Permanencia en la vereda, según los sectores

2.5

Chapter Summary

The review of robot locomotion shows that the wheel-leg concept is a readily available tech- nology which can significantly enhance the mobility of wheeled rovers on planetary surface exploration missions while offering a simpler, more reliable and more efficient alternative than legged locomotion at its current stage of development.

The review of soil sensing techniques indicates that, while necessary for mission planning and rover navigation, the usefulness of remote sensing for this research is limited due to its low spatial resolution, high computational requirements and inability to detect non- geometric hazards. Direct in-situ soil sensing devices do not suffer these issues, but require high power and long rover-stopping times. Indirect in-situ sensing techniques can avoid these drawbacks by analysing on-the-fly the interaction between the rover and the terrain, e.g. using thoroughly researched wheel slip and sinkage detection methods. However, the fact that non-geometric hazards are only detected upon traversal suggests the interest of addressing the unsolved problem of on-line wheel-leg slip and sinkage detection for soil characterisation and trafficability assessment, in order to prevent the rover from getting stuck in potential non-geometric hazards thanks to the higher mobility of wheel-legs.

The review of vehicle-terrain interaction modelling reveals abundant studies and approaches to simulate wheel-soil and leg-soil interaction, while little focus has been given to hybrid wheel-leg devices so far. The scarcely available models and studies do set a good basis for wheel-leg-soil interaction modelling. However, they raise unanswered research questions concerning: the effect of multi-legged wheel-legs on soil interaction, the relation between soil physical properties and the parameters of penetration force models, the relationship between slip and sinkage phenomena in wheel-leg-soil interaction, the influence of different foot designs and the comparability to classic Terramechanics approaches.

The review of terrain classification applications encourages their extension to the unexplored use of wheel-leg slip and sinkage indicators to classify dry, granular soils. The suitability of different classification algorithms to this specific problem needs to be compared. Overall, the literature review carried out confirms the motivation and novelty of this study and identifies the underlying principles and research niches addressed in this research.

Chapter 3

Wheel-Leg System Development

and Testing Methodology

The two main physical systems involved in this study are the wheel-leg system, formed by the rover’s wheel-leg locomotor and the hardware and software needed for wheel-leg driving and slip and sinkage sensing, and the terrain being traversed, fully characterised in terms of its physical properties. This chapter deals with the design and implementation of the wheel-leg system and the selection and characterisation of the soil types used to represent the deformable terrain addressed. These contents are introduced ahead of any theoretical modelling or algorithm development partly because they are strongly pre-conditioned by the mission requirements of the FASTER project. But the main reason behind it is that the definition of these materials and their quantitative and qualitative characteristics are pre-requisites for the elaboration of those models and algorithms.

Experimental work is paramount for this research, at different stages and with a variety of purposes. Preliminary, exploratory tests provide a practical insight to the mechanisms and phenomena involved in the interaction between a multi-legged wheel-leg and dry sand. Rigorous, extensive testing campaigns in controllable and repeatable conditions facilitate solid evidence to find empirical correlations or refine theoretical models based on analytical considerations. Finally, independent tests are required for system and algorithm validation, so as to evaluate the performance and accuracy of the solutions and models developed.

3.1. Wheel-Leg and Foot Designs

Therefore, the materials, set-ups and methodologies used for the experimentation phase of the research are also introduced in this chapter. This includes laboratory and field testing on a variety of soil types and conditions, using both simplified test beds, for empirical exploration and analysis, and a fully mobile wheel-legged rover, for final validation.

3.1

Wheel-Leg and Foot Designs

The key component of the robotic system addressed in this research is the wheel-leg itself, shown in Fig. 3.1 (a). It is closely based on previously used wheel-legs and further details of its design and fabrication processes can be found in [42, 126]. It counts with five uniformly distributed spokes, with 180 mm length and 14 mm thickness, each consisting of two full- length solid edges, bridged with the neighbouring legs for rigidity, and hollow in the middle. While slightly compliant, the high stiffness provided by the design makes the deflection negligible under the loads estimated for the FASTER SR used in this research (3.75 kg per wheel-leg).

