No label defined (Q4510178)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 1520611
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 1520611

      Statements

      0 references
      23 October 2000
      0 references
      collision avoidance
      0 references
      evolutionary algorithms
      0 references
      automatic design of robots
      0 references
      parametrized module configuration graph
      0 references
      metrics
      0 references
      Pareto-optimal individuals
      0 references
      mobile robots
      0 references
      Using evolutionary algorithms for the automatic design of robots seems to be an approach both daring and very natural. In this book -- his Ph.D. thesis at Carnegie Mellon -- Chris Leger is the first to present a design tool that provides evolutionary methods for the synthesis of very different kinds of robots. In the first part of the book the design tool, Darwin2K, is presented; the second part reports on experimental results. The design tool is based on a presentation of the robot by a set of parametrized modules. By the (rigid) physical connections between adjacent modules, the structure of an acyclic directed graph (the parametrized module configuration graph, PMCG) is introduced. This approach allows certain local features to be encapsuled in the modules while, at the same time, the use of parameters yields simpler descriptions than an enumeration of variables would do. Of the well-known genetic operators, the system uses crossover on both parameter strings and parts of the PMCG structure. Similarly, mutation applies to parameters, to properties of the connection between modules, and, by insert, delete, and permute, to the module structure. The fitness of an individual construction can be evaluated by various metrics, among them completion time for the task specified, load, power consumption, maximum torque on joints, etc. In general, Pareto-optimal individuals survive, if their qualities with respect to these metrics are above certain feasibility thresholds to be specified by the designer. Also, the designer can raise the importance of certain metrics if a certain percentage of the population is below the feasibility threshold. Metric computation requires both kinematic and dynamic simulation, e.g. for torque, mass, and link deflection. One of the metrics addresses self-collisions and collisions with obstacles. The corresponding collision detector and motion planner components of the simulator use a geometric module description by polyhedra. For planar mobile robots specified by one or more points, a cell decomposition of the workspace is used, and the standard Voronoi diagram structure is induced on these cells by waves expanding from the obstacles. I found reading this book quite interesting. Its main contribution is in providing engineers with an extendible, flexible system that has already performed well in some applications. The book is very nicely written and well structured.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references