OpenFlight/navp2p_v2_test_multi.py at main · running-mars/OpenFlight · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
"""
Description:
    - Actuator-level control for Fixed-wing UAVs towards minimum-time flight
    - Validate the performance with multiple targets

Authors:
    - Running-Mars
"""

import numpy as np
import pandas as pd
import time

from lib.jsbsimlib import JSBSimIO, X8Autopilot
from lib.jsbsim_aircraft import Aircraft, x8
from lib import jsbsim_properties as prp

from stable_baselines3 import PPO, SAC, DDPG, TD3, A2C


class SimLoop:
    def __init__(self,
                 sim_time: float = 1000.,
                 airspeed: float = 30.0,
                 sim_frequency_hz: float = 30.0,
                 aircraft: Aircraft = x8,
                 init_conditions: bool = None,
                 debug_level: int = 0):
        self.sim_time = sim_time
        self.airspeed = airspeed
        self.sim_frequency_hz = sim_frequency_hz
        self.aircraft = aircraft

        self.simIO: JSBSimIO = JSBSimIO(sim_frequency_hz, aircraft, init_conditions, debug_level)
        self.ap: X8Autopilot = X8Autopilot(self.simIO)

        self.state = {
            "goal": np.zeros(3),
            "vel": np.zeros(2),
            "attitude": np.zeros(3),
            "ang_vel": np.zeros(3),
            "collision": False,
        }
        self.goal_num = 64
        self.goal_position_all = np.zeros((self.goal_num, 3))

        for index in range(self.goal_num):
            distance_horiz = 500.
            angle_horiz = 2 * np.pi / 64 * 0.5 + index % 64 * 2 * np.pi / 64 + np.pi
            self.goal_position_all[index, 0] = distance_horiz * np.cos(angle_horiz)
            self.goal_position_all[index, 1] = distance_horiz * np.sin(angle_horiz)
            self.goal_position_all[index, 2] = 20. + (180. - 20.) * index / self.goal_num

        self.reach_radius = 30.                     # when the distance to the goal is less than reach_radius, say arrived!
        self.max_flight_time = 50.

        self.init_conditions = {prp.pitch_rad: 0.,
                                prp.roll_rad: 0.,
                                prp.heading_rad: 0.,
                                prp.lat_geod_deg: 0.,
                                prp.lng_geoc_deg: 0.,
                                prp.altitude_sl_ft: 100. * 3.28,
                                prp.u_fps: 52.48,       # 16.0 m/s
                                prp.v_fps: 0.,
                                prp.w_fps: 0.,
                                prp.p_radps: 0.,
                                prp.q_radps: 0.,
                                prp.r_radps: 0.}

        self.episode_init_time = None
        self.data_path = None

    def _setup_flight(self, index):
        self.simIO.reinitialise(self.init_conditions)
        self.episode_init_time = time.strftime("%Y%m%d-%H%M%S")
        self.data_path = './flight_data_' + str(index).zfill(3) + '.csv'
        df = pd.DataFrame(columns=['sim_time(s)',
                                   'latitude(m)', 'longitude(m)', 'altitude(m)',
                                   'pitch(rad)', 'roll(rad)', 'yaw(rad)',
                                   'velocity_north(m/s)', 'velocity_east(m/s)', 'velocity_down(m/s)',
                                   'velocity_body_front(m/s)', 'velocity_body_right(m/s)', 'velocity_body_down(m/s)',
                                   'roll_rate(rad/s)', 'pitch_rate(rad/s)', 'yaw_rate(rad/s)',
                                   'airspeed(m/s)', 'altitude_rate(m/s)'])
        df.to_csv(path_or_buf=self.data_path, mode='w', index=False)

    def navigate(self, radius=300., altitude=100.):
        for index in range(self.goal_position_all.shape[0]):
            self._setup_flight(index=index)
            self.goal_position = self.goal_position_all[index, :]

