A Free-Climbing Robot for Steep Crater Terrain

Dirk Spenneberg (DFKI Robotics Lab), Frank Kirchner (DFKI Robotics Lab & University of Bremen) Robert Hooke Str. 5

28329 Bremen Germany

dirk.spenneberg@dfki.de

Abstract

We present the goals, the approach, and the first results of the concept phase of the new project “Space-climber“ (funded by DLR, & ESA) which deals with the development of a biologically inspired, energy- efficient and adaptively free-climbing robot for steep slopes. We especially review he requirements given by the environment and certain assumptions which are legged robot specific. Furthermore, we present our combined simulation and optimization approach and the developed lunar crater test bed.

1. Introduction The goal of Spaceclimber is to prove that walking ro-botic systems present a solution for future missions on difficult terrain, in particular missions in craters or rock fissures. The robotic system that we intend to develop should be able to conquer irregular slopes of more than 80% and should be in a position to navigate with local autonomy using built-in sensors. The project started in July 2008. In the next sections we will pre-sent the logic of the concept phase, give an analysis of the crater terrain on moon (on which we focus for the design), discuss the requirements and assumptions, and discuss why simulation can play an important part al-ready in the concept phase of a development project. 2. Why Legged Robots Among the scientifically most interesting places on the moon and Mars are areas which require a very high degree of mobility, e.g. detrital fields, to carry out in-situ scientific investigations. Even more interesting, scientifically, are the craters and canyon walls, which are at the same time also more difficult to investigate. These places are highly suited for exogeological inves-tigations. The “Mars Express” mission has demonstrated that potential recent or fossilised habitats are mostly to be found in very deep-lying valleys with dark sedimenta-

tion. In addition, the “Mars Express” mission showed that water can be found in deep craters (see Fig. 1). Ice and other volatile substances probably also exist in the continually dark craters in the polar areas of the moon.

Fig. 1 Mars crater with water ice in the northerly low-lying area of Vastitas Borealis © ESA/DLR. Furthermore, it is assumed that the environmental con-ditions (such as the intensity of radiation) in deep caves and fissures are quite different to those on the surface. It is possible that there are microhabitats, ex-isting in the past or present, in which traces of life and very old stellar or interstellar particles could still be evident. At present it is only possible to investigate these loca-tions from a distance (e.g. by flying over them). If one is to carry out in-situ investigations in the future, it will be necessary to develop suitable systems which have the necessary mobility for negotiating such terrain. But current rover systems are not able to provide this mobility and it is unlikely that the already sophisticated rover chassis developments can be further improved to guarantee mobility on very steep and rough slopes.

Therefore we are focussing on a new development of a legged climbing robot in which we want to tackle the following major topics: • Intelligent morphology for robust, computa-

tionally simple, and energy-efficient locomo-tion: Skilful exploitation of their physical proper-ties within their construction enables biological systems to reduce significantly most of the com-plex control problems – governing a system with many degrees of freedom. Here we will focus on:

o Integration of visco-elastic elements into the joints/limbs of the walking robot

o Development of an intelligent self-adapting foot using “intelligent materi-als”

• Intelligent bio-inspired decentralized control concepts: Research in biology has shown that sys-tems like stick-insects rely heavily on a decentral-ized control architecture, which provides a very robust and fail-safe operation with a very low-demand of computational resources. Here we will focus on the development of a decentralized hard-ware architecture (each joint will have its own in-telligence) and corresponding decentralized data flow and software control architecture.

• Semi-autonomous control in a difficult and dy-namic environment: To bring the idea of a free-climbing robot to life, the issue of semi-autonomous navigation on natural slopes must be investigated thoroughly. In this context, semi-autonomous means that the robot is equipped with the ability to carry out local planning with dy-namic updating and to deal with unforeseen cir-cumstances, e.g. sliding on the slope, in order to attain the goals which it has been set by an exter-nal operator/scientist. In general, navigation and route planning of walking robots, which have more movement options and possibilities to deal with obstacles than driving systems, is a new field of research and will therefore play a crucial topic.

