Exemplo n.º 1
0
def lp_ineq_4D(K, k, ones_range=None):
    if (ones_range == None):
        ones_range = (0, K.shape[0])
    K, k = __normalize(K, k)
    K_1 = __compute_K_1(K, ones_range)
    cost = array([0., 0., 0., -1.])
    lb = array([-100000000. for _ in range(4)])
    ub = array([100000000. for _ in range(4)])
    Alb = array([-100000000. for _ in range(k.shape[0])])
    #~ K_1, k =  __normalize(K_1, k)
    solver = solver_LP_abstract.getNewSolver('qpoases',
                                             "dyn_eq",
                                             maxIter=10000,
                                             maxTime=10000.0,
                                             useWarmStart=False,
                                             verb=0)
    (status, res, rest) = solver.solve(cost,
                                       lb=lb,
                                       ub=ub,
                                       A_in=K_1,
                                       Alb=Alb,
                                       Aub=-k,
                                       A_eq=None,
                                       b=None)

    #problem solved or unfeasible
    status_ok = status == solver_LP_abstract.LP_status.OPTIMAL
    p_solved = status_ok and res[3] >= 0.
    return status_ok, p_solved, res
 def __init__ (self, name, contact_points, contact_normals, mu, g, mass, a, a0, maxIter=100, verb=0, solver='cvxopt'):
     self.name       = name;
     self.solver     = optim.getNewSolver(solver, name, maxIter=maxIter, verb=verb);
     self.maxIter    = maxIter;
     self.verb       = verb;
     self.m_in       = 6;
     self.a          = np.asarray(a).squeeze().copy();
     self.a0         = np.asarray(a0).squeeze().copy();
     
     # compute matrix A, which maps the force generator coefficients into the centroidal wrench
     (self.A, G4) = compute_centroidal_cone_generators(contact_points, contact_normals, mu);
     self.D = np.zeros((6,3));
     self.d = np.zeros(6);
     self.D[3:,:] = -mass*crossMatrix(g);
     self.d[:3]   = mass*g;
     self.n          = self.A.shape[1]+1;
     
     self.grad = np.zeros(self.n);
     self.grad[0] = -1.0;
     self.constrMat = np.zeros((self.m_in,self.n));
     self.constrMat[:,1:] = self.A;
     self.constrMat[:,0] = -np.dot(self.D, self.a);
     self.constrB = self.d + np.dot(self.D, self.a0);
     self.lb = np.zeros(self.n);
     self.lb[0] = -1e10;
     self.ub = np.array(self.n*[1e10,]);
     self.x = np.zeros(self.n);
    def __init__(self,
                 name,
                 contact_points,
                 contact_normals,
                 mu,
                 g,
                 mass,
                 contact_tangents=None,
                 maxIter=100,
                 verb=0,
                 solver='cvxopt'):
        g = np.asarray(g).squeeze()
        self.solver = optim.getNewSolver(solver,
                                         name,
                                         maxIter=maxIter,
                                         verb=verb)
        self.D = np.zeros((6, 3))
        self.d = np.zeros(6)
        self.D[3:, :] = -mass * crossMatrix(g)
        self.d[:3] = mass * g
        (G, tmp) = compute_centroidal_cone_generators(contact_points,
                                                      contact_normals, mu,
                                                      contact_tangents)

        self.name = name
        self.maxIter = maxIter
        self.verb = verb
        self.m_in = G.shape[1]
        self.n = 6
        #        self.iter       = 0;
        #        self.initialized = False;

