Real-Time Classification of Road Conditions
Abstract
Common navigation algorithms like A* or D* Lite rely on costs to determine an optimal path. Costs may incorporate distance, time, or energy consumption; however, they can include anything that affects travel along a path. Much research is done to improve planning algorithms based on a given cost, often without stating how to acquire that cost. Therefore, the focus of this research involves determining a method of accurately obtaining that cost in real-time by classifying environmental conditions. Specifically, this research employs K-Nearest Neighbor and Principal Component Analysis techniques to classify road conditions in order to determine the most informative parameters when measuring the cost of driving on those roads. This sensor-based classification approach may not only allow for improved automatic traction handling and path navigation, but also may be applied to any robotic system requiring real-time knowledge of environmental conditions.
Subject
Artificial IntelligenceMachine Learning
Robotics
Principal Component Analysis
Nearest Neighbors
Classification
Road Conditions
Paved
Unpaved
PCA
KNN
Citation
Weaver, Scott M (2016). Real-Time Classification of Road Conditions. Undergraduate Research Scholars Program. Available electronically from https : / /hdl .handle .net /1969 .1 /157644.