility, which is the important goal for the planetary engineering. The planetary perspective and success metrics will be completely different from typical industrial standards presented by the terrestrial systems.
Software for data processing and data fusion
Data management and data processing for the FASTER system are realized by complex subsystem joining hardware and software components across the system. The data management for the FASTER can be simply divided into hardware component, including communication subsystem on both rovers and enabling the transmission of the data between them, on-board computer (OBC), enabling data processing and calculation and software of the FASTER system plus separated devices. Data Fusion (DF) process in the systems integrating data from various sensors is crucial for overall system effectiveness and correct operations. The Data Fusion subsystem is located in between the soil sensors interface nodes, additional data (e.g. time, position, rover attitude etc.) and, on the other side, the NAV MAP node (as a part of the Navigation Subsystem).
Different types of data from sensors and additional data sources are used as input data on the same way. The specialized Data Fusion input interface simply read all data. The interface defined in this document is the base for all data incoming to DF module through input interface. Data Fusion is prepared as one ROS node, with additional possibility to prepare additional node before the DF module in case the need of further pre- processing data (e.g. problems with compatibility or timing of input data). The common definition of input interface makes the subsystem highly flexible. The DF module is ready to acquire data from additional sources, like now discussed, possible simple addition of GPR sensor which can be a good example of the potential benefits of such DF module development strategy.
The DF subsystem is capable to use both synchronous and asynchronous data as the input. Asynchronous data are handled by publisher-subscriber architecture and each sensor publishes messages to common channel, which is subscribed by DF module. In case of synchronous data (client-server architecture) the prioritising can be used to processed time-restricted data sources e.g. sensors deployed on demand. Asynchronous interface is based on ROS messages. On this channel, input data are buffered, by internal ROS mechanism, and processed in continuous loop. A synchronous channel is implemented as a ROS service with request of the same type as asynchronous messages and empty response. Data passed by this interface are processed with highest priority and are not buffered.
Data exchanged by ROS are always packed in proper messages, and there is not necessary to check its consistency. However, part of fields has defined constraints and this formal correctness is checked on the interface level and rejected when any problem occurred. Mainly, the trafficability percentage (%Tr) value is finally in 0 to 100 range. Pre-processing and analysis steps are provided for statistical algorithms and other methods, which pro- duce sensor related model parameters for each data source and to generalise data and check it for inconsistency with developed core fusion method.
While the input interface makes the Data Fusion module independent from the data format, this steps provides versatility with respect to measurement values, both over time and over value level. The fusion of the measurements should be done in two subsequent steps. The first step is called Recursive Update, in which the current measurement is fused with an estimation computed from previous readings from the same sensor. In this step, the algorithm fuses measurements of the same sensor over time. A Bayes Filter is a standard algorithm to compute that estimation. In a second step, the algorithm merges information from each sensor estimation over current cell. Because of different sensors measurement methods, which could cause different properties/obstacles detection over data from different sensors, it is not appropriate to use Bayes Filtering during this step. In this case, a simple solution called Worst Case Scenario Merging is implemented.
The main task for Data Fusion is to update the trafficability layer of the global map. This information and the terrain-cost map on a second layer are used as a base for path planning process. Because whole control system is managed by the scheduler, the DF module should update the global map on demand. For that interface, a ROS service (or ROS action) will be used. The general idea is to exchange the local part of the map. The map for update will be as large as the range of robots system perception and the distance to the next planned waypoint. We should take into account an additional buffer, which will be needed because of the Primary rover traverse. The work on the Data Fusion algorithm was focused not only on the direct management and processing of the incoming data, but also on the wide discussion on the uncertainty of the acquired data. Although the simple algorithm was finally chosen by the partners, the other possible algorithms and ideas on data comparison and validation were also discussed. Additionally, the further activities in the topic of fusion of soil were recognized and discussed.
Potential Impact:
In addition to a scientific contribution to the current state-of-the-art in sensor technology and co-operative robotics, the overall impact of the FASTER project relates mainly to three areas:
Creating a European network of experts in space exploration technology - The FASTER system for soil and terrain data acquisition for exploration rovers is complex and requires the cooperation of highly qualified specialists in multiple sectors, such as robotic rovers, sensor development, software development, planning of space exploration missions and others. Thus the cooperation of research groups from industry and academia in several European member states achieved in FASTER helped to form lasting alliances between key players in this sector that are likely to play a role in the development of key technologies for Space exploration in the future. This will help to maintain and improve the competitive edge of European research and development on space robotics towards the other space-faring nations.
Contribution to non-space robotics research and industry
Investment in space research has always had a significant impact on non-space markets and applications. The notorious Teflon pan is one example of a product that was developed for space, but has had a huge economic impact on Earth. The great economic potential of space technology for terrestrial applications is even more obvious with respect to space robotics. “The technologies applicable to space robotics will enable a wide range of Earth-based exploration and material processing activities from automated undersea inspection to mining and mineral extraction under hazardous conditions” (from: EUROP Strategic Research Agenda 2008).
The methods and technologies developed in FASTER do not only have the potential to make future planetary exploration missions more (cost) efficient, but also be of great value for terrestrial exploration robots. This applies in particular to the solutions for co-operative robots that were developed in FASTER. Here, spin-out opportunities, for example the co-operative monitoring of dikes and other structures, applications in humanitarian de-mining etc., were already defined and discussed during the FASTER dissemination workshops.
Market potential
The concepts and technologies developed and demonstrated in FASTER do have a significant potential for Space exploration. However, as this market is driven by government funding, i.e. missions planned by ESA, NASA and other national space agencies, it is very difficult to gauge the market potential here. The FASTER concept was explicitly developed to match as much as possible the requirements of a future ESA mission to Mars. If ESA is going ahead with this mission and if ESA decides to try new and innovative robotic and sensor concepts on this mission, then the FASTER project can play a key role in setting the stage for the development and space-qualification of such concepts. The consortium is convinced that concepts similar to the FASTER concept will eventually be used for planetary exploration. However, it is not possible at the moment to give a serious estimate about the time frame of such developments.
On the other hand, the terrestrial market for mobile robotics is growing steadily and analysts even predict a significant acceleration of growth in the years to come. This applies to industrial robotics, where flexible and mobile systems play an increasing role, but also to the sector of professional service robotics and even consumer robotics. In all these application domains, the co-operation of two or even multiple mobile robots to achieve a higher level of functionality is a key component. Here, the algorithms and technical solutions for robot co-operation developed in FASTER can be readily applied.
List of Websites:
https://www.faster-fp7-space.eu/
Contact: Dr. Thomas Vögele, DFKI Robotics Innovation Center, Robert-Hooke Str 1, 28359 Bremen, Germany
thomas.voegele@dfki.de, +49 421 17845 4130
