cl = robot.client # q = [x, y, theta] # (z not considered since planar) q1 = [-4, 4, 1, 0] q2 = [4, -4, 1, 0] # obstS 1 #q1 = [-4, 4, 0]; q2 = [4, -4, 0] # obstS 1 ps.setInitialConfig(q1) ps.addGoalConfig(q2) cl.obstacle.loadObstacleModel('potential_description', 'obstacles_concaves', 'obstacles_concaves') #ps.createOrientationConstraint ("orConstraint", "base_joint_rz", "", [1,0,0,0], [0,0,1]) #ps.setNumericalConstraints ("constraints", ["orConstraint"]) ps.selectPathPlanner("VisibilityPrmPlanner") #ps.selectPathValidation ("Dichotomy", 0.) ps.solve() ps.pathLength(0) ps.addPathOptimizer("GradientBased") ps.optimizePath(0) ps.numberPaths() ps.pathLength(ps.numberPaths() - 1) import matplotlib.pyplot as plt from mutable_trajectory_plot import planarPlot, addNodePlot from parseLog import parseCollConstrPoints num_log = 31891 contactPoints = parseCollConstrPoints(num_log, '77: contact point = (')
vf.loadObstacleModel("iai_maps", "room", "room") # with displayArrow parameter the viewer will display velocity and acceleration of the # center of the robot with 3D arrow. The length scale with the amplitude with a length # of 1 for the maximal amplitude. v = vf.createViewer(displayArrows=True) ps.setInitialConfig(q_init) ps.addGoalConfig(q_goal) ps.addPathOptimizer("RandomShortcut") # select kinodynamic methods : ps.selectSteeringMethod("Kinodynamic") ps.selectDistance("Kinodynamic") # the Kinodynamic steering method require a planner that build directionnal roadmap # (with oriented edges) as the trajectories cannot always be reverse. ps.selectPathPlanner("BiRRTPlanner") print(ps.solve()) # display the computed roadmap. Note that the edges are all represented as straight line # and may not show the real motion of the robot between the nodes : v.displayRoadmap("rm") # Alternatively, use the following line instead of ps.solve() to display the roadmap as # it's computed (note that it slow down a lot the computation) # v.solveAndDisplay('rm',1) # Highlight the solution path used in the roadmap v.displayPathMap("pm", 0) # remove the roadmap for the scene :
from hpp.gepetto import Viewer, PathPlayer r = Viewer (ps) pp = PathPlayer (robot.client, r) #r.loadObstacleModel ("puzzle_description","decor_very_easy","decor_very_easy") r.loadObstacleModel ("puzzle_description","decor_easy","decor_easy") r(q2) # r(q1) #q1bis = q2; q2bis = [0.0, 0.0, -0.8, 1.0, 0.0, 0.0, 0.0] #ps.resetGoalConfigs (); ps.setInitialConfig (q1bis); ps.addGoalConfig (q2bis); ps.solve () # ps.resetGoalConfigs (); ps.setInitialConfig (q1); ps.addGoalConfig (q2bis); ps.solve () #i = ps.numberPaths()-1 ps.setInitialConfig (q1); ps.addGoalConfig (q2) ps.selectPathPlanner ("VisibilityPrmPlanner") #ps.selectPathValidation ("Dichotomy", 0.) #ps.saveRoadmap ('/local/mcampana/devel/hpp/data/puzzle_veryEasy_PRM.rdm') #ps.readRoadmap ('/local/mcampana/devel/hpp/data/puzzle_easy_RRT.rdm') #ps.readRoadmap ('/local/mcampana/devel/hpp/data/puzzle_easy_PRM1.rdm') # srand # problem ? #ps.readRoadmap ('/local/mcampana/devel/hpp/data/puzzle_easy_PRM1.rdm') # srand 1453110445(909sec) [COLL!] #ps.readRoadmap ('/local/mcampana/devel/hpp/data/puzzle_easy_PRM2.rdm') # srand # just after solve, GB OK. But after readroadmap+solve, segfault quaternions.... ps.readRoadmap ('/local/mcampana/devel/hpp/data/puzzle_easy_PRM_test1.rdm') #srand 1454520599 working0.05 # srand 1454521537 (no RM saved) works 7 -> 5, best 0.2 ps.solve () ps.pathLength(0) len(ps.getWaypoints (0)) r(q1)
#q_goal[0:3]=[-4.5,-4.8,0.4]# harder goal position #set goal velocity (along x,y,z axis) : q_goal[-6:-3]=[0,0,0] vf.loadObstacleModel ("iai_maps", "room", "room") # with displayArrow parameter the viewer will display velocity and acceleration of the center of the robot with 3D arrow. The length scale with the amplitude with a length of 1 for the maximal amplitude v = vf.createViewer(displayArrows = True) ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.addPathOptimizer ("RandomShortcut") #select kinodynamic methods : ps.selectSteeringMethod("Kinodynamic") ps.selectDistance("Kinodynamic") # the Kinodynamic steering method require a planner that build directionnal roadmap (with oriented edges) as the trajectories cannot always be reversed. ps.selectPathPlanner("BiRRTPlanner") print (ps.solve ()) # display the computed roadmap. Note that the edges are all represented as straight line and may not show the real motion of the robot between the nodes : v.displayRoadmap("rm") #Alternatively, use the following line instead of ps.solve() to display the roadmap as it's computed (note that it slow down a lot the computation) #v.solveAndDisplay('rm',1) # Highlight the solution path used in the roadmap v.displayPathMap('pm',0) # remove the roadmap for the scene : #v.client.gui.removeFromGroup("rm",v.sceneName)
ps = ProblemSolver (robot) v = Viewer (ps) q_init = robot.getCurrentConfig () q_init[0:3] = [0.5, 1, 1] #z=0.4 pour sphere v (q_init) q_goal = q_init [::] q_goal [0:3] = [5,1, 1] #v (q_goal) print("chargement map") v.loadObstacleModel ("iai_maps", "tunnel", "tunnel") ps.selectPathPlanner("rrtPerso") print("Debut motion planning") ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) v.solveAndDisplay(white,0.02,1,yellow) ps.solve () v.displayRoadmap(white,0.02,1,yellow) v.client.gui.addXYZaxis("test0",white,0.05,1)
class AbstractPathPlanner: rbprmBuilder = None ps = None v = None afftool = None pp = None extra_dof_bounds = None robot_node_name = None # name of the robot in the node list of the viewer def __init__(self): self.v_max = -1 # bounds on the linear velocity for the root, negative values mean unused self.a_max = -1 # bounds on the linear acceleration for the root, negative values mean unused self.root_translation_bounds = [ 0 ] * 6 # bounds on the root translation position (-x, +x, -y, +y, -z, +z) self.root_rotation_bounds = [ -3.14, 3.14, -0.01, 0.01, -0.01, 0.01 ] # bounds on the rotation of the root (-z, z, -y, y, -x, x) # The rotation bounds are only used during the random sampling, they are not enforced along the path self.extra_dof = 6 # number of extra config appended after the joints configuration, 6 to store linear root velocity and acceleration self.mu = 0.5 # friction coefficient between the robot and the environment self.used_limbs = [ ] # names of the limbs that must be in contact during all the motion self.size_foot_x = 0 # size of the feet along the x axis self.size_foot_y = 0 # size of the feet along the y axis self.q_init = [] self.q_goal = [] @abstractmethod def load_rbprm(self): """ Build an rbprmBuilder instance for the correct robot and initialize it's extra config size """ pass def set_configurations(self): self.rbprmBuilder.client.robot.setDimensionExtraConfigSpace( self.extra_dof) self.q_init = self.rbprmBuilder.getCurrentConfig() self.q_goal = self.rbprmBuilder.getCurrentConfig() self.q_init[2] = self.rbprmBuilder.ref_height self.q_goal[2] = self.rbprmBuilder.ref_height def compute_extra_config_bounds(self): """ Compute extra dof bounds from the current values of v_max and a_max By default, set symmetrical bounds on x and y axis and bounds z axis values to 0 """ # bounds for the extradof : by default use v_max/a_max on x and y axis and 0 on z axis self.extra_dof_bounds = [ -self.v_max, self.v_max, -self.v_max, self.v_max, 0, 0, -self.a_max, self.a_max, -self.a_max, self.a_max, 0, 0 ] def set_joints_bounds(self): """ Set the root translation and rotation bounds as well as the the extra dofs bounds """ self.rbprmBuilder.setJointBounds("root_joint", self.root_translation_bounds) self.rbprmBuilder.boundSO3(self.root_rotation_bounds) self.rbprmBuilder.client.robot.setExtraConfigSpaceBounds( self.extra_dof_bounds) def set_rom_filters(self): """ Define which ROM must be in collision at all time and with which kind of affordances By default it set all the roms in used_limbs to be in contact with 'support' affordances """ self.rbprmBuilder.setFilter(self.used_limbs) for limb in self.used_limbs: self.rbprmBuilder.setAffordanceFilter(limb, ['Support']) def init_problem(self): """ Load the robot, set the bounds and the ROM filters and then Create a ProblemSolver instance and set the default parameters. The values of v_max, a_max, mu, size_foot_x and size_foot_y must be defined before calling this method """ self.load_rbprm() self.set_configurations() self.compute_extra_config_bounds() self.set_joints_bounds() self.set_rom_filters() self.ps = ProblemSolver(self.rbprmBuilder) # define parameters used by various methods : if self.v_max >= 0: self.ps.setParameter("Kinodynamic/velocityBound", self.v_max) if self.a_max >= 0: self.ps.setParameter("Kinodynamic/accelerationBound", self.a_max) if self.size_foot_x > 0: self.ps.setParameter("DynamicPlanner/sizeFootX", self.size_foot_x) if self.size_foot_y > 0: self.ps.setParameter("DynamicPlanner/sizeFootY", self.size_foot_y) self.ps.setParameter("DynamicPlanner/friction", 0.5) # sample only configuration with null velocity and acceleration : self.ps.setParameter("ConfigurationShooter/sampleExtraDOF", False) def init_viewer(self, env_name, env_package="hpp_environments", reduce_sizes=[0, 0, 0], visualize_affordances=[]): """ Build an instance of hpp-gepetto-viewer from the current problemSolver :param env_name: name of the urdf describing the environment :param env_package: name of the package containing this urdf (default to hpp_environments) :param reduce_sizes: Distance used to reduce the affordances plan toward the center of the plane (in order to avoid putting contacts closes to the edges of the surface) :param visualize_affordances: list of affordances type to visualize, default to none """ vf = ViewerFactory(self.ps) self.afftool = AffordanceTool() self.afftool.setAffordanceConfig('Support', [0.5, 0.03, 0.00005]) self.afftool.loadObstacleModel("package://" + env_package + "/urdf/" + env_name + ".urdf", "planning", vf, reduceSizes=reduce_sizes) self.v = vf.createViewer(ghost=True, displayArrows=True) self.pp = PathPlayer(self.v) for aff_type in visualize_affordances: self.afftool.visualiseAffordances(aff_type, self.v, self.v.color.lightBrown) def init_planner(self, kinodynamic=True, optimize=True): """ Select the rbprm methods, and the kinodynamic ones if required :param kinodynamic: if True, also select the kinodynamic methods :param optimize: if True, add randomShortcut path optimizer (or randomShortcutDynamic if kinodynamic is also True) """ self.ps.selectConfigurationShooter("RbprmShooter") self.ps.selectPathValidation("RbprmPathValidation", 0.05) if kinodynamic: self.ps.selectSteeringMethod("RBPRMKinodynamic") self.ps.selectDistance("Kinodynamic") self.ps.selectPathPlanner("DynamicPlanner") if optimize: if kinodynamic: self.ps.addPathOptimizer("RandomShortcutDynamic") else: self.ps.addPathOptimizer("RandomShortcut") def solve(self): """ Solve the path planning problem. q_init and q_goal must have been defined before calling this method """ if len(self.q_init) != self.rbprmBuilder.getConfigSize(): raise ValueError( "Initial configuration vector do not have the right size") if len(self.q_goal) != self.rbprmBuilder.getConfigSize(): raise ValueError( "Goal configuration vector do not have the right size") self.ps.setInitialConfig(self.q_init) self.ps.addGoalConfig(self.q_goal) self.v(self.q_init) t = self.ps.solve() print("Guide planning time : ", t) def display_path(self, path_id=-1, dt=0.1): """ Display the path in the viewer, if no path specified display the last one :param path_id: the Id of the path specified, default to the most recent one :param dt: discretization step used to display the path (default to 0.1) """ if self.pp is not None: if path_id < 0: path_id = self.ps.numberPaths() - 1 self.pp.dt = dt self.pp.displayVelocityPath(path_id) def play_path(self, path_id=-1, dt=0.01): """ play the path in the viewer, if no path specified display the last one :param path_id: the Id of the path specified, default to the most recent one :param dt: discretization step used to display the path (default to 0.01) """ self.show_rom() if self.pp is not None: if path_id < 0: path_id = self.ps.numberPaths() - 1 self.pp.dt = dt self.pp(path_id) def hide_rom(self): """ Remove the current robot from the display """ self.v.client.gui.setVisibility(self.robot_node_name, "OFF") def show_rom(self): """ Add the current robot to the display """ self.v.client.gui.setVisibility(self.robot_node_name, "ON") @abstractmethod def run(self): """ Must be defined in the child class to run all the methods with the correct arguments. """ # example of definition: """ self.init_problem() # define initial and goal position self.q_init[:2] = [0, 0] self.q_goal[:2] = [1, 0] self.init_viewer("multicontact/ground", visualize_affordances=["Support"]) self.init_planner() self.solve() self.display_path() self.play_path() """ pass
