4. LAS NARRATIVAS DEL TERRITORIO
4.3. El presente de la vereda Siete Trojes: cambios de tiempo y espacio
(a) (b)
Figure 2.6: Wheel-legged locomotion: (a) ESA/EPFL’s SpaceCat [46] and (b) IMPASS robot [47]
2.2
Soil and Terrain Sensing Technologies
While it is important to study the range of available locomotion concepts and the mobility of conventional wheels has a central role in this research, the research objectives and novelties formulated in Chapter 1 don’t directly address wheel-leg design and mobility analysis. Instead, they have a strong focus on assessing the physical properties of the terrain by using the wheel-leg concept as a soil sensor in and of itself. Hence, a review of the state of the art in soil sensing is also necessary.
Using soil sensing techniques during rover navigation allows avoiding dangerous situations, optimising the traversal speed and power use and providing additional data of scientific interest about the terrain, hence maximising the scientific return and performance of the mission. The range and nature of the sensed terrain characteristics varies depending on the type of sensor used, as analysed in the survey performed in [48]. The strengths and weaknesses of each sensing technique make them useful for different purposes at different stages of the mission. The following review analyses the capabilities and limitations of each technique to identify the most suitable option for the purpose of this research.
Remote sensing is based on measuring the reaction of the terrain to different types of Electro Magnetic (EM) waves and relating the measurements to terrain properties. These sensors can provide data from a large extension of terrain, but with low resolution when mounted on orbiters. Some of them are mounted directly on rovers, providing a higher resolution but a more limited area.
2.2. Soil and Terrain Sensing Technologies
(a) (b)
Figure 2.7: Remote soil sensing: (a) infra-red thermal inertia [49] and (b) stereo-vision feature tracking on MER Spirit rover [53]
Infra-Red EM waves can be used to estimate the thermal inertia of the terrain [49, 50, 51], as shown in Fig. 2.7 (a). Radio EM waves can reveal geophysical properties of the subsurface, e.g. using the WISDOM GPR [52] under development for ExoMars. Visual EM waves can be used by cameras to extract images of the terrain, which yield very useful information for texture classification, feature detection (e.g. rocks) or stereo-vision [53], as shown in Fig. 2.7 (b). Light Detection and Ranging techniques (LIDAR) using laser scanners can be also used to obtain direct 3D data of terrain topography.
In-situ direct sensing performed by lander probes or rovers measures soil resistance forces under given loading conditions to determine a set of physical or empirical parameters that characterise the terrain. Some of the devices that can be used for this purpose are illustrated in Fig. 2.8.
The Static Cone Penetrometer (SCP) is a standard tool in terrestrial geotechnical surveying used to calculate the Cone Index (CI) as a measurement of soil strength and stiffness, translatable into a Vehicle Cone Index (VCI) as described in [57] for an estimation of terrain trafficability. For robot-mounted applications alternative actuation mechanisms are conceived to reduce the total mass an power needed, such as the high frequency percussive actuator used by the Percussive Dynamic Cone Penetrometer (PDCP) [58] and the self- propelled spring-shock mechanism used by the MOLE [59] and MMUM [56] devices. Bio- inspired low-power drilling mechanisms [55, 60] can also be used to measure soil resistance forces and strength.
2.2. Soil and Terrain Sensing Technologies
(a) (b)
Figure 2.8: In-situ direct soil sensing: (a) bevameter plate and shear test tools (top) [54] and dual-reciprocating drilling (bottom) [55] and (b) MMUM device [56]
The Bevameter test proposed by Bekker [61] combines plate loading and shear loading devices to replicate the stress conditions created by a wheeled vehicle, allowing to estimate semi-empirical parameters of the Bekker vehicle-soil interaction model. This type of tests yield a detailed characterisation, but even compact designs like the one proposed by PNFI in [54] are complex and difficult to implement autonomously on planetary rovers.
Alternatively, in-situ indirect soil sensing techniques can give an insight of the physical properties of traversed terrain without the need of directly applying and measuring forces with specialised equipment. Examples of these techniques are seen in Fig. 2.9.
Seismic refraction surveying techniques, described in [62] allow in-situ characterisation of large areas of terrain combining a seismic source and geophone sensors placed at known locations, revealing information about the stiffness and density of surface and subsurface soil or rock layers. In spite of its complexity in terms of communication and accuracy requirements this technique has been used in the Apollo missions to the Moon and several approaches to robot-based seismic refraction are under research, as described in [63].
2.2. Soil and Terrain Sensing Technologies
(a) (b)
Figure 2.9: In-situ indirect soil sensing: (a) seismic refraction with multiple robots [62] and (b) MER Opportunity wheel trenches analysis [64]
Finally, assuming the lack of dedicated sensors for terrain characterisation, other rover sub- systems can be used to infer data about the physical properties of the soil. Tactile and visual information from the mechanical interaction between rover devices, e.g. wheels or tools, and the soil can be post-processed to estimate some of the physical characteristics of soils and rocks in planetary environments. Data from the Lunokhod rovers [65], the Viking landers [66], the Soujourner rover [67] and the MER rovers [64] has been used to infer the friction angle, cohesion, density and soil type of lunar and Martian terrains, becoming an invaluable source of information when planning future missions due to the lack of dedicated instruments and direct measurements.
Each of the reviewed techniques has its own advantages and disadvantages. Remote methods are useful for mission planning and scientific research, but have limited resolution in orbiter- mounted sensors and are computationally intensive and power hungry in rover-mounted sensors. In-situ direct sensors can provide a detailed characterisation as well as useful scientific data about surface and subsurface composition and can even be used as a means to retrieve soil samples. However, these dedicated sensors imply significant additional mass, complexity, power and stopping time to carry out the test. Therefore, their use needs to be well justified. In-situ indirect sensors provide an effective and rapid means of assessing the performance and mobility of the rover on the soil currently being traversed.
2.2. Soil and Terrain Sensing Technologies
The challenges of this approach are to maximise the reliability of the soil-vehicle interac- tion characterisation, e.g. by combining several sensor modes, and to use the sensor data efficiently for on-line assessment. The latter depends on the algorithms used to detect situ- ations where mobility is at stake, i.e. when wheel sinkage and/or slip become too high. A brief review of research on sinkage and slip detection is presented below.
Sinkage is the result of deformable soil being compressed and displaced due to external loads. The preferred alternative for accurate sinkage detection is based on computer vision methods, where the wheel-soil interface is detected by a dedicated camera and used to estimate the level of sinkage as exemplified in Fig. 2.10. Approaches used so far are based on greyscale intensity to distinguish the wheel rim from the soil [68] and edge detection of a concentric black and white pattern [69]. Fewer solutions have been investigated for on-line detection of wheel-leg sinkage, such as the colour-based approach proposed in [15].
(a) (b)
Figure 2.10: Vision-based wheel sinkage detection: (a) rim intensity change [68] and (b) pattern edge detection [69]