Energy-Efficient Path Planning for Vehicles Using Static Terrain and Elevation Data

This report conducts a comparative performance analysis of Artificial Potential Fields (APF) against Q-Learning for efficiently navigating static elevated terrains. In the real world, often agents need to navigate rugged environments with power consumption constraints. With the aid of Q-Learning and APF, I determine general approaches for agents to navigate uneven real world environments from a known source and destination. In this paper, I start with a review of the current literature on efficient path planning, build a model to demonstrate these path-planning algorithms in action, and conclude with simulation results to show the effectiveness and efficiency of the different path-planning algorithms. Path planning is the problem of finding a collision-free path for an agent from a starting point to a destination in an efficient manner. Efficiency for the purposes of this paper is defined as minimizing the amount of energy consumed to get from the source to the destination. All the simulation code is written in Python and is available in Appendix B - I. Terrains are represented as grayscale 2-dimensional images corresponding to elevation. The height maps are real land masses collected from the web[1]. I use the combination of multiple equations to create a cost function to analyze and understand how the cost (energy consumption) of a given path varies with respect to the cost function parameters for each model. Using the results from select runs on the height maps, I provide my findings and the potential positive and negative aspects of each methodology. The goal is to use these findings to make navigation inferences for vehicles in real world environments that will minimize energy consumption.