コード例 #1
0
ファイル: distance.py プロジェクト: simondlevy/gym-copter
    def reset(self):

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics
        state = np.zeros(12)
        self.dynamics.setState(state)

        return self.step(np.array([0, 0, 0, 0]))[0]
コード例 #2
0
    def reset(self):

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics with random perturbations
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_X] = self.INITIAL_RANDOM_OFFSET * np.random.randn()
        state[d.STATE_Y] = self.INITIAL_RANDOM_OFFSET * np.random.randn()
        state[d.STATE_Z] = -self.INITIAL_ALTITUDE
        self.dynamics.setState(state)

        return self.step(np.array([0, 0, 0, 0]))[0]
コード例 #3
0
ファイル: lander2d.py プロジェクト: coletta1/gym-copter
    def reset(self):

        self._destroy()

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics with random perturbation
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_Y] = 0
        state[d.STATE_Z] = -self.INITIAL_ALTITUDE
        self.dynamics.setState(state)
        self.dynamics.perturb(
            np.array([0, self._perturb(),
                      self._perturb(), 0, 0, 0]))

        return self.step(np.array([0, 0]))[0]
コード例 #4
0
ファイル: distance.py プロジェクト: simondlevy/gym-copter
class Distance(gym.Env, EzPickle):

    FRAMES_PER_SECOND = 50

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': FRAMES_PER_SECOND
    }

    def __init__(self):

        EzPickle.__init__(self)
        self.seed()

        # Observation is all state values except yaw and its derivative
        self.observation_space = spaces.Box(-np.inf,
                                            np.inf,
                                            shape=(10, ),
                                            dtype=np.float32)

        # Action is motor values
        self.action_space = spaces.Box(-1, +1, (4, ), dtype=np.float32)

        # Support for rendering
        self.renderer = None
        self.pose = None

        self.reset()

    def seed(self, seed=None):
        self.np_random, seed = seeding.np_random(seed)
        return [seed]

    def reset(self):

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics
        state = np.zeros(12)
        self.dynamics.setState(state)

        return self.step(np.array([0, 0, 0, 0]))[0]

    def step(self, action):

        # Abbreviation
        d = self.dynamics

        d.setMotors(action)
        d.update()

        # Get new state from dynamics
        posx, velx, posy, vely, posz, velz, phi, velphi, theta, veltheta, psi, _ = d.getState(
        )

        # Set pose in display
        self.pose = posx, posy, posz, phi, theta, psi

        # Convert state to usable form
        state = np.array(
            [posx, velx, posy, vely, posz, velz, phi, velphi, theta, veltheta])

        # Reward is a simple penalty for overall distance and velocity
        shaping = np.sqrt(posx**2 + posy**2)

        reward = (shaping -
                  self.prev_shaping) if (self.prev_shaping is not None) else 0

        self.prev_shaping = shaping

        done = False

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):

        from gym_copter.rendering.threed import ThreeDDistanceRenderer

        # Create renderer if not done yet
        if self.renderer is None:
            self.renderer = ThreeDDistanceRenderer(self, self.LANDING_RADIUS)

        return self.renderer.render()

    def close(self):

        return
コード例 #5
0
ファイル: copter.py プロジェクト: simondlevy/gym-copter
    def _reset(self):

        self.state = np.zeros(self.statedims)
        self.dynamics = DJIPhantomDynamics()
        self.tick = 0
        self.done = False
コード例 #6
0
ファイル: lander2d.py プロジェクト: coletta1/gym-copter
class Lander2D(gym.Env, EzPickle):

    # Parameters to adjust
    INITIAL_RANDOM_FORCE = 30  # perturbation for initial position
    INITIAL_ALTITUDE = 10
    LANDING_RADIUS = 2
    PENALTY_FACTOR = 12  # designed so that maximal penalty is around 100
    BOUNDS = 10
    OUT_OF_BOUNDS_PENALTY = 100
    INSIDE_RADIUS_BONUS = 100
    FRAMES_PER_SECOND = 50

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': FRAMES_PER_SECOND
    }

    def __init__(self):

        EzPickle.__init__(self)
        self.seed()
        self.viewer = None

        self.prev_reward = None

        # useful range is -1 .. +1, but spikes can be higher
        self.observation_space = spaces.Box(-np.inf,
                                            np.inf,
                                            shape=(6, ),
                                            dtype=np.float32)

        # Action is two floats [throttle, roll]
        self.action_space = spaces.Box(-1, +1, (2, ), dtype=np.float32)

        # Support for rendering
        self.viewer = None
        self.pose = None
        self.spinning = False

        self.reset()

    def seed(self, seed=None):
        self.np_random, seed = seeding.np_random(seed)
        return [seed]

    def reset(self):

        self._destroy()

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics with random perturbation
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_Y] = 0
        state[d.STATE_Z] = -self.INITIAL_ALTITUDE
        self.dynamics.setState(state)
        self.dynamics.perturb(
            np.array([0, self._perturb(),
                      self._perturb(), 0, 0, 0]))

        return self.step(np.array([0, 0]))[0]

    def step(self, action):

        # Abbreviation
        d = self.dynamics
        status = d.getStatus()

        # Stop motors after safe landing
        if status == d.STATUS_LANDED:
            d.setMotors(np.zeros(4))
            self.spinning = False

        # In air, set motors from action
        else:
            m = np.clip(action, 0, 1)  # keep motors in interval [0,1]
            d.setMotors([m[0], m[1], m[1], m[0]])
            self.spinning = sum(m) > 0
            d.update()

        # Get new state from dynamics
        _, _, posy, vely, posz, velz, phi, velphi = d.getState()[:8]

        # Set lander pose for viewer
        self.pose = posy, posz, phi

        # Convert state to usable form
        state = np.array([posy, vely, posz, velz, phi, velphi])

        # A simple penalty for overall distance and velocity
        shaping = -self.PENALTY_FACTOR * np.sqrt(np.sum(state[0:4]**2))

        reward = ((shaping - self.prev_shaping) if
                  (self.prev_shaping is not None) else 0)

        self.prev_shaping = shaping

        # Assume we're not done yet
        done = False

        # Lose bigly if we go outside window
        if abs(posy) >= self.BOUNDS:
            done = True
            reward -= self.OUT_OF_BOUNDS_PENALTY

        else:

