Пример #1
0
def init_vrep():
    vrep.simxFinish(-1)
    client_id = vrep.simxStart('127.0.0.1',19997,True,True,5000,5)
    assert client_id != -1, 'Failed to connect to remote API server'
    vrep.simxSynchronous(client_id, 1)
    ENV_VAR['client_id'] = client_id
    print('Initialized simulation environment')
Пример #2
0
 def start(self):
     vrep.simxFinish(-1)
     self.client_id = vrep.simxStart('127.0.0.1', 19997, True, True, 5000,
                                     5)
     if self.client_id == 1:
         raise ConnectionError
     vrep.simxSynchronous(self.client_id, 1)
     print('Initialized simulation environment with client id {}'.format(
         self.client_id))
Пример #3
0
def incline_obj_f(x):
    print('\nParameters: ' + str(x))
    try:
        CLIENTID = ENV_VAR['client_id']
        load_scene('scenes/inclines/' + str(int(x[3])) + 'deg.ttt')
        walker = ENV_VAR['walker']

        vrep.simxSynchronous(CLIENTID, 1)
        vrep.simxStartSimulation(CLIENTID, vrep.simx_opmode_blocking)

        errorCode, start = vrep.simxGetObjectPosition(CLIENTID, \
            walker.base_handle, -1, vrep.simx_opmode_blocking)
        assert errorCode in VALID_ERROR_CODES, 'Error getting object position'

        gait = DualTripod
        gait.f = x[0]
        gait.phase_offset = x[1]
        gait.coupling_phase_biases = gait.coupling_phase_biases = \
            gait.generate_coupling(gait.phase_groupings)

        cpg = CpgController(gait)
        for _ in range(1000):
            cpg.update(plot=False)

        print('Running trial...')
        for _ in range(100):
            output = cpg.output()
            for i in range(6):
                walker.legs[i].extendZ(output[i][0] * x[2])
                walker.legs[i].extendX(output[i][1] * x[2])
            vrep.simxSynchronousTrigger(CLIENTID)
            cpg.update()

        errorCode, end = vrep.simxGetObjectPosition(CLIENTID, \
            walker.base_handle, -1, vrep.simx_opmode_blocking)

        vrep.simxStopSimulation(CLIENTID, vrep.simx_opmode_blocking)
        vrep.simxGetPingTime(CLIENTID)
        vrep.simxClearIntegerSignal(CLIENTID, '', vrep.simx_opmode_blocking)
        vrep.simxClearStringSignal(CLIENTID, '', vrep.simx_opmode_blocking)
        vrep.simxClearFloatSignal(CLIENTID, '', vrep.simx_opmode_blocking)
        # to be maximized
        print('Objective: ' + str(start[0] - end[0]))
        return np.array([start[0] - end[0]])
    except Exception:
        print('Encountered an exception, disconnecting from remote API server')
        vrep.simxStopSimulation(CLIENTID, vrep.simx_opmode_oneshot)
        vrep.simxGetPingTime(CLIENTID)
        vrep.simxFinish(CLIENTID)
        traceback.print_exc()
        exit()
Пример #4
0
        def objective(x):
            x = np.asarray(x)[0]
            print('\nParameters: ' + str(x))

            try:
                # Clean up VREP
                self.run()

                # Get starting position
                start = self.get_pos(self.walker.base_handle)

                # Assign parameters
                self.walker.set_left_bias(x[2])
                self.walker.set_right_bias(x[3])
                self.set_cpg(gait, x[0], x[1])
                for _ in range(1000):
                    self.cpg.update(plot=False)

                # Run the simulation
                print('Running trial...')
                self.walk(max_steps)

                # Calculate how far the robot walked
                end = self.get_pos(self.walker.base_handle)
                dist = self.calc_dist(start, end)
                print('Distance traveled: ' + str(dist))