3. Lunar Crater Analysis As a reference terrain for the Spaceclimber design, we studied lunar craters. The moon, especially the polar areas, are at the moment of high interest, e.g. they are currently addressed in the ESA NEXT Phase A Lunar Lander Study [1]. In principle, the moon consists of two different kinds of regions, which are

• the Maria: Large, dark, basaltic plains on Earth's Moon, which are mostly on the visible side

• the Terrae (Highlands): A hilly landscape which is more heavily cratered.

The crust of the moon is shaped mainly by three ef-fects: impact of projectiles forming various sized cra-ters, volcanism (in the Mare), and craters tectonics resulting in wrinkled ridges. By studying more deeply the craters, we can identify that depending on the chosen crater all different kinds of terrain occur. At best, we have a simple crater (found mainly in the Maria) which has mainly a bowl- shaped depression, a mean slope of ~25°, and a aver-age diameter of less than 15km. At worst, we have a complex crater, with shallow floors, central uplifts, fissures, and terraces. Typically, these are larger than 15km. Their average slope is about 35°, but because of the fissures and terraces it is likely that the robot has to face terrain of up to 60° or even more. An example for this worst kind of craters is the famous Tycho cra-ter (see Fig. 2)

Fig. 2 Picture of Tycho Crater © NASA. This crater is about 4.8km deep and 85km wide and is covered with a lot of terraces and other harsh terrain. In addition, the surface can vary between pure rocky and melted material and the very fine and dusty Re-golith. Rocks with different site from some cm³ up to m³ oc-cur. And last but not least, large parts of the craters are typically lying in the dark, especially at the poles, where the sunlight surface angle is around 1.5°.

Fig. 3 SCORPION robot

Fig. 4 ARAMIES robot

3. Requirements and Assumptions From this environmental analysis and based on our experiences from the SCORPION (see Fig. 3) and ARAMIES (see Fig. 4) robots [2,3] we deduced cer-tain requirements and assumptions for a legged crater climbing robot. From the environmental ground conditions we got the following major requirements for the robotic sys-tem: (EN-R1) The system has to cope with 30°-35° (av-

erage) slopes.

(EN-R2) The system has to work in Regolith (loose substrate).

(EN-R3) The system has to work in partially rocky terrain.

(EN-R4) The robot has to be equipped with robust housing (dust, collisions).

(EN-R5) The navigation and control S/S has to work under high-contrast lightning conditions.

EN-R1 is based on the lunar crater analysis. Since 30°-35° is the average, the Spaceclimber has to be able to climb still steeper slopes. EN-R2 and EN-R3 address the different terrain the system has to tackle. These are the driving require-ments for the foot-design of the system. On the one hand, we have to develop a foot which is suitable in soft terrain, like the SCORPION’s, and on the other hand, we have to develop the foot in a way that it also gets hold in steep terrain, like the active ARAMIES foot. EN-R4 addresses the problem that the system has to be shielded against the extreme fine dust and also pro-vides a solid robustness, if it for example collides with a rock. EN-R5 refers to the extreme light conditions; in order to deal with the high contrasts, a highly dynamic cam-era system is needed. Furthermore, for navigation in the dark, active vision (e.g. laser scanner) has to be integrated into the system. In our list of environmental requirements, we focused on the ground conditions. Naturally, more environ-mental effects have to be considered, like the thermal conditions (which are due to the variety and the dark areas a challenging problem), the radiation conditions, the non-existing atmosphere, the lower gravity, and so on. These are taken into account in the design of the Spaceclimber robot, but as a first step, we are focus-sing on the development of a demonstrator which will be able to demonstrate this novel mobility under earth-conditions. Therefore the environmental ground condi-tions are for this phase the driving requirements. Dur-ing the development process, we have of course all other requirements regarding the environmental condi-tions in mind to avoid later problems to build all com-ponents space compliant. Furthermore we made some assumptions for the op-erations of the system. (O-A1) The operational range of the system has

to be at least ~1 km (O-A2) A reference payload should be integrated

(O-A3) The target weight should be less than 20 kg

(O-A4) The system should have a very small stow-ing volume

(O-A5) The system has to operate in shadowed areas

The Assumption O-A1 results from the assumption that it is likely that a walking robot will be used as an additional scout system which is attached to a larger rover taking it to the site (see Fig. 5).