            # It's all over once we're on the ground
            if status == d.STATUS_LANDED:

                done = True
                self.spinning = False

                # Win bigly we land safely between the flags
                if abs(posy) < self.LANDING_RADIUS:

                    reward += self.INSIDE_RADIUS_BONUS

            elif status == d.STATUS_CRASHED:

                # Crashed!
                done = True
                self.spinning = False

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):

        from gym_copter.rendering.twod import TwoDLanderRenderer

        # Create viewer if not done yet
        if self.viewer is None:
            self.viewer = TwoDLanderRenderer(self.LANDING_RADIUS)

        return self.viewer.render(mode, self.pose, self.spinning)

    def close(self):
        if self.viewer is not None:
            self.viewer.close()
            self.viewer = None

    def _perturb(self):

        return np.random.uniform(-self.INITIAL_RANDOM_FORCE,
                                 +self.INITIAL_RANDOM_FORCE)

    def _destroy(self):
        if self.viewer is not None:
            self.viewer.close()
コード例 #7
0
ファイル: lander1d.py プロジェクト: coletta1/gym-copter
class Lander1D(gym.Env, EzPickle):

    # Parameters to adjust
    INITIAL_ALTITUDE = 10
    LANDING_RADIUS = 2  # for display purposes only
    PENALTY_FACTOR = 12  # designed so that maximal penalty is around 100
    BOUNDS = 10
    SAFE_LANDING_BONUS = 100
    FRAMES_PER_SECOND = 50

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': FRAMES_PER_SECOND
    }

    def __init__(self):

        EzPickle.__init__(self)
        self.viewer = None

        self.prev_reward = None

        # Observation is altitude and its rate of change
        self.observation_space = spaces.Box(-np.inf,
                                            np.inf,
                                            shape=(2, ),
                                            dtype=np.float32)

        # Action is one value [throttle]
        self.action_space = spaces.Box(-1, +1, (1, ), dtype=np.float32)

        # Support for rendering
        self.renderer = None
        self.pose = None
        self.spinning = False

        self.reset()

    def reset(self):

        self._destroy()

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics with random perturbation
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_Y] = 0
        state[d.STATE_Z] = -self.INITIAL_ALTITUDE
        self.dynamics.setState(state)

        return self.step(np.array([0]))[0]

    def step(self, action):

        # Abbreviation
        d = self.dynamics
        status = d.getStatus()

        # Stop motors after safe landing
        if status == d.STATUS_LANDED:
            d.setMotors(np.zeros(4))
            self.spinning = False

        # In air, set motors from action
        else:

            # Keep action in interval [0,1]
            t = np.clip(action[0], 0, 1)
            d.setMotors([t, t, t, t])
            self.spinning = t > 0
            d.update()

        # Get new state from dynamics
        _, _, posy, _, posz, velz, phi, _ = d.getState()[:8]

        # Set lander pose for renderer
        self.pose = posy, posz, phi

        # Convert state to usable form
        state = np.array([posz, velz])

        # Reward is a simple penalty for overall distance and velocity
        shaping = -self.PENALTY_FACTOR * np.sqrt(np.sum(state[0:4]**2))

        reward = (shaping -
                  self.prev_shaping) if (self.prev_shaping is not None) else 0

        self.prev_shaping = shaping

        # Assume we're not done yet
        done = False

        # Lose bigly if we go outside window
        if abs(posy) >= self.BOUNDS:
            done = True

        else:

            # It's all over once we're on the ground
            if status == d.STATUS_LANDED:

                done = True

                reward += self.SAFE_LANDING_BONUS

            elif status == d.STATUS_CRASHED:

                # Crashed!
                done = True

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):

        from gym_copter.rendering.twod import TwoDLanderRenderer

        # Create renderer if not done yet
        if self.renderer is None:
            self.renderer = TwoDLanderRenderer(self.LANDING_RADIUS)

        d = self.dynamics

        return self.renderer.render(mode, self.pose, self.spinning)

    def close(self):
        if self.renderer is not None:
            self.renderer.close()
            self.renderer = None

    def _destroy(self):
        if self.renderer is not None:
            self.renderer.close()
コード例 #8
0
class Lander3D(gym.Env, EzPickle):

    # Parameters to adjust
    INITIAL_ALTITUDE = 5
    INITIAL_RANDOM_OFFSET = 2.5
    XY_PENALTY_FACTOR = 25  # designed so that maximal penalty is around 100
    PITCH_ROLL_PENALTY_FACTOR = 0  # 250
    YAW_PENALTY_FACTOR = 50
    MOTOR_PENALTY_FACTOR = 0.03
    BOUNDS = 10
    OUT_OF_BOUNDS_PENALTY = 100
    RESTING_DURATION = 1.0  # render a short while after successful landing
    FRAMES_PER_SECOND = 50
    MAX_ANGLE = 45  # big penalty if roll or pitch angles go beyond this
    EXCESS_ANGLE_PENALTY = 100

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': FRAMES_PER_SECOND
    }

    def __init__(self):

        EzPickle.__init__(self)
        self.seed()

        self.prev_reward = None

        # Observation is all state values
        self.observation_space = (spaces.Box(-np.inf,
                                             np.inf,
                                             shape=(12, ),
                                             dtype=np.float32))

        # Action is four floats (one per motor)
        self.action_space = spaces.Box(-1, +1, (4, ), dtype=np.float32)

        # Support for rendering
        self.viewer = None
        self.pose = None

        # Pre-convert max-angle degrees to radian
        self.max_angle = np.radians(self.MAX_ANGLE)

        self.reset()

    def seed(self, seed=None):
        self.np_random, seed = seeding.np_random(seed)
        return [seed]

    def reset(self):

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics with random perturbations
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_X] = self.INITIAL_RANDOM_OFFSET * np.random.randn()
        state[d.STATE_Y] = self.INITIAL_RANDOM_OFFSET * np.random.randn()
        state[d.STATE_Z] = -self.INITIAL_ALTITUDE
        self.dynamics.setState(state)

        return self.step(np.array([0, 0, 0, 0]))[0]

    def step(self, action):

        # Abbreviation
        d = self.dynamics
        status = d.getStatus()

        # Keep motors in interval [0,1]
        motors = (np.zeros(4) if status == d.STATUS_LANDED else np.clip(
            action, 0, 1))

        d.setMotors(motors)

