Avoiding Inevitable Collision States: Safety and Computational Efficiency in Replanning with Sampling-based Algorithms

TitleAvoiding Inevitable Collision States: Safety and Computational Efficiency in Replanning with Sampling-based Algorithms
Publication TypeReport
Year of Publication2010
AuthorsBekris, KE
Series TitleWorkshop on "Guaranteeing Safe Navigation in Dynamic Environments", International Conference on Robotics and Automation (ICRA-10)
Date PublishedMay 2010
CityAnchorage, AK
Abstract

Safety concerns arise when planning for systems with dynamics among moving obstacles, where a collision-free trajectory leads to an Inevitable Collision State (ICS). Identifying whether a state is ICS, however, is computationally challenging. This has led to approximations, varying from conservative schemes, which never characterize an ICS as a safe state, to schemes with weaker guarantees but fast online resolution of an ICS query. The computational cost of the approach is critical in problems that require replanning. This report presents various alternatives for identifying whether a state is ICS from the related literature. It also discusses different ways for integrating such schemes with sampling-based planners in safe replanning frameworks so as to reduce the computational overhead of avoiding ICS.

URLhttp://www.cs.rutgers.edu/~kb572/pubs/ics_tradeoffs.pdf