Abstract:
Over the last two decades, we have seen driverless cars conquer the Mojave desert, drive on mars and operate on our streets and warehouses. One of the most fundamental requirements of such robots is their ability to navigate their environment with minimal human oversight.

As more robots graduate from the confines of laboratories to real world, it is inevitable that their assumptions about the environment that they operate are violated in more ways than expected. In many situations, rather than purely being reactive, having the predictive ability to understand the effects of its actions on future states helps the robot from making poor decisions. We address the problem of path following in challenging terrain using a Model Predictive Control framework.

Sometimes, harsh environmental conditions pose further restrictions in terms of Size, Weight and Power on the robot’s hardware. We look at the use case of local planning for the M2020 rover where severe computational constraints render most of the algorithmic advances in robot motion planning unusable. We discuss an adaptive, cache-based, meta-planning algorithm that produces high quality solutions, uses less resources and in many cases, runs faster than the baseline algorithm.

Even the best engineered robotic system requires the ability to adapt to new situations that it encounters. Model adaptation is a double-edge sword where quick changes to the model needs to be carefully traded-off with control instability and the dangers of losing past experiences. In a complex system such as a wheeled mobile robot on challenging terrain, blame assignment is a necessary feature. We discuss preliminary results in characterizing a state-of-the-art slip model identification with data from a high-fidelity vehicle dynamics simulator. We conclude with a list of proposed tasks: a) improvements to the slip model identification to deal with low mu surfaces, b) evaluating and implementing an adaptive controller framework that uses the improved slip model for wheeled mobile robot navigation.

Thesis Committee Members:
Alonzo J. Kelly, Chair
David Wettergreen
Matthew T. Mason
Oktay Arslan, Airbus

Thesis Proposal Document