Fig. 5 Rover-Scout Scenario. The picture shows the SCORPION robot which was detached from a rover observing and supporting the SCORPION operations. The tether can be seen as an option for very short ranges. Therefore it is likely that an operational range of 1km is already enough for the robot to go into a crater, take samples, and come back for recharging. Of course, a tether might also be an option but proved to be difficult for distances larger than 100m. Because of the expected shadows, a solar powered system in principle will be difficult, and battery power is preferred, which means that higher distances result in higher mass of the system. This has to be weighted against the mass available for scientific payload. If a compact payload like the Nanokhod GEP system is carried (~ 1kg), enough mass budget would remain to significantly increase the operational range. O-A2 refers to the need to operate a scientific payload. O-A3 and O-A4 are mainly chosen due to the fact that the system can be used in the Mother-Child scenario as depicted in Fig. 5. O-A5 addresses the problem that depending on the crater, the system has to operate par-tially in the dark, which means that solar power alone is no option and other means like batteries or RTGs

have to be used. The latter is, due to the problematic handling and the pure weight in comparison to the power output, not an option for a mini-robot like the planned Spaceclimber. There are of course other requirements for the system, e.g. to provide communication to the operator segment etc.., but they are more generic and not legged-robot or crater-mission specific and therefore not listed here. 4. The DFKI Crater Exploration Test Bed To develop a system with compliance to the above mentioned requirements, serious testing is necessary. Therefore, the DFKI set up a crater exploration test bed which at the moment is mainly used in the LU-NARES project (funded by the DLR and BIG Bre-men). The aim of this project is the evaluation of state-of-the-art robot technologies for future cooperative, heteroge-neous, extraterrestrial missions with reconfigurable robots. In cooperation with our partners, a reconfigur-able robot system consisting of a Lander (OHB) with a manipulator, a Rover (EADS Astrium), and the climb-ing robot SCORPION (DFKI) will be developed based on already existing robot systems. To test its versatility and robustness, a replication of a lunar crater was developed.

Fig. 6 CAD design of the crater test bed showing also the two light sources which will be used.

Average 30-35° slope

45° slope

Visitor area

Fig. 7 Upper part of the crater comprising mini- and micro-craters as well as rocks. The design is based on the lunar crater analysis. The CAD sketch of the test bed is shown in Fig. 6, the depicted room including the visitor area on the right, In Fig. 7, the upper part of the crater is shown. It re-sembles very well the structure of a real lunar crater. The design of the test bed is very flexible and allows to use different substrate-like basaltic split of different sizes which reasonably resembles Regolith. Further-more, we can add fixed as well as loose rocks on the crater slope to change the difficulty from a simple cra-ter to a more complex one. To test also different angles of inclination, the test bed starts on the left with an inclination of about 30° and ends on the right with an inclination of 45°.

Fig. 8 Light test on the basalt surface with special illumination for high contrast. The angle of the light was 5°, which results in the typical long shad-ows. To model realistic light conditions, we are illuminating the test bed with two light sources. Because of side shutters, we can prevent the problem of core- and half- shadows. Due to a very high color temperature of 6000K and a high luminance of 14500 lux at 12°/10m, we get high lunar-like contrast to challenge our vision systems (an example is shown in Fig. 8). For an analysis of the performance of the robot, we have integrated a 6D high speed IR-tracking system, which can track up to 150 different markers at the same time. In addition, we have integrated 4 pan-tilt motor zoom cameras to track all operations on video. For gravity simulation, we have a 2 axis portal crane available which can carry 200kg and move with 1m/s. Via a special mechanism, this can be used to move parallel to the center of mass of the robot and attach a counter weight to reduce the weight of the robot. Of course, this is only a very limited lower gravity emula-tion (e.g. because the ground substrates behaviour does