                # Clean up VREP
                self.stop()

                return np.array([dist])

            except (ConnectionError, SceneLoadingError, WalkerLoadingError,
                    MotorLoadingError, HandleLoadingError) as e:
                print("Encountered an exception: {} "
                      "disconnecting from remote API server".format(e))
                vrep.simxFinish(self.client_id)
                traceback.print_exc()
                self.stop()
                raise e
Пример #5
0
    def objective(x):
        try:
            walker.reset()
            vrep.simxStartSimulation(CLIENTID, vrep.simx_opmode_blocking)
            errorCode, start = vrep.simxGetObjectPosition(CLIENTID, \
                walker.base_handle, -1, vrep.simx_opmode_blocking)
            assert errorCode in VALID_ERROR_CODES, 'Walker not found'

            x = np.asarray(x)[0]
            print('\nParameters: ' + str(x))
            cpg_gait = gait
            if parameter_mode is 'normal':
                cpg_gait.f = x[0]
                cpg_gait.phase_offset = x[1]
                walker.set_left_bias(x[2])
                walker.set_right_bias(x[3])
            elif parameter_mode is 'contextual':
                cpg_gait.f = x[0]
                cpg_gait.phase_offset = x[1]
                walker.set_left_bias(x[2])
                walker.set_right_bias(x[2])
            elif parameter_mode is 'discovery':
                cpg_gait.f = x[0]
                cpg_gait.phase_offset = x[1]
                cpg_gait.phase_groupings = {
                    0: x[2],
                    1: x[3],
                    2: x[4],
                    3: x[5],
                    4: x[6],
                    5: x[7]
                }
            else:
                raise Exception('Invalid parameter mode')

            cpg_gait.coupling_phase_biases = \
                cpg_gait.generate_coupling(cpg_gait.phase_groupings)

            cpg = CpgController(cpg_gait)
            for _ in range(1000):
                cpg.update(plot=False)

            print('Running trial...')
            for _ in range(steps):
                output = cpg.output()
                if objective_mode is not 'normal':
                    walker.update_energy()
                for i in range(6):
                    walker.legs[i].extendZ(output[i][0])
                    walker.legs[i].extendX(output[i][1])
                wait(1)
                cpg.update()

            errorCode, end = vrep.simxGetObjectPosition(CLIENTID, \
                walker.base_handle, -1, vrep.simx_opmode_blocking)


            if objective_mode is not 'single':
                total_energy = walker.calculate_energy() / 5

            vrep.simxStopSimulation(CLIENTID, vrep.simx_opmode_blocking)
            vrep.simxGetPingTime(CLIENTID)
            vrep.simxClearIntegerSignal(CLIENTID, '', vrep.simx_opmode_blocking)
            vrep.simxClearStringSignal(CLIENTID, '', vrep.simx_opmode_blocking)
            vrep.simxClearFloatSignal(CLIENTID, '', vrep.simx_opmode_blocking)

            if objective_mode is 'target':
                target = x[-2], x[-1]
                return np.array([distance(end, target)])
            else:
                distance = start[0] - end[0]
                if penalize_offset: distance += (.1 * np.abs(end[1] - start[1]))
                if objective_mode is 'single':
                    print('Distance traveled: ' + str(distance))
                    return np.array([distance])
                else:
                    print('Distance traveled: ' + str(distance))
                    print('Total power: ' + str(total_energy) + 'W')
                    return np.array([distance, total_energy])
        except Exception as e:
            print('Encountered an exception, disconnecting from remote API server')
            vrep.simxStopSimulation(CLIENTID, vrep.simx_opmode_oneshot)
            vrep.simxGetPingTime(CLIENTID)
            vrep.simxFinish(CLIENTID)
            traceback.print_exc()
            exit()
Пример #6
0
def exit_vrep():
    vrep.simxFinish(ENV_VAR['client_id'])
    print('Exited simulation environment')
Пример #7
0
def turning_obj_f(x):
    print('\nParameters: ' + str(x))
    try:
        CLIENTID = ENV_VAR['client_id']
        walker = ENV_VAR['walker']
        errorCode,baseCollectionHandle=\
            vrep.simxGetCollectionHandle(CLIENTID, 'Base', \
            vrep.simx_opmode_blocking)
        vrep.simxSynchronous(CLIENTID, 1)
        vrep.simxStartSimulation(CLIENTID, vrep.simx_opmode_blocking)

        walker.reset()

