Exemplo n.º 1
0
class CurvedPathStaticDynamicObstacles(BaseShipScenario):
    def generate(self, rng):
        L = 400
        a = 2 * np.pi * (rng.rand() - 0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L * cos(a), L * sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = Vessel2D(angle, x, y)

        straightpath = ParamLine(startpoint=self.path(0).flatten(), endpoint=self.path.get_endpoint().flatten())

        for i in range(rng.randint(5, 20)):
            self.static_obstacles.append(StaticObstacle(
                self.path(0.9*self.path.length*(rng.rand() + 0.1)).flatten() + 100*(rng.rand(2)-0.5), radius=10*(rng.rand()+0.5) ))

        for i in range(rng.randint(-10, 3)):
            init_s = 0.9 * self.path.length * (rng.rand() + 0.1)
            speed = (rng.rand()+1/6)*6
            p = rng.choice([self.path, reversed(self.path), straightpath, reversed(straightpath)])
            self.dynamic_obstacles.append(DynamicObstacle(path=p, speed=speed, init_s=init_s))
Exemplo n.º 2
0
    def generate(self, rng):
        L = 400
        a = 2*np.pi*(rng.rand()-0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L*cos(a), L*sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = Vessel2D(angle, x, y, linearising_feedback=self.linFB)
Exemplo n.º 3
0
    def generate(self, rng):
        L = 400
        a = 2*np.pi*(rng.rand()-0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L*cos(a), L*sin(a))))
        self.speed = 1.2

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = AUV2D(angle, x, y)
Exemplo n.º 4
0
    def generate(self, rng):
        L = 400
        a = 2*np.pi*(rng.rand()-0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L*cos(a), L*sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)

        self.dynamic_obstacles.append(DynamicObstacle(self.path, speed=2, init_s=20))
        self.ship = Vessel2D(angle, x, y)
Exemplo n.º 5
0
    def generate(self, rng):
        L = 400
        a = 2 * np.pi * (rng.rand() - 0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L * cos(a), L * sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = AUV2D(angle, x, y, linearising_feedback=False)

        for i in range(10):
            self.static_obstacles.append(StaticObstacle(
                self.path(0.9*self.path.length*(rng.rand() + 0.1)).flatten() + 100*(rng.rand(2)-0.5), radius=10*(rng.rand()+0.5) ))
Exemplo n.º 6
0
class CurvedPathShipCollisionScenario(BaseShipScenario):
    def generate(self, rng):
        L = 400
        a = 2*np.pi*(rng.rand()-0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L*cos(a), L*sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = Vessel2D(angle, x, y)

        self.dynamic_obstacles.append(DynamicObstacle(reversed(self.path), speed=2, init_s=20))
Exemplo n.º 7
0
class CurvedPathScenario(BaseShipScenario):
    def __init__(self, linearising_feedback=True):
        self.linFB = linearising_feedback
        super().__init__()

    def generate(self, rng):
        L = 400
        a = 2*np.pi*(rng.rand()-0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L*cos(a), L*sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = Vessel2D(angle, x, y, linearising_feedback=self.linFB)
Exemplo n.º 8
0
class CurvedPathStaticObstaclesAUV(BaseShipScenario):
    def generate(self, rng):
        L = 400
        a = 2 * np.pi * (rng.rand() - 0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L * cos(a), L * sin(a))))
        self.speed = 4

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = AUV2D(angle, x, y, linearising_feedback=False)

        for i in range(10):
            self.static_obstacles.append(StaticObstacle(
                self.path(0.9*self.path.length*(rng.rand() + 0.1)).flatten() + 100*(rng.rand(2)-0.5), radius=10*(rng.rand()+0.5) ))

    def step_reward(self, action, obs, ds):
        done = False
        x, y = self.ship.position
        step_reward = 0

