Off-Nominal Rover Driving: Terrain Manipulation and Degraded Mobility Compensation

This dissertation develops new mobility and manipulation capabilities for planetary exploration rovers through comprehensive wheel-soil interaction modeling and then demonstrates these new techniques on rovers. 

While wheeled rovers are the current paradigm for robotic space exploration, wheeled mobile robot performance is only well understood for normal driving conditions. A better understanding of wheel-soil interaction lets us do more with rovers in two key ways: 1) add manipulation capabilities through the use of nonprehensile manipulation, e.g. by using wheels to dig, and 2) retain mobility when experiencing degradation of the mobility system, e.g. loss of a wheel motor. Both of these areas require knowledge of wheel-soil interaction forces beyond the scope of existing methods. 

First, we present the concept of Nonprehensile Terrain Manipulation (NPTM) and illustrate the scope of potential applications for planetary exploration rovers. We then select one NPTM action, wheel-based trench excavation in soft soil, as a candidate action that is achievable on current rovers with no hardware addition but that requires new modeling. 

Next, we present a closed-form model of soil flow around a wheel driving in regolith that can be used both to model trenches dug by rover wheels and to improve terramechanics models. We then detail a new terramechanics model that covers all slip angles and all ranges of slip and skid and validate it with tests on two wheels operating over a wide range of states. 

Then, we demonstrate the feasibility of NPTM, the need to account for mobility system failure, and the viability of recovering from failure on full-scale rovers operating in a variety of environments. This is done through a series of demonstrations on NASA rover prototypes in lunar simulant and a Martian analog environment. 

Finally, we implement an optimization framework to automatically generate driving strategies for both digging trenches in soil and recovering from multiple types of mobility system failure. We demonstrate the generated driving strategies on a miniature rover in soft soil and use the optimization to predict overall rover mobility in these modes.