System of Terrain Analysis, Energy Estimation and Path Planning for Planetary Exploration by Robot Teams

Publication Year:
2012
Usage 747
Downloads 604
Abstract Views 143
Repository URL:
https://ir.lib.uwo.ca/etd/1078; https://ir.lib.uwo.ca/cgi/viewcontent.cgi?article=2319&context=etd
Author(s):
Michel, David C
Tags:
Path-Planning; Energy Estimation; Soil Identification; Planetary Exploration; Robotics
article description
NASA’s long term plans involve a return to manned moon missions, and eventually sending humans to mars. The focus of this project is the use of autonomous mobile robotics to enhance these endeavors. This research details the creation of a system of terrain classification, energy of traversal estimation and low cost path planning for teams of inexpensive and potentially expendable robots.The first stage of this project was the creation of a model which estimates the energy requirements of the traversal of varying terrain types for a six wheel rocker-bogie rover. The wheel/soil interaction model uses Shibly’s modified Bekker equations and incorporates a new simplified rocker-bogie model for estimating wheel loads. In all but a single trial the relative energy requirements for each soil type were correctly predicted by the model.A path planner for complete coverage intended to minimize energy consumption was designed and tested. It accepts as input terrain maps detailing the energy consumption required to move to each adjacent location. Exploration is performed via a cost function which determines the robot’s next move. This system was successfully tested for multiple robots by means of a shared exploration map. At peak efficiency, the energy consumed by our path planner was only 56% that used by the best case back and forth coverage pattern.After performing a sensitivity analysis of Shibly’s equations to determine which soil parameters most affected energy consumption, a neural network terrain classifier was designed and tested. The terrain classifier defines all traversable terrain as one of three soil types and then assigns an assumed set of soil parameters. The classifier performed well over all, but had some difficulty distinguishing large rocks from sand.This work presents a system which successfully classifies terrain imagery into one of three soil types, assesses the energy requirements of terrain traversal for these soil types and plans efficient paths of complete coverage for the imaged area. While there are further efforts that can be made in all areas, the work achieves its stated goals.