        # Update dynamics if airborne
        if status != d.STATUS_LANDED:
            d.update()

        # Get new state from dynamics
        state = np.array(d.getState())

        # Extract pose from state
        x, y, z, phi, theta, psi = state[0::2]

        # Set lander pose in display
        self.pose = x, y, z, phi, theta, psi

        # Reward is a simple penalty for overall distance and angle and their
        # first derivatives, plus a bit more for running motors (discourage
        # endless hover)
        shaping = -(
            self.XY_PENALTY_FACTOR * np.sqrt(np.sum(state[0:6]**2)) +
            self.PITCH_ROLL_PENALTY_FACTOR * np.sqrt(np.sum(state[6:10]**2)) +
            self.YAW_PENALTY_FACTOR * np.sqrt(np.sum(state[10:12]**2)) +
            self.MOTOR_PENALTY_FACTOR * np.sum(motors))

        reward = ((shaping - self.prev_shaping) if
                  (self.prev_shaping is not None) else 0)
        self.prev_shaping = shaping

        # Assume we're not done yet
        done = False

        # Lose bigly if we go out of bounds
        if abs(x) >= self.BOUNDS or abs(y) >= self.BOUNDS:
            done = True
            reward = -self.OUT_OF_BOUNDS_PENALTY

        # Lose bigly for excess roll or pitch
        if abs(phi) >= self.max_angle or abs(theta) >= self.max_angle:
            done = True
            reward = -self.OUT_OF_BOUNDS_PENALTY

        # It's all over once we're on the ground
        if status == d.STATUS_LANDED:

            done = True

            # Different subclasses add different bonuses for proximity to
            # center
            reward += self._get_bonus(x, y)

        elif status == d.STATUS_CRASHED:

            # Crashed!
            done = True

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):
        '''
        Returns None because we run viewer on a separate thread
        '''
        return None

    def close(self):
        return

    def _get_bonus(self, x, y):

        # Bonus is proximity to center
        return self.BOUNDS - np.sqrt(x**2 + y**2)
コード例 #9
0
class Lander3DSimple(gym.Env, EzPickle):

    # Parameters to adjust  
    INITIAL_RANDOM_OFFSET      = 2.5  # perturbation factor for initial horizontal position
    INITIAL_ALTITUDE           = 5
    LANDING_RADIUS             = 2
    XY_PENALTY_FACTOR          = 25   # designed so that maximal penalty is around 100
    PITCH_ROLL_PENALTY_FACTOR  = 250   
    BOUNDS                     = 10
    OUT_OF_BOUNDS_PENALTY      = 100
    INSIDE_RADIUS_BONUS        = 100
    RESTING_DURATION           = 1.0  # for rendering for a short while after successful landing
    FRAMES_PER_SECOND          = 50
    MAX_ANGLE                  = 45   # big penalty if roll or pitch angles go beyond this
    EXCESS_ANGLE_PENALTY       = 100

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second' : FRAMES_PER_SECOND
    }

    def __init__(self):

        EzPickle.__init__(self)
        self.seed()

        self.prev_reward = None

        # Observation is all state values
        self.observation_space = spaces.Box(-np.inf, np.inf, shape=(10,), dtype=np.float32)

        # Action is three floats (throttle, roll, pitch)
        self.action_space = spaces.Box(-1, +1, (3,), dtype=np.float32)

        # Support for rendering
        self.renderer = None
        self.pose = None

        # Pre-convert max-angle degrees to radian
        self.max_angle = np.radians(self.MAX_ANGLE)

        self.reset()

    def seed(self, seed=None):
        self.np_random, seed = seeding.np_random(seed)
        return [seed]

    def reset(self):

        self.prev_shaping = None

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics(self.FRAMES_PER_SECOND)

        # Initialize custom dynamics with random perturbation
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_X] =  self.INITIAL_RANDOM_OFFSET #* np.random.randn()
        state[d.STATE_Y] =  self.INITIAL_RANDOM_OFFSET #* np.random.randn()
        state[d.STATE_Z] = -self.INITIAL_ALTITUDE
        self.dynamics.setState(state)

        return self.step(np.array([-1, 0, 0]))[0]

    def step(self, action):

        # Use mixer to convert demands into motor values
        t = np.clip(action[0], 0, 1) # truncate throttle below 0
        r = action[1]
        p = action[2]
        motors = [t-r-p, t+r+p, t+r-p, t-r+p] 

        # Abbreviation
        d = self.dynamics
        status = d.getStatus()

        # Stop motors after safe landing
        if status == d.STATUS_LANDED:
            d.setMotors(np.zeros(4))

        # In air, set motors from action
        else:
            d.setMotors(np.clip(motors, 0, 1))    # keep motors in interval [0,1]
            d.update()

        # Get new state from dynamics, removing yaw
        state = np.array(d.getState())[:10]

        # Extract pose from state
        x, y, z, phi, theta = state[0::2]

        # Set lander pose in display (with zero for yaw)
        self.pose = x, y, z, phi, theta, 0

        # Reward is a simple penalty for overall distance and angle and their first derivatives
        shaping = -(
                self.XY_PENALTY_FACTOR * np.sqrt(np.sum(state[0:6]**2)) + 
                self.PITCH_ROLL_PENALTY_FACTOR * np.sqrt(np.sum(state[6:10]**2))
                )
                                                                  
        reward = (shaping - self.prev_shaping) if (self.prev_shaping is not None) else 0
        self.prev_shaping = shaping

        # Assume we're not done yet
        done = False

        # Lose bigly if we go out of bounds
        if abs(x) >= self.BOUNDS or abs(y) >= self.BOUNDS:
            done = True
            reward = -self.OUT_OF_BOUNDS_PENALTY

        # Lose bigly for excess roll or pitch 
        if abs(phi) >= self.max_angle or abs(theta) >= self.max_angle:
            done = True
            reward = -self.OUT_OF_BOUNDS_PENALTY

        # It's all over once we're on the ground
        if status == d.STATUS_LANDED:

            done = True

            # Win bigly we land safely between the flags
            if x**2+y**2 < self.LANDING_RADIUS**2: 

                reward += self.INSIDE_RADIUS_BONUS

        elif status == d.STATUS_CRASHED:

            # Crashed!
            done = True

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):

        from gym_copter.rendering.threed import ThreeDLanderRenderer

        # Create renderer if not done yet
        if self.renderer is None:
            self.renderer = ThreeDLanderRenderer(self, self.LANDING_RADIUS)

        return self.renderer.render()

    def close(self):

        return
コード例 #10
0
ファイル: lander2d.py プロジェクト: GayanSamuditha/gym-copter
class CopterLander2D(gym.Env, EzPickle):

