Abstract
The use of mobile robots has formed the core of numerous exploration missions to the Moon and Mars. However, estimation of rover traction on the loose, granular regolith found widely on the lunar and martian surfaces remains challenging. While it is relatively easy to describe the forces imparted on the terrain by a rover, characterisation of the terrain response and its trafficability has proven to be more difficult. The validation of future rover designs and laboratory-based experiments to aid current mission operations will rely on a comprehensive understanding of the deformation mechanics of regolith-like materials. This thesis presents a novel terrain characterisation methodology and analysis which extends current approaches to trafficability analysis using the theory of terramechanics. The presented approach primarily focuses on identifying the effects of changes in relative density on the deformation behaviours of frictional soils under low loading conditions. Motivated by previous work that produced Surrey Space Centre (SSC) regolith simulants, a study of in-situ martian and lunar regolith properties is presented. A classification method by particle size distribution (PSD) is demonstrated and three Engineering Soil (ES) simulants suitable for use with the ExoMars rover testbed are selected based on the PSD. Preparation and characterisation methods are developed to provide samples of each simulant at three discrete relative densities and measure their shear and normal load failure responses. Finally, these data are used to study the trafficability characteristics of two simulants using a simple microrover single wheel testbed. Direct shear tests using samples of the ES simulants demonstrated little sensitivity in the residual stress to initial density. However, the magnitude of the simulant brittleness (the brittleness index) was found increase with PSD under low loading conditions similar to those of the ExoMars rover. Furthermore, the sensitivity of the brittleness index to sample density demonstrated that the measured residual stress does not reflect the critical state strength in high PSD simulants. Subsequent pressure-sinkage tests demonstrated that failure planes developed due to normal load do not reflect the residual strength of the soil. Therefore, where brittleness index is greater than approximately 33 %, the standard Bekker/Reece-Wong pressure-sinkage curves do not fully capture the normal failure response. By splitting the pressure-sinkage data into pre- and post-failure regions, parameter goodness of fit was found to improve by up to 13% in high brittleness ES-3. A more modest improvement of up to 3% was seen in the lower brittleness SSC-2. Using these parameters, model performance predictions were computed and compared with preliminary single wheel testbed data. While the precision of the data support the presented methodologies, future improvements to the apparatus and additional data will aid in verifying the accuracy of these measurements.