        #        self.Hess = np.zeros((self.n,self.n));
        self.grad = np.zeros(self.n)
        self.constrInMat = np.zeros((self.m_in, self.n))
        self.constrInMat[:, :] = G.T
        self.constrInUB = np.zeros(self.m_in) + 1e100
        self.constrInLB = np.zeros(self.m_in)
        self.constrEqMat = np.dot(np.ones(G.shape[1]), G.T).reshape((1, 6))
        self.constrEqB = np.ones(1)
        self.lb = np.array(self.n * [
            -1e10,
        ])
        self.ub = np.array(self.n * [
            1e10,
        ])
        self.x = np.zeros(self.n)
Exemplo n.º 4
0
 def __init__(self,
              name,
              c0,
              v,
              contact_points,
              contact_normals,
              mu,
              g,
              mass,
              maxIter=10000,
              verb=0,
              regularization=1e-5,
              solver='qpoases'):
     ''' Constructor
         @param c0 Initial CoM position
         @param v Opposite of the direction in which you want to maximize the CoM acceleration (typically that would be
                                                                                                the CoM velocity direction)
         @param g Gravity vector
         @param regularization Weight of the force minimization, the higher this value, the sparser the solution
         @param solver Specify which LP solver to use (qpoases, cvxopt or scipy)
     '''
     self.name = name
     self.verb = verb
     self.m_in = 6
     self.solver = optim.getNewSolver(solver,
                                      name,
                                      maxIter=maxIter,
                                      verb=verb)
     self.hessian_regularization = INITIAL_HESSIAN_REGULARIZATION
     self.b = np.zeros(6)
     self.d = np.empty(6)
     self.c0 = np.empty(3)
     self.v = np.empty(3)
     self.constrUB = np.zeros(self.m_in) + 1e100
     #        self.constrLB = np.zeros(self.m_in)-1e100;
     self.set_problem_data(c0, v, contact_points, contact_normals, mu, g,
                           mass, regularization)
Exemplo n.º 5
0
    def __init__(self,
                 name,
                 c0,
                 dc0,
                 contact_points,
                 contact_normals,
                 mu,
                 g,
                 mass,
                 kinematic_constraints=None,
                 angular_momentum_constraints=None,
                 contactTangents=None,
                 maxIter=1000,
                 verb=0,
                 regularization=1e-5,
                 solver='qpoases'):
        ''' Constructor
            @param c0 Initial CoM position
            @param dc0 Initial CoM velocity
            @param contact points A matrix containing the contact points
            @param contact normals A matrix containing the contact normals
            @param mu Friction coefficient (either a scalar or an array)
            @param g Gravity vector
            @param mass The robot mass            
            @param kinematic constraints couple [A,b] such that the com is constrained by  A x <= b
            @param regularization Weight of the force minimization, the higher this value, the sparser the solution
        '''
        assert mass > 0.0, "Mass is not positive"
        assert mu > 0.0, "Friction coefficient is not positive"
        assert np.asarray(
            c0).squeeze().shape[0] == 3, "Com position vector has not size 3"
        assert np.asarray(
            dc0).squeeze().shape[0] == 3, "Com velocity vector has not size 3"
        assert np.asarray(
            contact_points).shape[1] == 3, "Contact points have not size 3"
        assert np.asarray(
            contact_normals).shape[1] == 3, "Contact normals have not size 3"
        assert np.asarray(contact_points).shape[0] == np.asarray(
            contact_normals
        ).shape[
            0], "Number of contact points do not match number of contact normals"
        self._name = name
        self._maxIter = maxIter
        self._verb = verb
        self._c0 = np.asarray(c0).squeeze().copy()
        self._dc0 = np.asarray(dc0).squeeze().copy()
        self._mass = mass
        self._g = np.asarray(g).squeeze().copy()
        self._gX = skew(self._g)
        #        self._regularization    = regularization;
        self.set_contacts(contact_points, contact_normals, mu, contactTangents)
        if kinematic_constraints != None:
            self._kinematic_constraints = kinematic_constraints[:]
        else:
            self._kinematic_constraints = None
        if angular_momentum_constraints != None:
            self._angular_momentum_constraints = angular_momentum_constraints[:]
        else:
            self._angular_momentum_constraints = None

        self._lp_solver = getNewSolver('qpoases',
                                       "name",
                                       useWarmStart=False,
                                       verb=0)
        self._qp_solver = qp_solver(3,
                                    0,
                                    solver='qpoases',
                                    accuracy=1e-6,
                                    maxIter=100,
                                    verb=0)