Applying modern soil mechanics to small mobile robots

Abstract

Developing a robot that can traverse diffcult terrain requires many decisions to be made during the design stages. Numerous locomotion strategies exist and many architectures of each strategy can be created. We introduce a model that provides a platform for comparing these locomotion methods by calculating their expected energy use in traversing compliant terrain. Furthermore, our method can be used to aid the operation of a robot. A large proportion of the locomotion strategies in the literature include additional methods to improve their effcacy in diffcult terrain. A wheel-leg hybrid robot is a point in case. Our model can be used to determine when it is most effcient to employ these augmentations. To fnd the energy requirements of a terrain, the resistance-to-motion the robot faces is determined. Our model focuses on two locomotion mechanisms, wheels, and singlelink single-joint rotary legs. The method developed within this thesis also shows much promise for the analysis of other locomotion strategies. For wheeled locomotion this approach allows the modelling of the terrain interaction with both driven or towed wheels. Our model is a signifcant improvement for small mobile robots, those weighing less than 50 kilograms and with a wheel diameter or leg length under half a metre, for which the state of the art fails to accurately determine dynamic sinkage or the resistance to motion. The proposed model addresses many of the issues apparent with the classical terramechanics approach to calculating resistance to motion. Primarily these models were created for large agricultural or military vehicles, those with wheel diameters greater than half a metre and weighing over half a tonne. Additionally the classical model decouples the effects of multi-directional loading on the terrain. These both contribute to inaccurate predictions of the terrain interaction for small mobile robots. We address these problems by utilising the soil strength failure envelope method to determine bearing capacity. This approach more accurately models the interlinked forces and moments that exist at the soil-wheel/soil-leg interface. The solution amalgamates various physical phenomenon previously calculated independently, such as compaction, bulldozing, and rolling resistance, greatly simplifying the analysis. The number of parameters is also signifcantly reduced to a single term for undrained soils and two terms for drained soils. These parameters are well established in the soil mechanics literature. The model is augmented with a term that evaluates the work required to overcome unavoidable, small, repeated obstacles found in diffcult terrain such as uneven ground or fora. A natural scaling law is used to predict the distribution of various sized obstacles and the equivalent average resistance-to-motion that these obstacles present. As a whole, the model predicts the minimum work needed to traverse a particular terrain populated with repetitive obstacles as might be found in environments of interest such as forests, felds, rubble, deserts, or tundra. Additionally, feld testing in a forested environment with a wheeled robot and the development of a wheel-leg unit during the undertaking of this research ensures the model has a practical framing. Within this thesis is presented a novel method to evaluate a robot’s resistance-tomotion. This model represents a signifcant contribution to the feld as it replaces terramechanics models which have little practical use due to their inaccuracy and which are misleading for small robots. Our model can be utilised to select physical robot parameters such as wheel radius, or width as well as aiding in the operation of locomotion augmentations. Additionally, our model provides a platform for comparison of different robot architectures where no previous adequate platform exists. We believe our model will prove useful to the robotics community as it is the frst application of modern soil mechanics to determine resistance-to-motion. The increased fdelity to the soil interaction enables such a model to be utilised in both design and operation of small mobile robots.

More information

Identifiers

Catalogue record

Export record