    FPS = 50
    SCALE = 30.0  # affects how fast-paced the game is, forces should be adjusted as well

    # Criteria for a successful landing
    LANDING_POS_Y = 4.15
    LANDING_VEL_X = 2.0
    LANDING_ANGLE = 0.05

    # Initial velocity perturbation factor
    INITIAL_RANDOM_VELOCITY = 0.75

    # Vehicle display properties ---------------------------------------------------------

    LANDER_POLY = [(-14, +17), (-17, 0), (-17, -10), (+17, -10), (+17, 0),
                   (+14, +17)]

    LEG_X = 12
    LEG_Y = -7
    LEG_W = 3
    LEG_H = 20

    MOTOR_X = 25
    MOTOR_Y = 7
    MOTOR_W = 4
    MOTOR_H = 5

    BLADE_X = 25
    BLADE_Y = 8
    BLADE_W = 20
    BLADE_H = 2

    BLADE1L_POLY = [
        (BLADE_X, BLADE_Y),
        (BLADE_X - BLADE_W / 2, BLADE_Y + BLADE_H),
        (BLADE_X - BLADE_W, BLADE_Y),
        (BLADE_X - BLADE_W / 2, BLADE_Y + -BLADE_H),
    ]

    BLADE1R_POLY = [
        (BLADE_X, BLADE_Y),
        (BLADE_X + BLADE_W / 2, BLADE_Y + BLADE_H),
        (BLADE_X + BLADE_W, BLADE_Y),
        (BLADE_X + BLADE_W / 2, BLADE_Y + -BLADE_H),
    ]

    BLADE2L_POLY = [
        (-BLADE_X, BLADE_Y),
        (-(BLADE_X + BLADE_W / 2), BLADE_Y + BLADE_H),
        (-(BLADE_X + BLADE_W), BLADE_Y),
        (-(BLADE_X + BLADE_W / 2), BLADE_Y + -BLADE_H),
    ]

    BLADE2R_POLY = [
        (-BLADE_X, BLADE_Y),
        (-BLADE_X + BLADE_W / 2, BLADE_Y + BLADE_H),
        (-BLADE_X + BLADE_W, BLADE_Y),
        (-BLADE_X + BLADE_W / 2, BLADE_Y + -BLADE_H),
    ]

    HULL_POLY = [
        (-30, 0),
        (-4, +4),
        (+4, +4),
        (+30, 0),
        (+4, -14),
        (-4, -14),
    ]

    LEG1_POLY = [(-LEG_X, LEG_Y), (-LEG_X + LEG_W, LEG_Y),
                 (-LEG_X + LEG_W, LEG_Y - LEG_H), (-LEG_X, LEG_Y - LEG_H)]

    LEG2_POLY = [(+LEG_X, LEG_Y), (+LEG_X + LEG_W, LEG_Y),
                 (+LEG_X + LEG_W, LEG_Y - LEG_H), (+LEG_X, LEG_Y - LEG_H)]

    MOTOR1_POLY = [(+MOTOR_X, MOTOR_Y), (+MOTOR_X + MOTOR_W, MOTOR_Y),
                   (+MOTOR_X + MOTOR_W, MOTOR_Y - MOTOR_H),
                   (+MOTOR_X, MOTOR_Y - MOTOR_H)]

    MOTOR2_POLY = [(-MOTOR_X, MOTOR_Y), (-MOTOR_X + MOTOR_W, MOTOR_Y),
                   (-MOTOR_X + MOTOR_W, MOTOR_Y - MOTOR_H),
                   (-MOTOR_X, MOTOR_Y - MOTOR_H)]

    VIEWPORT_W = 600
    VIEWPORT_H = 400

    TERRAIN_CHUNKS = 11

    SKY_COLOR = 0.5, 0.8, 1.0
    GROUND_COLOR = 0.5, 0.7, 0.3
    FLAG_COLOR = 0.8, 0.0, 0.0
    VEHICLE_COLOR = 1.0, 1.0, 1.0
    MOTOR_COLOR = 0.5, 0.5, 0.5
    PROP_COLOR = 0.0, 0.0, 0.0
    OUTLINE_COLOR = 0.0, 0.0, 0.0

    # -------------------------------------------------------------------------------------

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': FPS
    }

    def __init__(self):
        EzPickle.__init__(self)
        self.seed()
        self.viewer = None

        self.world = Box2D.b2World()
        self.ground = None
        self.lander = None

        self.prev_reward = None

        # useful range is -1 .. +1, but spikes can be higher
        self.observation_space = spaces.Box(-np.inf,
                                            np.inf,
                                            shape=(6, ),
                                            dtype=np.float32)

        # Action is two floats [main engine, left-right engines].
        # Main engine: -1..0 off, 0..+1 throttle from 50% to 100% power. Engine can't work with less than 50% power.
        # Left-right:  -1.0..-0.5 fire left engine, +0.5..+1.0 fire right engine, -0.5..0.5 off
        self.action_space = spaces.Box(-1, +1, (2, ), dtype=np.float32)

        self.reset()

    def seed(self, seed=None):
        self.np_random, seed = seeding.np_random(seed)
        return [seed]

    def _destroy(self):
        if not self.ground: return
        self.world.contactListener = None
        self.world.DestroyBody(self.ground)
        self.ground = None
        self.world.DestroyBody(self.lander)
        self.lander = None

    def reset(self):
        self._destroy()
        self.prev_shaping = None

        self.startpos = None

        W = self.VIEWPORT_W / self.SCALE
        H = self.VIEWPORT_H / self.SCALE

        # terrain
        height = self.np_random.uniform(0,
                                        H / 2,
                                        size=(self.TERRAIN_CHUNKS + 1, ))
        chunk_x = [
            W / (self.TERRAIN_CHUNKS - 1) * i
            for i in range(self.TERRAIN_CHUNKS)
        ]
        self.helipad_x1 = chunk_x[self.TERRAIN_CHUNKS // 2 - 1]
        self.helipad_x2 = chunk_x[self.TERRAIN_CHUNKS // 2 + 1]
        self.helipad_y = H / 4
        height[self.TERRAIN_CHUNKS // 2 - 2] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 - 1] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 + 0] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 + 1] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 + 2] = self.helipad_y
        smooth_y = [
            0.33 * (height[i - 1] + height[i + 0] + height[i + 1])
            for i in range(self.TERRAIN_CHUNKS)
        ]