q_init = robot.getCurrentConfig () q_goal = q_init [::] q_init [0:2] = [-3.2, -4] rank = robot.rankInConfiguration ['torso_lift_joint'] q_init [rank] = 0.2 r (q_init) q_goal [0:2] = [-3.2, -4] rank = robot.rankInConfiguration ['l_shoulder_lift_joint'] q_goal [rank] = 0.5 rank = robot.rankInConfiguration ['l_elbow_flex_joint'] q_goal [rank] = -0.5 rank = robot.rankInConfiguration ['r_shoulder_lift_joint'] q_goal [rank] = 0.5 rank = robot.rankInConfiguration ['r_elbow_flex_joint'] q_goal [rank] = -0.5 r (q_goal) ps.loadObstacleFromUrdf ("iai_maps", "kitchen_area", "") ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.selectPathPlanner ("PRM") ps.solve () from hpp_ros import PathPlayer pp = PathPlayer (robot.client, r) pp (0) pp (1)
from hpp.corbaserver import ProblemSolver from hpp.gepetto import Viewer from hpp.gepetto import PathPlayer white = [1.0, 1.0, 1.0, 1.0] green = [0.23, 0.75, 0.2, 0.5] yellow = [0.85, 0.75, 0.15, 1] pink = [1, 0.6, 1, 1] orange = [1, 0.42, 0, 1] robot = Robot('box') robot.setJointBounds("base_joint_xyz", [0, 5, 0, 2, 0, 2]) robot.client.robot.setDimensionExtraConfigSpace( robot.getNumberDof()) # extraDof for velocitiy ps = ProblemSolver(robot) ps.selectPathPlanner("dyn") v = Viewer(ps) #v.loadObstacleModel ("iai_maps", "tunnel", "tunnel") v.loadObstacleModel("iai_maps", "abstract", "abstract") q_init = [0, 1, 1, 1, 0, 0, 0] q_goal = [5, 1, 1, 0.9239, 0, -0.3827, 0] v(q_init) ps.setInitialConfig(q_init) ps.addGoalConfig(q_goal) ps.addPathOptimizer("RandomShortcut") ps.solve()
solver.addGoalConfig (qG) #publisher(q) #print q #sys.exit(0) qqinit = solver.getInitialConfig()[0:3] qqgoal = solver.getGoalConfigs()[0][0:3] print "planning from",qqinit,"to",qqgoal #print "Irreducible Planner" #solver.selectPathPlanner("PRM") #solver.selectPathPlanner("DiffusingPlanner") print "solve" start_time = float(time.time()) solver.selectPathPlanner("IrreduciblePlanner") solver.solve () end_time = float(time.time()) seconds = end_time - start_time hours = seconds/3600 print("--- %s seconds ---" % seconds) print("--- %s hours ---" % hours) fname0 = "../data-traj/rrt-wall0.tau" fname1 = "../data-traj/rrt-wall1.tau" pathplayer = PathPlayer (robot.client, publisher) pathplayer(0) pathplayer(1) pathplayer.toFile(0,fname0)
return v = FakeViewer() #afftool.visualiseAffordances('Support', v, v.color.lightBrown) q_init = rbprmBuilder.getCurrentConfig (); q_init[:7] = cfg.INIT_CONFIG_ROOT # Choosing RBPRM shooter and path validation methods. psGuide.selectConfigurationShooter("RbprmShooter") psGuide.selectPathValidation("RbprmPathValidation",0.05) # Choosing kinodynamic methods : psGuide.selectSteeringMethod("RBPRMKinodynamic") psGuide.selectDistance("Kinodynamic") psGuide.selectPathPlanner("DynamicPlanner") # Solve the planning problem : # move the robot out of the view before computing the contacts from hpp.corbaserver import Client #~ #DEMO code to play root path and final contact plan cl = Client() cl.problem.selectProblem("rbprm_path") rbprmBuilder2 = Robot ("toto") ps2 = ProblemSolver( rbprmBuilder2 ) cl.problem.selectProblem("default") #~ cl.problem.movePathToProblem(pId,"rbprm_path",rbprmBuilder.getAllJointNames()[1:]) r2 = Viewer (ps2, viewerClient=v.client) q_far = q_init[::] q_far[2] = -2
from hpp.environments import Buggy robot = Buggy("buggy") robot.setJointBounds ("base_joint_xy", [-5, 16, -4.5, 4.5]) from hpp.corbaserver import ProblemSolver ps = ProblemSolver (robot) from hpp.gepetto import ViewerFactory gui = ViewerFactory (ps) gui.loadObstacleModel ('hpp_environments', "scene", "scene") q_init = robot.getCurrentConfig () q_goal = q_init [::] q_init[0:2] = [-3.7, -4]; gui (q_init) q_goal [0:2] = [15,2] gui (q_goal) ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.selectSteeringMethod ("ReedsShepp") ps.selectPathPlanner ("DiffusingPlanner") ps.addPathOptimizer ("RandomShortcut") t = ps.solve () print ("solving time", t)
