Investigated the behavior of Motion planning algorithms under faults by implementing the Fault Injector tool in C++ using Intel’s PIN tool, which instruments dynamic faults during the compile time in the Motion planning Algorithm (RRT and RRT*).
Proposed application patching techniques (using C ) that allows faults to prevent catastrophic failures and increases the robustness of Motion Planning algorithms.
Identified that Motion planning algorithms tend to contain a large degree of inherent error tolerance (e.g. sampling from random distributions and calculating approximate distances).
Set of low overhead algorithm-based resilience techniques can be used to allow for the parts of application which are not inherently error tolerant (e.g. kd-tree control flow), to gracefully fail.
Presented research at IEEE Texas Workshop on Integrating System Exploration (TexasWISE) 2015 at Austin.