        self.ground = self.world.CreateStaticBody(shapes=edgeShape(
            vertices=[(0, 0), (W, 0)]))
        self.sky_polys = []
        for i in range(self.TERRAIN_CHUNKS - 1):
            p1 = (chunk_x[i], smooth_y[i])
            p2 = (chunk_x[i + 1], smooth_y[i + 1])
            self.ground.CreateEdgeFixture(vertices=[p1, p2],
                                          density=0,
                                          friction=0.1)
            self.sky_polys.append([p1, p2, (p2[0], H), (p1[0], H)])

        self.startpos = self.VIEWPORT_W / self.SCALE / 2, self.VIEWPORT_H / self.SCALE

        self.lander = self.world.CreateDynamicBody(
            position=self.startpos,
            angle=0.0,
            fixtures=[
                fixtureDef(shape=polygonShape(vertices=[(x / self.SCALE,
                                                         y / self.SCALE)
                                                        for x, y in poly]),
                           density=0.0)
                for poly in [
                    self.HULL_POLY, self.LEG1_POLY, self.LEG2_POLY,
                    self.MOTOR1_POLY, self.MOTOR2_POLY, self.BLADE1L_POLY,
                    self.BLADE1R_POLY, self.BLADE2L_POLY, self.BLADE2R_POLY
                ]
            ])

        # By showing props periodically, we can emulate prop rotation
        self.props_visible = 0

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics()

        # Initialize custom dynamics with slight velocity perturbation
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_Y] = self.startpos[
            0]  # 3D copter Y comes from 2D copter X
        state[d.STATE_Z] = -self.startpos[
            1]  # 3D copter Z comes from 2D copter Y, negated for NED
        state[d.STATE_Y_DOT] = self.INITIAL_RANDOM_VELOCITY * np.random.randn()
        state[d.STATE_Z_DOT] = self.INITIAL_RANDOM_VELOCITY * np.random.randn()
        self.dynamics.setState(state)

        return self.step(np.array([0, 0]))[0]

    def step(self, action):

        # Map throttle demand from [-1,+1] to [0,1]
        throttle = (action[0] + 1) / 2

        # Abberviation
        d = self.dynamics

        # Set motors from demands
        roll = action[1]
        d.setMotors(
            np.clip([
                throttle - roll, throttle + roll, throttle + roll,
                throttle - roll
            ], 0, 1))

        # Update dynamics
        d.update(1. / self.FPS)

        # Get new state from dynamics
        x = d.getState()

        # Parse out state into elements
        posx = x[d.STATE_Y]
        posy = -x[d.STATE_Z]
        velx = x[d.STATE_Y_DOT]
        vely = -x[d.STATE_Z_DOT]
        angle = x[d.STATE_PHI]
        angularVelocity = x[d.STATE_PHI_DOT]

        # Set lander pose in display
        self.lander.position = posx, posy
        self.lander.angle = -angle

        # Convert state to usable form
        state = ((posx - self.VIEWPORT_W / self.SCALE / 2) /
                 (self.VIEWPORT_W / self.SCALE / 2),
                 (posy -
                  (self.helipad_y)) / (self.VIEWPORT_H / self.SCALE / 2),
                 velx * (self.VIEWPORT_W / self.SCALE / 2) / self.FPS,
                 vely * (self.VIEWPORT_H / self.SCALE / 2) / self.FPS, angle,
                 20.0 * angularVelocity / self.FPS)

        # Shape the reward
        reward = 0
        shaping = 0
        shaping -= 100 * np.sqrt(
            state[0]**2 +
            state[1]**2)  # Lose points for altitude and vertical drop rate'
        shaping -= 100 * np.sqrt(
            state[2]**2 + state[3]**2
        )  # Lose points for distance from X center and horizontal velocity

        if self.prev_shaping is not None:
            reward = shaping - self.prev_shaping
        self.prev_shaping = shaping

        # Assume we're not done yet
        done = False

        # Lose bigly if we go outside window
        if abs(state[0]) >= 1.0:
            done = True
            reward = -100
        '''
        print('posy=%3.3f (%3.3f)\tvelx=%+3.3f (%3.3f)\tang=%+3.3f (%3.3f) %d' % 
                (posy, self.LANDING_POS_Y, velx, self.LANDING_VEL_X, self.lander.angle, self.LANDING_ANGLE, 
                    self.helipad_x1 < posx < self.helipad_x2))
        '''

        # It's all over once we're on the ground
        if self._on_ground():

            done = True

            landed_safely = abs(velx) < self.LANDING_VEL_X and abs(
                self.lander.angle
            ) < self.LANDING_ANGLE and self.helipad_x1 < posx < self.helipad_x2

            # More reward if we land safely, less otherwise
            reward = reward + 100 if landed_safely else -50

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):

        from gym.envs.classic_control import rendering

        if self.viewer is None:
            self.viewer = rendering.Viewer(self.VIEWPORT_W, self.VIEWPORT_H)
            self.viewer.set_bounds(0, self.VIEWPORT_W / self.SCALE, 0,
                                   self.VIEWPORT_H / self.SCALE)

        # Draw ground as background
        self.viewer.draw_polygon([(0, 0), (self.VIEWPORT_W, 0),
                                  (self.VIEWPORT_W, self.VIEWPORT_H),
                                  (0, self.VIEWPORT_H)],
                                 color=self.GROUND_COLOR)

        # Draw sky
        for p in self.sky_polys:
            self.viewer.draw_polygon(p, color=self.SKY_COLOR)

        # Draw flags
        for x in [self.helipad_x1, self.helipad_x2]:
            flagy1 = self.helipad_y
            flagy2 = flagy1 + 50 / self.SCALE
            self.viewer.draw_polyline([(x, flagy1), (x, flagy2)],
                                      color=(1, 1, 1))
            self.viewer.draw_polygon(
                [(x, flagy2), (x, flagy2 - 10 / self.SCALE),
                 (x + 25 / self.SCALE, flagy2 - 5 / self.SCALE)],
                color=self.FLAG_COLOR)

        # Draw copter
        self._show_fixture(1, self.VEHICLE_COLOR)
        self._show_fixture(2, self.VEHICLE_COLOR)
        self._show_fixture(0, self.VEHICLE_COLOR)
        self._show_fixture(3, self.MOTOR_COLOR)
        self._show_fixture(4, self.MOTOR_COLOR)