from hpp.corbaserver.lydia import Robot from hpp.corbaserver import ProblemSolver from hpp.gepetto import Viewer, PathPlayer robot = Robot ('lydia') robot.setJointBounds('base_joint_xyz', [-0.9, 0.9, -0.9, 0.9, -1.1, 1.1]) ps = ProblemSolver (robot) r = Viewer (ps) r.loadObstacleModel ("hpp_benchmark", "obstacle", "obstacle") q_init = robot.getCurrentConfig () q_goal = q_init [::] q_init [2] = -0.6 q_goal [2] = 0.6 ps.selectPathPlanner ("VisibilityPRM") ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.solve ()
from hpp.corbaserver import ProblemSolver ps = ProblemSolver (robot) from hpp.gepetto import Viewer gui = Viewer (ps) gui.loadObstacleModel ('tp-rrt', "scene", "scene") q_init = robot.getCurrentConfig () q_goal = q_init [::] q_init [0:2] = q_init[0:2]=[-3.7, -4]; gui(q_init) gui (q_init) q_goal [0:2] = [15,2] gui (q_goal) #~ ps.loadObstacleFromUrdf ("iai_maps", "kitchen_area", "") ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.selectPathPlanner ("PlannerTP") ps.addPathOptimizer ("RandomShortcut") t = ps.solve () print ("solving time", t) from hpp.gepetto import PathPlayer pp = PathPlayer (robot.client, gui)
q1 = [0, 0, 0.9, 0, 0, 0]; q2 = [0, 0, -0.9, 0, 0, 0] from hpp.gepetto import Viewer, PathPlayer r = Viewer (ps) pp = PathPlayer (robot.client, r) r.loadObstacleModel ("puzzle_description","decor_very_easy","decor_very_easy") r(q1) ps.setInitialConfig (q1); ps.addGoalConfig (q2) # Load box obstacle in HPP for collision avoidance cl.obstacle.loadObstacleModel('puzzle_description','decor_very_easy','') #cl.obstacle.loadObstacleModel('puzzle_description','decor_easy','') #cl.obstacle.loadObstacleModel('puzzle_description','decor','') ps.selectPathPlanner ("VisibilityPrmPlanner") # 26min solve time begin=time.time() ps.solve () end=time.time() print "Solving time: "+str(end-begin) ps.pathLength(0) ps.addPathOptimizer("GradientBased") begin=time.time() ps.optimizePath (0) end=time.time() print "Optim time: "+str(end-begin) cl.problem.getIterationNumber () ps.pathLength(1) #vPRM: 24 iter, 25.3s->21s, weighted Cost + comp linear error #RRTc: 30 iter, 33.4s->14.9s, weighted Cost + comp linear error
q1=[0,1.5,0.64870180180254433111, 0.9249114088877176,0,0,-0.38018270043406405,0,0,0,0,0.26179900000000000393,0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0.26179900000000000393,-0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0,0,-0.45378600000000002268,0.87266500000000002402,-0.41887900000000000134,0,0,0,-0.45378600000000002268,0.87266500000000002402,-0.41887900000000000134,0] #q1=[0,1.5,0.64870180180254433111, 1,0,0,0 ,0,0,0,0,0.26179900000000000393,0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0.26179900000000000393,-0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927,0,0,-0.45378600000000002268,0.87266500000000002402,-0.41887900000000000134,0,0,0,-0.45378600000000002268,0.87266500000000002402,-0.41887900000000000134,0] #original from antonio q2=[0,-2.5,0.64870180180254433111, 0.7101853756232854, 0, 0, -0.7040147244559684, 0,0,0,0,0.26179900000000000393, 0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927, -0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927, 0.1745319999999999927,-0.1745319999999999927,0.26179900000000000393, -0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927, -0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927, 0.1745319999999999927,-0.1745319999999999927,0,0,-0.45378600000000002268, 0.87266500000000002402,-0.41887900000000000134,0,0,0,-0.45378600000000002268, 0.87266500000000002402,-0.41887900000000000134,0] #q2=[0,-2.5,0.64870180180254433111, 1,0,0,0, 0,0,0,0,0.26179900000000000393, 0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927, -0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927, 0.1745319999999999927,-0.1745319999999999927,0.26179900000000000393, -0.1745299999999999907,0,-0.52359900000000003661,0,0,0.1745319999999999927, -0.1745319999999999927,0.1745319999999999927,-0.1745319999999999927, 0.1745319999999999927,-0.1745319999999999927,0,0,-0.45378600000000002268, 