Autonomous Navigation in Unknown Environments using Sparse Kernel-based Occupancy Mapping

This paper focuses on real-time occupancy mapping and collision checking\nonboard an autonomous robot navigating in an unknown environment. We propose a\nnew map representation, in which occupied and free space are separated by the\ndecision boundary of a kernel perceptron classifier. We develop an online\ntraining algorithm that maintains a very sparse set of support vectors to\nrepresent obstacle boundaries in configuration space. We also derive conditions\nthat allow complete (without sampling) collision-checking for piecewise-linear\nand piecewise-polynomial robot trajectories. We demonstrate the effectiveness\nof our mapping and collision checking algorithms for autonomous navigation of\nan Ackermann-drive robot in unknown environments.\n