        # Simulate spinning props by alternating show/hide
        if self.props_visible:
            for k in range(5, 9):
                self._show_fixture(k, self.PROP_COLOR)

        self.props_visible = self._on_ground(0.05) or (
            (self.props_visible + 1) % 3)

        # Pause briefly to show vehicle on ground
        if self._on_ground():
            sleep(0.5)

        return self.viewer.render(return_rgb_array=mode == 'rgb_array')

    def close(self):
        if self.viewer is not None:
            self.viewer.close()
            self.viewer = None

    def _on_ground(self, tolerance=0):
        return self.lander.position.y < (self.LANDING_POS_Y + tolerance)

    def _show_fixture(self, index, color):
        fixture = self.lander.fixtures[index]
        trans = fixture.body.transform
        path = [trans * v for v in fixture.shape.vertices]
        self.viewer.draw_polygon(path, color=color)
        path.append(path[0])
        self.viewer.draw_polyline(path, color=self.OUTLINE_COLOR, linewidth=1)
コード例 #11
0
ファイル: lander2d.py プロジェクト: GayanSamuditha/gym-copter
    def reset(self):
        self._destroy()
        self.prev_shaping = None

        self.startpos = None

        W = self.VIEWPORT_W / self.SCALE
        H = self.VIEWPORT_H / self.SCALE

        # terrain
        height = self.np_random.uniform(0,
                                        H / 2,
                                        size=(self.TERRAIN_CHUNKS + 1, ))
        chunk_x = [
            W / (self.TERRAIN_CHUNKS - 1) * i
            for i in range(self.TERRAIN_CHUNKS)
        ]
        self.helipad_x1 = chunk_x[self.TERRAIN_CHUNKS // 2 - 1]
        self.helipad_x2 = chunk_x[self.TERRAIN_CHUNKS // 2 + 1]
        self.helipad_y = H / 4
        height[self.TERRAIN_CHUNKS // 2 - 2] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 - 1] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 + 0] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 + 1] = self.helipad_y
        height[self.TERRAIN_CHUNKS // 2 + 2] = self.helipad_y
        smooth_y = [
            0.33 * (height[i - 1] + height[i + 0] + height[i + 1])
            for i in range(self.TERRAIN_CHUNKS)
        ]

        self.ground = self.world.CreateStaticBody(shapes=edgeShape(
            vertices=[(0, 0), (W, 0)]))
        self.sky_polys = []
        for i in range(self.TERRAIN_CHUNKS - 1):
            p1 = (chunk_x[i], smooth_y[i])
            p2 = (chunk_x[i + 1], smooth_y[i + 1])
            self.ground.CreateEdgeFixture(vertices=[p1, p2],
                                          density=0,
                                          friction=0.1)
            self.sky_polys.append([p1, p2, (p2[0], H), (p1[0], H)])

        self.startpos = self.VIEWPORT_W / self.SCALE / 2, self.VIEWPORT_H / self.SCALE

        self.lander = self.world.CreateDynamicBody(
            position=self.startpos,
            angle=0.0,
            fixtures=[
                fixtureDef(shape=polygonShape(vertices=[(x / self.SCALE,
                                                         y / self.SCALE)
                                                        for x, y in poly]),
                           density=0.0)
                for poly in [
                    self.HULL_POLY, self.LEG1_POLY, self.LEG2_POLY,
                    self.MOTOR1_POLY, self.MOTOR2_POLY, self.BLADE1L_POLY,
                    self.BLADE1R_POLY, self.BLADE2L_POLY, self.BLADE2R_POLY
                ]
            ])

        # By showing props periodically, we can emulate prop rotation
        self.props_visible = 0

        # Create cusom dynamics model
        self.dynamics = DJIPhantomDynamics()

        # Initialize custom dynamics with slight velocity perturbation
        state = np.zeros(12)
        d = self.dynamics
        state[d.STATE_Y] = self.startpos[
            0]  # 3D copter Y comes from 2D copter X
        state[d.STATE_Z] = -self.startpos[
            1]  # 3D copter Z comes from 2D copter Y, negated for NED
        state[d.STATE_Y_DOT] = self.INITIAL_RANDOM_VELOCITY * np.random.randn()
        state[d.STATE_Z_DOT] = self.INITIAL_RANDOM_VELOCITY * np.random.randn()
        self.dynamics.setState(state)

        return self.step(np.array([0, 0]))[0]
コード例 #12
0
ファイル: box2d.py プロジェクト: GayanSamuditha/gym-copter
class CopterBox2D(gym.Env):

    START_X = 10
    START_Y = 13

    FPS = 50
    SCALE = 30.0  # affects how fast-paced the game is, forces should be adjusted as well

    LEG_X = 12
    LEG_Y = -7
    LEG_W = 3
    LEG_H = 20

    MOTOR_X = 25
    MOTOR_Y = 7
    MOTOR_W = 4
    MOTOR_H = 5

    BLADE_X = 25
    BLADE_Y = 8
    BLADE_W = 20
    BLADE_H = 2

    BLADE1L_POLY = [
        (BLADE_X, BLADE_Y),
        (BLADE_X - BLADE_W / 2, BLADE_Y + BLADE_H),
        (BLADE_X - BLADE_W, BLADE_Y),
        (BLADE_X - BLADE_W / 2, BLADE_Y + -BLADE_H),
    ]

    BLADE1R_POLY = [
        (BLADE_X, BLADE_Y),
        (BLADE_X + BLADE_W / 2, BLADE_Y + BLADE_H),
        (BLADE_X + BLADE_W, BLADE_Y),
        (BLADE_X + BLADE_W / 2, BLADE_Y + -BLADE_H),
    ]

    BLADE2L_POLY = [
        (-BLADE_X, BLADE_Y),
        (-(BLADE_X + BLADE_W / 2), BLADE_Y + BLADE_H),
        (-(BLADE_X + BLADE_W), BLADE_Y),
        (-(BLADE_X + BLADE_W / 2), BLADE_Y + -BLADE_H),
    ]

    BLADE2R_POLY = [
        (-BLADE_X, BLADE_Y),
        (-BLADE_X + BLADE_W / 2, BLADE_Y + BLADE_H),
        (-BLADE_X + BLADE_W, BLADE_Y),
        (-BLADE_X + BLADE_W / 2, BLADE_Y + -BLADE_H),
    ]