(a) (b) (c)

Figure 3.1: Wheel-leg equipped with one FRF, two CIF and two LTF wheel-leg (a), individual pictures (b) of an FRF (bottom), an LTF (centre) and a CIF (top) and two detailed isometric CAD views of the LTF design (c). The wheel-leg, CIF and FRF devices are courtesy of DFKI, Bremen

3.1. Wheel-Leg and Foot Designs

Interchangeable feet can be attached at the end of each leg. Three different designs are considered in this research, pictured in Fig. 3.1 (b). The Flexible Rubber Feet (FRF), similar to those used by the Asguard robots [126], provide good load distribution, terrain adaptability and traction. The “Camel Inspired” Feet (CIF), similar to those used by CESAR [42], have a much bigger contact area, considerably lowering the contact pressure in deformable terrain.

Finally, the Load Testing Feet (LTF) were specifically designed for the EU FP7 FASTER project. Their main design driver is to replicate the static load contact pressure below one of the 10 cm-wide and 15 cm-radius wheels of the 350 kg FASTER PR with the reduced 15 kg mass of the SR. Figure 3.2 (a) plots as a blue line the average contact pressure below a PR wheel in Earth’s gravity, geometrically calculated for different sinkage levels in the 0-120 mm range. The initial width of the LTF is calculated to mimic the contact pressure with a 50% safety margin, shown as a dashed blue line, to ensure that the LTF would penetrate through any duricrust that is weak enough to break below the pressure of a PR wheel.

(a) (b)

Figure 3.2: Pressure-sinkage curves geometrically calculated for PR wheel and a SR LTF and ex- perimentally obtained for firm ground, quartz sand and nepheline powder (a) and static wheel-leg pressure-sinkage test set-up (b)

3.1. Wheel-Leg and Foot Designs

Due to the low mass of the SR, the resulting width is overly narrow. As a cautionary measure, to prevent excessive sinkage, the depth of this narrow blade is limited to 7 mm. This is enough to penetrate any sub-centimetre weak surface duricrusts as those reported by earlier Mars missions [150, 151], and acts as a maximum sinkage threshold for certainly safe terrain (‘GO’). Thereafter, the LTF is provided with a wider plateau, whose thickness is calculated to mimic the static pressure of a PR wheel at a sinkage of 75 mm. This is the specified maximum allowable sinkage for the FASTER PR, equal to half the wheel’s radius, and acts as the minimum sinkage for certainly unsafe terrain (‘NO-GO’). Any sinkage level in between those thresholds is treated as uncertain (‘MAYBE’), as the contact pressure beneath a SR wheel-leg is significantly smaller than below a PR wheel in that region.

The final design, shown in two CAD isometric views in Fig. 3.1 (c), is asymmetric both longitudinally and transversally. The rear part of the foot is cut-off and rounded, to avoid excessive digging of the heel on foot impact with the ground and to mimic the negligible rut recovery [99] of rigid wheels on deformable terrain. The blade is placed adjacent to one edge of the foot rather than centred, laying flat with the inner face of the leg, so that the whole profile of the foot is observable to enable vision-based sinkage detection in the full range. The variable, stepped width of the blade aims at accounting for the variations in the attack angle of the leg to maintain the desired contact pressure at low sinkages. However, the design of this feature lies beyond the scope of this thesis, and its effect is negligible with sinkages above 7 mm.

As a result of this design, the geometrically calculated contact pressure below an LTF loaded with the SR nominal mass is shown as a black line in Fig. 3.2 (a). The red, yellow and green lines in the same plot correspond to experimental measurements of quasi-static pressure-sinkage tests on different types of materials using the LTF and the modified PST set-up shown in Fig. 3.2 (b).

The initial peaks in the experimental curves correspond to the high pressure due to the small contact area of the blade. This is followed by a drop due to the higher contact area of the plateau and a constant increase thereafter, as expected from the pressure-sinkage behaviour of soils. The intersections between these lines and the geometrically calculated ones (triangle and star markers) correspond to the expected static sinkage of a SR wheel-leg