        errorCode, start = vrep.simxGetObjectPosition(CLIENTID, \
            walker.base_handle, -1, vrep.simx_opmode_blocking)
        assert errorCode in VALID_ERROR_CODES, 'Error getting object position'

        q1 = vrep.simxGetObjectGroupData(CLIENTID, baseCollectionHandle, \
            7, vrep.simx_opmode_blocking)
        o1 = quatToDirection(q1[3]) + np.pi

        x[2] = 15 + (3 * x[2])
        x[3] = x[3] * 2 * np.pi / 10

        gait = DualTripod
        gait.f = x[2]
        gait.phase_offset = x[3]
        gait.coupling_phase_biases=gait.generate_coupling(gait.phase_groupings)

        walker.set_left_bias(x[0] / 10)
        walker.set_right_bias(x[1] / 10)

        cpg = CpgController(gait)
        for _ in range(1000):
            cpg.update(plot=False)

        print('Running trial...')
        for _ in range(100):
            output = cpg.output()
            walker.update_energy()
            for i in range(6):
                walker.legs[i].extendZ(output[i][0])
                walker.legs[i].extendX(output[i][1])
            vrep.simxSynchronousTrigger(CLIENTID)
            cpg.update()

        total_energy = walker.calculate_energy() / 5

        errorCode, end = vrep.simxGetObjectPosition(CLIENTID, \
            walker.base_handle, -1, vrep.simx_opmode_blocking)
        q2 = vrep.simxGetObjectGroupData(CLIENTID, baseCollectionHandle, \
            7, vrep.simx_opmode_blocking)
        o2 = quatToDirection(q2[3]) + np.pi

        vrep.simxStopSimulation(CLIENTID, vrep.simx_opmode_blocking)
        vrep.simxGetPingTime(CLIENTID)
        vrep.simxClearIntegerSignal(CLIENTID, '', vrep.simx_opmode_blocking)
        vrep.simxClearStringSignal(CLIENTID, '', vrep.simx_opmode_blocking)
        vrep.simxClearFloatSignal(CLIENTID, '', vrep.simx_opmode_blocking)
        displacement = np.array(end) - np.array(start)
        orientation = o2 - o1

        info = np.array([total_energy, displacement[0], displacement[1], \
            orientation])

        objective = distance(displacement, (x[4], x[5]))

        print('Orientation change: ' + str(orientation))
        print('Total power: ' + str(total_energy) + 'W')
        print('Target offset: ' + str(objective))
        return np.array([objective]), info

    except Exception:
        print('Encountered an exception, disconnecting from remote API server')
        vrep.simxStopSimulation(CLIENTID, vrep.simx_opmode_oneshot)
        vrep.simxGetPingTime(CLIENTID)
        vrep.simxFinish(CLIENTID)
        traceback.print_exc()
        exit()
Пример #8
0
        def objective(x, cache_walker=True):
            """
            x[0] = cpg frequency [1, 45]
            x[1] = cpg phase offset [-np.pi, np.pi]
            x[2] = left bias of walker [0, 1]
            x[3] = right bias of walker [0, 1]
            x[4] = front pair leg scaling [.8, 1.2]
            x[5] = middle pair leg scaling [.8, 1.2]
            x[6] = rear pair leg scaling [.8, 1.2]
            :return:
            """
            self.start()
            x = np.asarray(x)[0]
            print('\nParameters: ' + str(x))
            try:
                # Clean up VREP
                if cache_walker and \
                        self.last_walker_params is not None and \
                        np.array_equal(x[4:], self.last_walker_params):
                    self.run()
                    self.load_prev_walker()
                else:
                    self.exit()
                    self.start()
                    self.load_scene()
                    self.load_walker()
                    self.run()

                    # Assign parameters
                    self.scale_walker(x[4:])

                    # Cache walker
                    self.walker_filename = "logs/models/{}.ttm" \
                        .format(time.strftime("%Y.%m.%d-%H.%M.%S"))
                    self.last_walker_params = x[4:]
                    self.save_walker(self.walker_filename)

                # Set the other parameters
                self.set_cpg(gait, x[0], x[1])
                self.walker.set_left_bias(x[2])
                self.walker.set_right_bias(x[3])