    HULL_POLY = [
        (-30, 0),
        (-4, +4),
        (+4, +4),
        (+30, 0),
        (+4, -14),
        (-4, -14),
    ]

    LEG1_POLY = [(-LEG_X, LEG_Y), (-LEG_X + LEG_W, LEG_Y),
                 (-LEG_X + LEG_W, LEG_Y - LEG_H), (-LEG_X, LEG_Y - LEG_H)]

    LEG2_POLY = [(+LEG_X, LEG_Y), (+LEG_X + LEG_W, LEG_Y),
                 (+LEG_X + LEG_W, LEG_Y - LEG_H), (+LEG_X, LEG_Y - LEG_H)]

    MOTOR1_POLY = [(+MOTOR_X, MOTOR_Y), (+MOTOR_X + MOTOR_W, MOTOR_Y),
                   (+MOTOR_X + MOTOR_W, MOTOR_Y - MOTOR_H),
                   (+MOTOR_X, MOTOR_Y - MOTOR_H)]

    MOTOR2_POLY = [(-MOTOR_X, MOTOR_Y), (-MOTOR_X + MOTOR_W, MOTOR_Y),
                   (-MOTOR_X + MOTOR_W, MOTOR_Y - MOTOR_H),
                   (-MOTOR_X, MOTOR_Y - MOTOR_H)]

    VIEWPORT_W = 600
    VIEWPORT_H = 400

    SKY_COLOR = 0.5, 0.8, 1.0
    GROUND_COLOR = 0.5, 0.7, 0.3
    FLAG_COLOR = 0.8, 0.0, 0.0
    VEHICLE_COLOR = 1.0, 1.0, 1.0
    MOTOR_COLOR = 0.5, 0.5, 0.5
    PROP_COLOR = 0.0, 0.0, 0.0
    OUTLINE_COLOR = 0.0, 0.0, 0.0

    metadata = {
        'render.modes': ['human', 'rgb_array'],
        'video.frames_per_second': FPS
    }

    def __init__(self, observation_size, action_size):

        self.seed()
        self.viewer = None

        self.world = Box2D.b2World()
        self.ground = None
        self.lander = None

        self.prev_reward = None

        # Useful range is -1 .. +1, but spikes can be higher
        self.observation_space = spaces.Box(-np.inf,
                                            np.inf,
                                            shape=(observation_size, ),
                                            dtype=np.float32)

        # [-1,+1] will be rescaled to [0,1] for dynamics input
        self.action_space = spaces.Box(-1,
                                       +1, (action_size, ),
                                       dtype=np.float32)

        self._reset(0, 0)

    def seed(self, seed=None):
        self.np_random, seed = seeding.np_random(seed)
        return [seed]

    def _destroy(self):
        if not self.ground: return
        self.world.contactListener = None
        self.world.DestroyBody(self.ground)
        self.ground = None
        self.world.DestroyBody(self.lander)
        self.lander = None

    def _reset(self, xoff=0, yoff=0):
        self._destroy()
        self.world.contactListener_keepref = ContactDetector(self)
        self.world.contactListener = self.world.contactListener_keepref
        self.landed = False
        self.prev_shaping = None
        self.rendering = False

        W = self.VIEWPORT_W / self.SCALE
        H = self.VIEWPORT_H / self.SCALE

        # Turn off gravity so we can run our own dynamics
        self.world.gravity = 0, 0

        # terrain
        CHUNKS = 11
        height = self.np_random.uniform(0, H / 2, size=(CHUNKS + 1, ))
        chunk_x = [W / (CHUNKS - 1) * i for i in range(CHUNKS)]
        self.helipad_x1 = chunk_x[CHUNKS // 2 - 1]
        self.helipad_x2 = chunk_x[CHUNKS // 2 + 1]
        self.helipad_y = H / 4
        height[CHUNKS // 2 - 2] = self.helipad_y
        height[CHUNKS // 2 - 1] = self.helipad_y
        height[CHUNKS // 2 + 0] = self.helipad_y
        height[CHUNKS // 2 + 1] = self.helipad_y
        height[CHUNKS // 2 + 2] = self.helipad_y
        smooth_y = [
            0.33 * (height[i - 1] + height[i + 0] + height[i + 1])
            for i in range(CHUNKS)
        ]

        self.ground = self.world.CreateStaticBody(shapes=edgeShape(
            vertices=[(0, 0), (W, 0)]))
        self.sky_polys = []
        for i in range(CHUNKS - 1):
            p1 = (chunk_x[i], smooth_y[i])
            p2 = (chunk_x[i + 1], smooth_y[i + 1])
            self.ground.CreateEdgeFixture(vertices=[p1, p2],
                                          density=0,
                                          friction=0.1)
            self.sky_polys.append([p1, p2, (p2[0], H), (p1[0], H)])

        initial_y = self.VIEWPORT_H / self.SCALE

        self.lander = self.world.CreateDynamicBody(
            position=(self.VIEWPORT_W / self.SCALE / 2, initial_y),
            angle=0.0,
            fixtures=[
                fixtureDef(shape=polygonShape(vertices=[(x / self.SCALE,
                                                         y / self.SCALE)
                                                        for x, y in poly]),
                           density=1.0)
                for poly in [
                    self.HULL_POLY, self.LEG1_POLY, self.LEG2_POLY,
                    self.MOTOR1_POLY, self.MOTOR2_POLY, self.BLADE1L_POLY,
                    self.BLADE1R_POLY, self.BLADE2L_POLY, self.BLADE2R_POLY
                ]
            ])

        self.dynamics = DJIPhantomDynamics()

        # Start at top center, plus optional offset
        state = np.zeros(12)
        state[
            self.dynamics.
            STATE_Y] = self.START_X + xoff  # 3D copter Y comes from 2D copter X
        state[self.dynamics.STATE_Z] = -(
            self.START_Y + yoff)  # 3D copter Z comes from 2D copter Y

        self.dynamics.setState(state)

        # By showing props periodically, we can emulate prop rotation
        self.show_props = 0

        # Support showing vehicle while on ground
        self.ground_count = 0

        return self.step(np.array([0, 0]))[0]

    def step(self, action):

        motors = self._action_to_motors(action)

        # Set motors and compute dynamics
        self.dynamics.setMotors(motors)
        self.dynamics.update(1.0 / self.FPS)
        state = self.dynamics.getState()