0.87266500000000002402,-0.41887900000000000134,0,0,0,-0.45378600000000002268, 0.87266500000000002402,-0.41887900000000000134,0] publisher = ScenePublisher(robot) publisher(q2) time.sleep(1) publisher(q1) ## Add constraints solver = ProblemSolver (robot) print "diffusing planner (general RRT implementation)" solver.selectPathPlanner("DiffusingPlanner") solver.loadObstacleFromUrdf("hpp_ros","wall") solver.setInitialConfig (q1) solver.addGoalConfig (q2) #pc.invokeIrreduciblePlanner() #mpc = MPClient() #pc = mpc.precomputation #cnames = pc.getNumericalConstraints (robot.getInitialConfig()) #solver.setNumericalConstraints ("stability-constraints", cnames) ## add obstacles print "solve" start_time = float(time.time()) solver.solve () end_time = float(time.time())
class Agent(Client): robot = None platform = None index = 0 ps = None # we load other agents as ghosts to reduce the computation time while planning ghosts = [] ghost_urdf = '' ghost_package = '' # to avoid confusion, we use start and end instead of init and goal start_config = [] end_config = [] current_config = [] # this is not used for now permitted_plan = [] # this is the plan generated repeat = 0 # to help the selectProblem function # an agent should have a robot and the start and end configuration # to avoid confusion, we use start_config instead of init_config and # end_confi instead of goal_config def __init__(self, robot, start, end): Client.__init__(self) self.repeat = 0 # print 'creating an agent of type ', robotType self.robot = robot self.start_config = start self.end_config = end self.current_config = self.start_config self.__plan_proposed = [] # once all agents are generated, we may register the agents to a platform def registerPlatform(self, platform, index): self.platform = platform self.index = index # this function gives some information about the agent and robot it is managing def printInformation(self): print '-------------------------------------------' print 'name of the robot:\t', self.robot.name print 'configuration size:\t', self.robot.getConfigSize() print 'degree of freedom:\t', self.robot.getNumberDof() print 'mass of the robot:\t', self.robot.getMass() print 'the center of mass:\t', self.robot.getCenterOfMass() config = self.robot.getCurrentConfig() nm = self.robot.getJointNames() print 'there are ', len(nm), 'joint names in total. They are:' for i in range(len(nm)): lower = self.robot.getJointBounds(nm[i])[0] upper = self.robot.getJointBounds(nm[i])[1] print 'joint name: ', nm[ i], '\trank in configuration:', self.robot.rankInConfiguration[ nm[i]], print '\tlower bound: {0:.3f}'.format( lower), '\tupper bound: {0:.3f}'.format(upper) # set up the environment def setEnvironment(self): if self.platform.env != None: self.ps.loadObstacleFromUrdf(self.platform.env.packageName, self.platform.env.urdfName, self.platform.env.name) # self.ps.moveObstacle('airbase_link_0', [0,0, -3, 1,0,0,0]) # load the other agents to the problem solver def loadOtherAgents(self): # print 'There are ', len(self.platform.agents), 'agents' #load ghost agents for a in self.platform.agents: if (a.index != self.index): # if it is not itself then load a ghost agent g = Ghost() self.ps.loadObstacleFromUrdf( g.packageName, g.urdfName, a.robot.name) # it's the robot's name!!! # and then place it at the initial location of the agent # print self.robot.name, ' is now loading ', a.robot.name, ' as a ghost' config = a.current_config spec = self.getMoveSpecification(config) spec[2] = 0.3 self.obstacle.moveObstacle(a.robot.name + 'base_link_0', spec) # load agents from the node def loadOtherAgentsFromNode(self, node): print 'There are ', len(self.platform.agents), 'agents' #load ghost agents for a in self.platform.agents: if (a.index != self.index): # if it is not itself then load a ghost agent g = Ghost() self.ps.loadObstacleFromUrdf( g.packageName, g.urdfName, a.robot.name) # it's the robot's name!!! # and then place it at the initial location of the agent config = node.getAgentCurrentConfig(a.index) spec = self.getMoveSpecification(config) self.obstacle.moveObstacle(a.robot.name + 'base_link_0', spec) print self.robot.name, ' is