This effort will extend the domain of applicability of existing nonholonomic trajectory-generation algorithms from indoor to outdoor rough-terrain environments. Additionally, existing local-navigation algorithms that perform full-configuration space search under typical wheeled-vehicle nonholonomic constraints will be extended to exhibit real-time replanning and to apply directly to rover rocker-bogy mobility systems.
Under a collaborative agreement with investigators at CMU, we are integrating the developed software with the CLARAty architecture during the term of the program in each of three predefined milestones.
Experience with long-traverse terrestrial autonomous vehicles clearly indicates that high-fidelity models of terrain following and vehicle maneuverability are necessary in both the trajectory-generation and the local-motion-planning levels of autonomy. Therefore, the work is expected to significantly improve the performance and enhance the utility of rovers in both autonomous and semi-autonomous modes of operation. In particular, the results of the work are expected to have a large impact given the long excursions that are being contemplated in the MSL mission.
JPL has a collaborative role for this effort, which is led by Alonzo Kelly of Carnegie Mellon University.