        # Run one tick of Box2D simulator
        self.world.Step(1.0 / self.FPS, 6 * 30, 2 * 30)

        # Copy dynamics kinematics out to lander, negating Z for NED => ENU
        dyn = self.dynamics
        self.lander.position = state[dyn.STATE_Y], -state[dyn.STATE_Z]
        self.lander.angle = -state[dyn.STATE_PHI]
        self.lander.angularVelocity = -state[dyn.STATE_PHI_DOT]
        self.lander.linearVelocity = (state[dyn.STATE_Y_DOT],
                                      -state[dyn.STATE_Z_DOT])

        state, reward, done = self._get_state_reward_done()

        return np.array(state, dtype=np.float32), reward, done, {}

    def render(self, mode='human'):

        from gym.envs.classic_control import rendering

        # Helps with a little extra time at the end
        self.rendering = True

        if self.viewer is None:
            self.viewer = rendering.Viewer(self.VIEWPORT_W, self.VIEWPORT_H)
            self.viewer.set_bounds(0, self.VIEWPORT_W / self.SCALE, 0,
                                   self.VIEWPORT_H / self.SCALE)

        self.viewer.draw_polygon([(0, 0), (self.VIEWPORT_W, 0),
                                  (self.VIEWPORT_W, self.VIEWPORT_H),
                                  (0, self.VIEWPORT_H)],
                                 color=self.GROUND_COLOR)

        for p in self.sky_polys:
            self.viewer.draw_polygon(p, color=self.SKY_COLOR)

        self._show_fixture(1, self.VEHICLE_COLOR)
        self._show_fixture(2, self.VEHICLE_COLOR)
        self._show_fixture(0, self.VEHICLE_COLOR)
        self._show_fixture(3, self.MOTOR_COLOR)
        self._show_fixture(4, self.MOTOR_COLOR)

        # Simulate spinning props by alernating
        if self.landed or self.show_props:
            for k in range(5, 9):
                self._show_fixture(k, self.PROP_COLOR)

        for x in [self.helipad_x1, self.helipad_x2]:
            flagy1 = self.helipad_y
            flagy2 = flagy1 + 50 / self.SCALE
            self.viewer.draw_polyline([(x, flagy1), (x, flagy2)],
                                      color=(1, 1, 1))
            self.viewer.draw_polygon(
                [(x, flagy2), (x, flagy2 - 10 / self.SCALE),
                 (x + 25 / self.SCALE, flagy2 - 5 / self.SCALE)],
                color=self.FLAG_COLOR)

        self.show_props = (self.show_props + 1) % 3

        return self.viewer.render(return_rgb_array=mode == 'rgb_array')

    def close(self):
        if self.viewer is not None:
            self.viewer.close()
            self.viewer = None

    def _show_fixture(self, index, color):
        fixture = self.lander.fixtures[index]
        trans = fixture.body.transform
        path = [trans * v for v in fixture.shape.vertices]
        self.viewer.draw_polygon(path, color=color)
        path.append(path[0])
        self.viewer.draw_polyline(path, color=self.OUTLINE_COLOR, linewidth=1)
コード例 #13
0
ファイル: box2d.py プロジェクト: GayanSamuditha/gym-copter
    def _reset(self, xoff=0, yoff=0):
        self._destroy()
        self.world.contactListener_keepref = ContactDetector(self)
        self.world.contactListener = self.world.contactListener_keepref
        self.landed = False
        self.prev_shaping = None
        self.rendering = False

        W = self.VIEWPORT_W / self.SCALE
        H = self.VIEWPORT_H / self.SCALE

        # Turn off gravity so we can run our own dynamics
        self.world.gravity = 0, 0

        # terrain
        CHUNKS = 11
        height = self.np_random.uniform(0, H / 2, size=(CHUNKS + 1, ))
        chunk_x = [W / (CHUNKS - 1) * i for i in range(CHUNKS)]
        self.helipad_x1 = chunk_x[CHUNKS // 2 - 1]
        self.helipad_x2 = chunk_x[CHUNKS // 2 + 1]
        self.helipad_y = H / 4
        height[CHUNKS // 2 - 2] = self.helipad_y
        height[CHUNKS // 2 - 1] = self.helipad_y
        height[CHUNKS // 2 + 0] = self.helipad_y
        height[CHUNKS // 2 + 1] = self.helipad_y
        height[CHUNKS // 2 + 2] = self.helipad_y
        smooth_y = [
            0.33 * (height[i - 1] + height[i + 0] + height[i + 1])
            for i in range(CHUNKS)
        ]

        self.ground = self.world.CreateStaticBody(shapes=edgeShape(
            vertices=[(0, 0), (W, 0)]))
        self.sky_polys = []
        for i in range(CHUNKS - 1):
            p1 = (chunk_x[i], smooth_y[i])
            p2 = (chunk_x[i + 1], smooth_y[i + 1])
            self.ground.CreateEdgeFixture(vertices=[p1, p2],
                                          density=0,
                                          friction=0.1)
            self.sky_polys.append([p1, p2, (p2[0], H), (p1[0], H)])

        initial_y = self.VIEWPORT_H / self.SCALE

        self.lander = self.world.CreateDynamicBody(
            position=(self.VIEWPORT_W / self.SCALE / 2, initial_y),
            angle=0.0,
            fixtures=[
                fixtureDef(shape=polygonShape(vertices=[(x / self.SCALE,
                                                         y / self.SCALE)
                                                        for x, y in poly]),
                           density=1.0)
                for poly in [
                    self.HULL_POLY, self.LEG1_POLY, self.LEG2_POLY,
                    self.MOTOR1_POLY, self.MOTOR2_POLY, self.BLADE1L_POLY,
                    self.BLADE1R_POLY, self.BLADE2L_POLY, self.BLADE2R_POLY
                ]
            ])

        self.dynamics = DJIPhantomDynamics()

        # Start at top center, plus optional offset
        state = np.zeros(12)
        state[
            self.dynamics.
            STATE_Y] = self.START_X + xoff  # 3D copter Y comes from 2D copter X
        state[self.dynamics.STATE_Z] = -(
            self.START_Y + yoff)  # 3D copter Z comes from 2D copter Y

        self.dynamics.setState(state)

        # By showing props periodically, we can emulate prop rotation
        self.show_props = 0

        # Support showing vehicle while on ground
        self.ground_count = 0

        return self.step(np.array([0, 0]))[0]