The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #93136

Title:Robots as Molecules: Statistical Mechanics for Velocity Control
Date: 1993
Abstract:Efficient locomotion is a crucial competence for any mobile system. The traditional approach to achieving this has been to employ path planning to produce a collision free path. however, efficient locomotion can also emerge from uncoupled obstacle avoidance and goal-directed competences, but this method is unable to guarantee efficient locomotion because it throws away the possibility of using global control. There is a tradeoff between time spent navigating towards a goal and time spent avoiding obstacles; obviously, one would like to maximize the former and minimize the latter to achieve locomotion which is as efficient as possible. This dissertation describes an investigation of the uncoupled obstacle avoidance and navigation strategy with an added velocity control module. By applying insights from statistical mechanics in gases, a model for velocity control in the robot is designed. If the velocity of a robot can be adjusted as a function of obstacle density, locomotion efficiency can be increased by allowing more time to navigate. The main concept is the mean free path for molecules in a gas. The mean free path can be predicted by statistical means, and the concept can be applied to robot/molecule and obstacles in an environment by simulation. The paper addresses kinetics in a gas, the implementation of a simulator (the AxS Robot Simulator), the experiments carried out using the simulator and the results and analysis of the findings.

[Search These Pages] [DAI Home Page] [Comment]