Skip to content

Latest commit

 

History

History
59 lines (49 loc) · 3.44 KB

developer_guide.md

File metadata and controls

59 lines (49 loc) · 3.44 KB
Raw

Jetleg Developer Guide

Nomenclature

Jetleg uses a couple of names different from the ones used in the literature:

  • IAR or compressed_hull ( or actuationRegions): refers to the feasible region(s) in Orsolino. et al. 2019
  • frictionRegion: refers to the support region in Bretl. et al. 2008;
  • actuation_polytope: refers to the force polytopes

Jetleg Flow

The communication:
  • Jetleg publishes to topics such as "/feasible_region/actuation_polygon" and ", "/feasible_region/support_region" , "/feasible_region/force_polygons" in the "communication.py" file (defined in init())
    • The publishers are named "pub_polygon", "pub_support_region", and "pub_force_polygons"
    • Polygon types are being published
  • CrawlCrontroller subscribes to the topics in the "/common_utilities/fr/src/FeasibleRegion.cpp" file.
    • The return variables are called "actuationRegionSub", "supportRegionSub", and "forcePolytopesSub"
  • Jetleg subscribes to the topic "*/hyq/debug" which are published in the logger
    • normals, possible foothold options, optimization_started flag, etc... are all received
The computation of the polygons:
  • talker() in "communication.py" calls "try_iterative_projection_bretl()" or "iterative_projection_bretl()" directly in "computational_dynamics.py"
    • Returns a compressed_hull (support/feasible region as requested), the actuation polygons (force polytopes), and the computation_time
  • "iterative_projection_bretl()" calls "setup_iterative_projection()" and then "pypoman.project_polytope()"
    • "setup_iterative_projection()" prepares the equality constraint (gravity action), inequality constraints (friction and actuation limits), actuation polygons (force polygons)
    • "pypoman.project_polytope()" is used to compute the vertices of the polygon. This is Caron's library and includes Bretl's iterative projection algorithm
Foothold planning:
  • talker() in "communication.py" calls "selectMaximumFeasibleArea()"
    • Returns foothold with max area (foothold_params.option_index), and the feasible regions (actuationRegions)

Important Jetleg Modules

Communication:
  • "examples/communication.py"
    • The main module that communicates with the C++ framework and calls the required computations
  • "jet_leg/iterative_projection_parameters.py"
    • Stores iterative projection parameters from "Debug" topic
  • "jet_leg/foothold_planning_interface.py"
    • Stores footplanning parameters from "Debug" topic
Foothold Planning:
  • "jet_leg/foothold_planning.py"
    • Selects foothold option with maximum feasible region (selectMaximumFeasibleArea()). Uses foothold_planning parameters (including possible foothold options positions) and also IP parameters.
IP Algorithm:
  • "jet_leg/computational_dynamics.py"
    • Sets up the IP algorithm (setup_iterative_projection()) and starts the IP aglorithm (iterative_projection_bretl())
Tools:
  • "jet_leg/constraints.py"
    • Sets up the inequality constraints (getInequalities())and computes the "actuation polygon" (computeActuationPolygons())
  • "jet_leg/computational_geometry.py"
    • Computes halfspace representation for the constraints and also computes polygon area for foothold planning
  • "jet_leg/hyq_kinematics.py"
    • Compute forward and inverse kinematics
  • "jet_leg/geometry.py"
    • Sorts vertices of a polygon
  • "jet_leg/math_tools.py"
    • Performs algebric and geometric computations