Generation and Organization of Behaviors for Autonomous Robots

摘要

In this thesis, the generation and organization of behaviors forautonomous robots is studied within the framework ofbehavior-based robotics (BBR).Several different behavioral architectures have been considered inapplications involving both bipedal and wheeled robots. In thecase of bipedal robots, generalized finite-state machines (GFSMs)were used for generating a smooth gait for a (simulated) five-link bipedalmodel, constrained to move in the sagittal plane. In addition,robust balancing was achieved, even in the presence of perturbations.Furthermore, in simulations of a three-dimensional bipedal robot,gaits were generated using clusters of central patterngenerators (CPGs) connected via a feedback network. A thirdarchitecture, namely a recurrent neural network (RNN), was usedfor generating several behaviors in a simulated, one-legged hoppingrobot. In all cases, evolutionary algorithms (EAs) were used foroptimizing the behaviors.The important problem of behavioral organization has been studied usingthe utility function (UF) method, in which behavior selection isobtained through evolutionary optimization of utility functions thatprovide a common currency for the comparison of behaviors. In general,the UF method requires the use of simulations. Thus, an importantpart of this thesis has been the development of a general-purposesoftware library (UFLib) implementing the UF method. In order to study the properties of the UF method, several behavioralorganization problems, mostly involving wheeled robots, have been considered. Most importantly, it was found that the UF methodgreatly simplifies the search for solutions to a wide varietyof behavioral organization problems and requires a minimum of hand-coding.Furthermore, the results show that the use of multiple simulations(for the evaluation of a robot) significantly improves the abilityof the robot to select appropriate behaviors. For the EA, it was found that the standard crossover procedure, whichswaps entire utility functions between individuals, performed at least as well as several modified operators, and that the mutation rateshould be set so as to generate around three parameter modificationsper individual. Finally, some early results are presented concerning the use of the UF method in connection with a robot intended fortransportation and delivery.