            while True:
                sim_time_s = self.simIO.get_time()
                if sim_time_s > self.max_flight_time:
                    break
                position = self.simIO.get_local_position_m()  # return [latitude, longitude, altitude] meter
                orientation = self.simIO.get_local_orientation()  # return [pitch, roll, yaw] rad
                orientation[2] = self.simIO.clip2pi(orientation[2])  # from [0, 2*pi] -> [-pi, pi] rad
                v_ned_mps = self.simIO.get_velocity_mps()  # [v_n, v_e, v_d] meter/s
                v_body_mps = self.simIO.get_velocity_body_mps()
                v_ang_radps = self.simIO.get_angular_velocity_radps()  # roll pitch yaw rate in [-2 * pi, 2 * pi]
                airspeed = self.simIO.get_airspeed_mps()
                altitude_rate = self.simIO.get_altitude_rate_mps()

                goal_yaw = np.arctan2(self.goal_position[1] - position[1], self.goal_position[0] - position[0])
                goal_horiz = np.sqrt(
                    (self.goal_position[1] - position[1]) ** 2 + (self.goal_position[0] - position[0]) ** 2)
                goal_vert = self.goal_position[2] - position[2]

                self.state["goal"][0] = self.simIO.angle_minus(minuend=goal_yaw, subtrahend=orientation[2]) / (
                            2 * np.pi)
                self.state["goal"][1] = np.tanh(goal_horiz * 0.01)  # normalize (especially when d < 200m)
                self.state["goal"][2] = np.tanh(goal_vert * 0.02)  # normalize (especially when d < 100m)

                self.state["vel"][0] = np.tanh((airspeed - 20.) * 0.05)
                self.state["vel"][1] = np.tanh(v_ned_mps[2] * 0.1)

                self.state["attitude"][0] = orientation[0] / (2 * np.pi)
                self.state["attitude"][1] = orientation[1] / (2 * np.pi)
                self.state["attitude"][2] = orientation[2] / (2 * np.pi)

                self.state["ang_vel"][0] = v_ang_radps[0] / (4 * np.pi)
                self.state["ang_vel"][1] = v_ang_radps[1] / (4 * np.pi)
                self.state["ang_vel"][2] = v_ang_radps[2] / (4 * np.pi)

                obs = np.zeros(11, )
                obs[0:3] = self.state["goal"]
                obs[3:5] = self.state["vel"]
                obs[5:8] = self.state["attitude"]
                obs[8:11] = self.state["ang_vel"]

                action, _ = model.predict(obs, deterministic=True)

                # control with Api
                self.simIO[prp.aileron_cmd] = action[0]
                self.simIO[prp.elevator_cmd] = action[1]
                self.simIO[prp.throttle_cmd] = action[2] * 0.5 + 0.5

                # run the Flight Dynamics Model
                self.simIO.run_fdm()

                # flight data recording
                flight_data = {'sim_time(s)': sim_time_s,
                               'latitude(m)': [position[0]],
                               'longitude(m)': [position[1]],
                               'altitude(m)': [position[2]],
                               'pitch(rad)': [orientation[0]],
                               'roll(rad)': [orientation[1]],
                               'yaw(rad)': [orientation[2]],
                               'velocity_north(m/s)': [v_ned_mps[0]],
                               'velocity_east(m/s)': [v_ned_mps[1]],
                               'velocity_down(m/s)': [v_ned_mps[2]],
                               'velocity_body_front(m/s)': [v_body_mps[0]],
                               'velocity_body_right(m/s)': [v_body_mps[1]],
                               'velocity_body_down(m/s)': [v_body_mps[2]],
                               'roll_rate(rad/s)': [v_ang_radps[0]],
                               'pitch_rate(rad/s)': [v_ang_radps[1]],
                               'yaw_rate(rad/s)': [v_ang_radps[2]],
                               'airspeed(m/s)': airspeed,
                               'altitude_rate(m/s)': altitude_rate}
                df = pd.DataFrame(flight_data)
                df.to_csv(path_or_buf=self.data_path, mode='a', index=False, header=False)

                goal_distance = np.sqrt((self.simIO.get_local_position_m()[0] - self.goal_position[0]) ** 2 +
                                        (self.simIO.get_local_position_m()[1] - self.goal_position[1]) ** 2 +
                                        (self.simIO.get_local_position_m()[2] - self.goal_position[2]) ** 2)

                if goal_distance <= self.reach_radius:
                    print("Arrived!")
                    break

            print("* Flight time : %f" % (self.simIO.get_time()))


if __name__ == '__main__':
    model_path = "./nav_p2p_policy_20250304-163217.zip"
    model = A2C.load(model_path)

    simLoop = SimLoop()
    simLoop.navigate()