not change) but it gives very good hints as to what we have to expect if the robot has to face different gravity conditions. For more realistic tests under such condi-tions, we have established a simulation for modelling our systems. 5. Using Simulation and Learning Especially in the concept phase, a simulation is of high benefit. It allows to test at a very early state if a robot concept is feasible. This is of very high importance in the design of legged systems because they have a high degree of design parameters, e.g. joint torques, seg-ment length, attaching position for the legs, etc.). The developed multi-body simulation allows testing the robot in static as well as in dynamic (e.g. walking) situations and allows among other signals (e.g. body pitch and roll angle) to derive average as well as peak torques in the joints. Because of the high degrees of freedom, it is very dif-ficult and time-consuming to manually optimize and test different robotic concepts. Therefore, an automated optimization plug-in has been developed which allows changing a parametric model and testing its performance in different benchmark tests, e.g walking on flat or elevated ground. Based on certain goal functions (e.g. minimizing the energy con-sumption and maximize the stability), a learning algo-rithm can be used to optimize the structure of the sys-tem, for example, to find the best segment length for the legs. At the moment, this new design approach where learning algorithms are integrated into the con-cept feasibility analysis looks very promising, but it is too early to give already detailed results. 6. Resume and Next Steps We reviewed the requirements for a crater exploration robot and discussed the implications for a legged solu-tion. It is clear that the system needs a high dexterity to cope with the varying ground conditions. Furthermore to be able to walk freely in a slope it must be as flexi-ble as possible and combine the advantages of more specialized solutions like the ARAMIES and Scorpion robot. At the moment different concept options to achieve this flexibility are traded and evaluated regard-ing their feasibility by using our simulation and opti-mization framework. As soon as these steps are done we will go into the detailed design of the subsystems, which to some extend already started now to achieve a concurrent design a full system which meets the re-quirements. We plan to have a first breadboard system

ready by end of 2008 and the full functional prototype by end of 2009. 7. Acknowledgments The presented work was done by the Spaceclimber-Team to which belong apart from the authors, Fran-cesco Allegrini, Frank Beinersdorf, Sebastian Bartsch, Dr. Stefan Bosse, Florian Cordes, Felix Grimminiger, Jens Hilljegerdes, and Steffen Planthaber. The Spaceclimber project is funded by DLR (Fkz: 50RA0705) & ESA (ESA Contract 18116). The LUNARES project is funded by DLR (Fkz: DLR 50RA076) & BIG Bremen (Fkz: BIG INNO1036A). This project builds on the experiences of the ARA-MIES project (DLR Fkz 50JR0561 & ESA Contract 18116/04/NL/PA) and the SCORPION project (DARPA Grant No. N0014-99-1-0483 & NASA USRA Grant No. 8008-003-002-001).

8. References [1] Next Lunar Lander with In-Situ Science: Phase A Mis-sion Study, Mission Requirements Document (NEXT-LL-MRD-ESA(HME)-0001) ESA – The NEXT Team, Issue1, Revision 2, 2007 [2] D. Spenneberg, F. Kirchner ‘’ The Bio-Inspired Scorpion Robot: Design, Control & Lessons Learned’’, In Climbing & Walking Robots(Book), Ed: Houxiang Zhang, I-Tech Educa-tion and Publishing, 2007 [3] [D. Spenneberg, S. Bosse, J. Hilljegerdes, F. Kirchner, A. Strack, and H. Zschenker. “Control of a bio-inspired four-legged robot for exploration of uneven terrain”. In Proceed-ings of ASTRA 2006,.ESA 2006.