now loading ', a.robot.name, ' as a ghost', 'it is at ', spec[ 0], spec[1] # note that the default solver does not consider the position of other agents def startDefaultSolver(self): self.repeat += 1 name = self.robot.name self.problem.selectProblem(str(self.index) + ' ' + str(self.repeat)) self.robot = HyQ(name) self.ps = ProblemSolver(self.robot) self.ps.setInitialConfig(self.start_config) self.ps.addGoalConfig(self.end_config) self.ps.selectPathPlanner("VisibilityPrmPlanner") self.ps.addPathOptimizer("RandomShortcut") # initialise a solver from a node, the node contains information about other agents and itself # this method is used when proposing plans while interacting with platform for MAS path planning def startNodeSolver(self, node): self.repeat += 1 name = self.robot.name self.problem.selectProblem(str(self.index) + ' ' + str(self.repeat)) self.robot = HyQ(name) self.ps = ProblemSolver(self.robot) cfg = node.getAgentCurrentConfig(self.index) print 'this iteration, the agent', name, 'starts from ', cfg[0], cfg[1] self.ps.setInitialConfig(cfg) self.ps.addGoalConfig(self.end_config) self.ps.selectPathPlanner("VisibilityPrmPlanner") self.ps.addPathOptimizer("RandomShortcut") # this is only used when the agent takes too long (30 seconds) while planning def terminate_solving(self): self.problem.interruptPathPlanning() # the solve method for problem solver but with a time bound def solve(self): # try catch ------------------- try: t = Timer(30.0, self.terminate_solving) t.start() print 'solved: ', self.ps.solve() t.cancel() except Error as e: print e.msg print '***************\nfailed to plan within limited time\n**************' return -1 # self.repeat += 1 # store the path for two reasons: # 1. store as a default plan # 2. to continue the path def storePath(self, choice=0, segments=8): # always store the first one for now self.__plan_proposed = [] for p in range(int(round(segments * self.ps.pathLength(choice)))): self.__plan_proposed.append( self.ps.configAtParam(choice, p * 1.0 / segments)) # the last configuration is the goal configuration if self.ps.configAtParam( choice, self.ps.pathLength(choice)) == self.end_config: self.__plan_proposed.append(self.end_config) print 'stored; plan length: ', len(self.__plan_proposed) # this is hard colded for now for the airplane example, we should introduce an entry for the environment # class about it. def setBounds(self): self.robot.setJointBounds("base_joint_xy", [-35, 10, -2.6, 4.3]) #the rest are just some helping functions def getConfigOfProposedPlanAtTime(self, index): return self.__plan_proposed[index] def getConfigOfPermittedPlanAtTime(self, index): return self.permitted_plan[index] def getProposedPlanLength(self): return len(self.__plan_proposed) def setPermittedPlan(self, plan): self.permitted_plan = plan def getPermittedPlanLength(self): return len(self.permitted_plan) # export the permitted plan to a specific file in the format # agent X # config 1 # config 2 # etc def exportPermittedPlan(self, filename): f = open(filename, 'a+') f.write('agent ' + str(self.index) + '\n') for p in self.permitted_plan: f.write(str(p)[1:-1] + '\n') f.close() # for the sake of manipulation, we return a copy of it def obtainPermittedPlan(self): return copy.copy(self.permitted_plan) # we will get only a copy of it, not the original one # to remind the difference, we use 'obtain' instead of 'get' def obtainProposedPlan(self): return copy.copy( self.__plan_proposed ) #for some reason, sometimes the value would maybe changed??? # to transfer the specification from 2D to 3D def getMoveSpecification(self, config): x = config[0] y = config[1] th = atan2(config[3], config[2]) # print 'sin = ', self.init_config[3], ' cos = ', self.init_config[2], ' th = ', th return [x, y, 0, cos(th / 2), 0, 0, sin(th / 2)] # the function to compute a plan, exceptions are not handled in this simple demo def computePlan(self, node): self.startNodeSolver(node) self.setBounds() self.setEnvironment() self.loadOtherAgentsFromNode(node) if self.solve() != -1: self.storePath() else: self.__plan_proposed = self.__plan_proposed[node.progress_time::] [node.getAgentCurrentConfig(self.index)] print 'take the previous one and continue the searching' return -1