        # Living penalty
        # step_reward -= 0.001  # TODO Increase living penalty

        if not done and self.reward < -250:
            done = True

        if not done and abs(self.s - self.path.length) < 1:
            done = True

        for o in self.static_obstacles + self.dynamic_obstacles:
            if not done and distance(self.ship.position, o.position) < self.ship.radius + o.radius:
                done = True
                step_reward -= 50
                break

        if not done and distance(self.ship.position, self.path.get_endpoint()) < 20:
            done = True
            # step_reward += 50

        if not done:  # Reward progress along path, penalise backwards progress
            step_reward += 2*ds

        if not done:  # Penalise cross track error if too far away from path
            # state_error = obs[:6]
            # step_reward += (0.2 - np.clip(np.linalg.norm(state_error), 0, 0.4))/100
            # heading_err = state_error[2]
            # surge_err = state_error[3]
            # TODO Punish for facing wrong way / Reward for advancing along path

            for i, slot in self.active_static.items():
                closeness = obs[4 + 2*slot + 1]
                step_reward -= 0.2*closeness**2

            surge_error = obs[0]
            cross_track_error = obs[2]

            # step_reward -= abs(cross_track_error)*0.1
            # step_reward -= max(0, -surge_error)*0.1

            step_reward -= abs(cross_track_error)*0.5 + max(0, -surge_error)*0.5

            # step_reward -= (max(0.1, -obs[0]) - 0.1)*0.3
            # dist_from_path = np.sqrt(x_err ** 2 + y_err ** 2)
            # path_angle = self.path.get_angle(self.s)
            # If the reference is pointing towards the path, don't penalise
            # if dist_from_path > 0.25 and sign(float(Angle(path_angle - self.ship.ref[1]))) == sign(y_err):
            #     step_reward -= 0.1*(dist_from_path - 0.25)

        return done, step_reward
Exemplo n.º 9
0
class CurvedPathScenarioAUV(BaseShipScenario):
    def __init__(self):
        super().__init__()

    def generate(self, rng):
        L = 400
        a = 2*np.pi*(rng.rand()-0.5)
        self.path = RandomCurveThroughOrigin(rng, start=((L*cos(a), L*sin(a))))
        self.speed = 1.2

        x, y = self.path(0)
        angle = self.path.get_angle(0)
        x += 2*(rng.rand()-0.5)
        y += 2*(rng.rand()-0.5)
        angle += 0.1*(rng.rand()-0.5)
        self.ship = AUV2D(angle, x, y)
    
    def step_reward(self, action, obs, ds):
        done = False
        x, y = self.ship.position
        step_reward = 0

        # Living penalty
        # step_reward -= 0.001  # TODO Increase living penalty

        if not done and self.reward < -50:
            done = True

        if not done and abs(self.s - self.path.length) < 1:
            done = True

        if not done and distance(self.ship.position, self.path.get_endpoint()) < 20:
            done = True
            # step_reward += 50

        if not done:  # Reward progress along path, penalise backwards progress
            step_reward += ds/2


        if not done:  # Penalise cross track error if too far away from path
            # state_error = obs[:6]
            # step_reward += (0.2 - np.clip(np.linalg.norm(state_error), 0, 0.4))/100
            # heading_err = state_error[2]
            # surge_err = state_error[3]
            # TODO Punish for facing wrong way / Reward for advancing along path

            surge_error = obs[0]
            cross_track_error = obs[2]

            # step_reward -= abs(cross_track_error)*0.1
            # step_reward -= max(0, -surge_error)*0.1

            step_reward -= abs(cross_track_error)*0.5 + max(0, -surge_error)*0.5

            # step_reward -= (max(0.1, -obs[0]) - 0.1)*0.3
            # dist_from_path = np.sqrt(x_err ** 2 + y_err ** 2)
            # path_angle = self.path.get_angle(self.s)
            # If the reference is pointing towards the path, don't penalise
            # if dist_from_path > 0.25 and sign(float(Angle(path_angle - self.ship.ref[1]))) == sign(y_err):
            #     step_reward -= 0.1*(dist_from_path - 0.25)
        return done, step_reward