                # Get starting position
                start = self.get_pos(self.walker.base_handle)

                # Run the simulation
                print('Running trial...')
                self.walk(max_steps)

                # Calculate how far the robot walked
                end = self.get_pos(self.walker.base_handle)
                dist = self.calc_dist(start, end)
                print("Distance traveled: {}".format(dist))

                # Clean up VREP
                self.remove_walker()
                self.close_scene()
                self.stop()
                self.exit()

                return np.array([dist])

            except SceneLoadingError as e:
                print(str(e))
                self.stop()
                self.close_scene()

            except (ConnectionError, WalkerLoadingError, MotorLoadingError,
                    HandleLoadingError) as e:
                print("Encountered an exception: {} "
                      "disconnecting from remote API server".format(e))
                vrep.simxFinish(self.client_id)
                traceback.print_exc()
                self.stop()
                self.close_scene()
                raise e
Пример #9
0
 def exit(self):
     print("Exiting VREP")
     vrep.simxFinish(self.client_id)
Пример #10
0
        def objective(x, cache_walker=True):
            """
            8 sw, 6 hw
            x[0] = f # Frequency [1, 45]
            x[1] = phase_offset # phase difference btwn leg and foot. [-pi, pi]
            x[2] = R_l # Sequence of leg extension length. [0, .04]
            x[3] = R_f # Sequence of foot extension lengths. [0, .04]
            x[4] = X_l # Sequence of leg offset values. [0, .04]
            x[5] = X_f # Sequence of foot offset values. [0, .04]
            x[6] = left bias of walker [.5, 1]
            x[7] = right bias of walker [.5, 1]
            x[8] = front left leg scaling [.8, 1.2]
            x[9] = front right leg scaling [.8, 1.2]
            x[11] = middle left  leg scaling [.8, 1.2]
            x[12] = middle right leg scaling [.8, 1.2]
            x[13] = rear left pair leg scaling [.8, 1.2]
            x[14] = rear right pair leg scaling [.8, 1.2]
            walker params = x[8:]
            :return:
            """
            if not is_1D(x):
                x = np.asarray(x)[0]
            else:
                x = np.asarray(x)

            self.start()

            gait_params = x[:6]
            left_bias = x[6]
            right_bias = x[7]
            walker_params = x[8:]
            print("walker params", walker_params)

            print('\nParameters: ' + str(x))
            try:
                # Clean up VREP
                if cache_walker and \
                        self.last_walker_params is not None and \
                        np.array_equal(walker_params, self.last_walker_params):
                    self.run()
                    self.load_prev_walker()
                else:
                    self.exit()
                    self.start()
                    self.load_scene()
                    self.load_walker()
                    self.run()

                    # Assign parameters
                    self.scale_walker(walker_params)

                    # Cache walker
                    self.walker_filename = "logs/models/{}.ttm" \
                        .format(time.strftime("%Y.%m.%d-%H.%M.%S"))
                    self.last_walker_params = walker_params
                    self.save_walker(self.walker_filename)

                # Set the other parameters
                self.set_cpg(build_gait(gait_params))
                self.walker.set_left_bias(left_bias)
                self.walker.set_right_bias(right_bias)

                # Get starting position
                start = self.get_pos(self.walker.base_handle)

                # Run the simulation
                print('Running trial...')
                self.walk(max_steps)

                # Calculate how far the robot walked
                end = self.get_pos(self.walker.base_handle)
                dist = self.calc_dist(start, end)
                print("Distance traveled: {}".format(dist))

                # Clean up VREP
                self.remove_walker()
                self.close_scene()
                self.stop()
                self.exit()

                return np.array([dist])

            except SceneLoadingError as e:
                print(str(e))
                self.stop()
                self.close_scene()

            except (ConnectionError, WalkerLoadingError, MotorLoadingError,
                    HandleLoadingError) as e:
                print("Encountered an exception: {} "
                      "disconnecting from remote API server".format(e))
                vrep.simxFinish(self.client_id)
                traceback.print_exc()
                self.stop()
                self.close_scene()
                raise e