ps = ProblemSolver (robot) from hpp.gepetto import Viewer v = Viewer (ps) q_init = robot.getCurrentConfig () q_goal = q_init [::] q_init [0:3] = [1, 2, 1] q_goal [0:3] = [9.15, 7.5, 0.5] v.loadObstacleModel ("iai_maps", "env_simple", "simple") ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.clearRoadmap() ps.selectPathPlanner("interactive") white=[1.0,1.0,1.0,1.0] brown=[0.85,0.75,0.15,0.5] ps.addPathOptimizer ("RandomShortcut") v.solveAndDisplay("rm1",50,white,0.05,1,brown) #ps.solve() from hpp.gepetto import PathPlayer pp = PathPlayer (robot.client, v) v(q_goal) pp (0) pp (1)
ps.lockJoint ('base_joint_xy', [0,0]) ps.lockJoint ('base_joint_rz', [1,0]) res = ps.applyConstraints (q_init) if res [0]: q_init = res [1] else: raise RuntimeError ('Failed to apply constraints on init config.') res = ps.applyConstraints (q_goal) if res [0]: q_goal = res [1] else: raise RuntimeError ('Failed to apply constraints on goal config.') robot.setCurrentConfig (q_init) r = vf.createViewer (collisionURDF = True) r (q_init) ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.selectPathPlanner ("DiffusingPlanner") ps.selectPathValidation("Progressive", 0.01) #ps.selectPathValidation("Progressive", 0.05) #ps.readRoadmap ('/home/florent/devel/fiad/data/roadmap.rdm') # ps.solve () pp = PathPlayer (ps.client, r)
ps = ProblemSolver(robot) v = Viewer(ps) q_init = robot.getCurrentConfig() q_init[0:3] = [9, 7.5, 0.5] v(q_init) q_goal = q_init[::] q_goal[0:3] = [0, 1.5, 0.5] #v (q_goal) print("chargement map") v.loadObstacleModel("iai_maps", "labyrinth2", "labyrinth2") ps.setInitialConfig(q_init) ps.addGoalConfig(q_goal) ps.selectPathPlanner("rrtConnect") v.solveAndDisplay("rm1", 50, white, 0.02, 1, brown) ps.solve() v.displayRoadmap("rm1", white, 0.02, 1, brown) v.client.gui.removeFromGroup("rm1", v.sceneName) pp = PathPlayer(robot.client, v) #print("affichage solution") #pp (0) print("affichage solution optimise")
from hpp.corbaserver.box import Robot from hpp.corbaserver import ProblemSolver from hpp.gepetto import Viewer from hpp.gepetto import PathPlayer white=[1.0,1.0,1.0,1.0] green=[0.23,0.75,0.2,0.5] yellow=[0.85,0.75,0.15,1] pink=[1,0.6,1,1] orange=[1,0.42,0,1] robot = Robot ('box') robot.setJointBounds ("base_joint_xyz", [0,5,0,2,0,2]) robot.client.robot.setDimensionExtraConfigSpace(robot.getNumberDof()) # extraDof for velocitiy ps = ProblemSolver (robot) ps.selectPathPlanner("dyn") v = Viewer (ps) #v.loadObstacleModel ("iai_maps", "tunnel", "tunnel") v.loadObstacleModel ("iai_maps", "abstract", "abstract") q_init = [0,1,1,1,0,0,0] q_goal = [5,1,1,0.9239,0,-0.3827,0] v (q_init) ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.addPathOptimizer ("RandomShortcut")
ps.setInitialConfig(q_init) ps.addGoalConfig(q_goal) # write problem in files : f = open(statusFilename, "w") f.write("q_init= " + str(q_init) + "\n") f.write("q_goal= " + str(q_goal) + "\n") f.close() # Choosing RBPRM shooter and path validation methods. ps.selectConfigurationShooter("RbprmShooter") ps.selectPathValidation("RbprmPathValidation", 0.05) # Choosing kinodynamic methods : ps.selectSteeringMethod("RBPRMKinodynamic") ps.selectDistance("Kinodynamic") ps.selectPathPlanner("DynamicPlanner") # Solve the planning problem : success = ps.client.problem.prepareSolveStepByStep() if not success: print "planning failed." import sys sys.exit(1) ps.client.problem.finishSolveStepByStep() try: # display solution : from hpp.gepetto import PathPlayer pp = PathPlayer(v)
q_init = robot.getCurrentConfig () q_init[0:3] = [6, 7, 0.5] #z=0.4 pour sphere v (q_init) q_goal = q_init [::] q_goal [0:3] = [-2.5, -3.5, 0.5] #v (q_goal) print("chargement map") v.loadObstacleModel ("iai_maps", "room", "room") ps.setInitialConfig (q_init) ps.addGoalConfig (q_goal) ps.selectPathPlanner("test2") print("debut motion plannning") tps1 = time.clock() ps.solve () tps2 = time.clock() print(tps2-tps1) from hpp.gepetto import PathPlayer pp = PathPlayer (robot.client, v) #print("affichage solution") #pp (0) while 1 : print("Appuyer sur entree pour afficher la solution")