You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
AirTrapMine/src/GUI/mission_planner.py

538 lines
21 KiB
Python

#!/usr/bin/env python3
"""
飛行任務規劃模組
支援: M-Formation, Circle, Leader-Follower (Bezier 轉彎), Grid Sweep
"""
import math
from typing import List, Tuple, Optional, Dict, Any
from enum import Enum
class MissionType(Enum):
"""任務類型"""
M_FORMATION = "m_formation"
CIRCLE_FORMATION = "circle_formation"
LEADER_FOLLOWER = "leader_follower"
GRID_SWEEP = "grid_sweep"
class CoordinateConverter:
"""GPS 座標與 Local 座標的轉換器"""
def __init__(self, origin_lat: float, origin_lon: float, origin_alt: float = 0):
self.origin_lat = origin_lat
self.origin_lon = origin_lon
self.origin_alt = origin_alt
self.R = 6371000.0
def gps_to_local(self, lat: float, lon: float, alt: float) -> Tuple[float, float, float]:
lat_rad = math.radians(lat)
lon_rad = math.radians(lon)
origin_lat_rad = math.radians(self.origin_lat)
origin_lon_rad = math.radians(self.origin_lon)
dlat = lat_rad - origin_lat_rad
dlon = lon_rad - origin_lon_rad
x = self.R * dlon * math.cos((lat_rad + origin_lat_rad) / 2)
y = self.R * dlat
z = alt - self.origin_alt
return x, y, z
def local_to_gps(self, x: float, y: float, z: float) -> Tuple[float, float, float]:
origin_lat_rad = math.radians(self.origin_lat)
origin_lon_rad = math.radians(self.origin_lon)
lat_rad = origin_lat_rad + (y / self.R)
lon_rad = origin_lon_rad + (x / (self.R * math.cos((lat_rad + origin_lat_rad) / 2)))
lat = math.degrees(lat_rad)
lon = math.degrees(lon_rad)
alt = self.origin_alt + z
return lat, lon, alt
class FormationPlanner:
"""隊形規劃器"""
def __init__(self, spacing: float = 5.0,
base_altitude: float = 10.0,
altitude_diff: float = 2.0):
self.spacing = spacing
self.base_altitude = base_altitude
self.altitude_diff = altitude_diff
self.current_origin = None
self.converter = None
def plan_formation_mission(self,
drone_gps_positions: List[Tuple[float, float, float]],
target_gps: Tuple[float, float, float],
mission_type: MissionType = MissionType.M_FORMATION,
params: Optional[Dict[str, Any]] = None
) -> Tuple[List[List[Tuple[float, float, float]]],
Tuple[float, float, float]]:
if len(drone_gps_positions) == 0:
raise ValueError("無人機位置列表不能為空")
center_lat = sum(pos[0] for pos in drone_gps_positions) / len(drone_gps_positions)
center_lon = sum(pos[1] for pos in drone_gps_positions) / len(drone_gps_positions)
center_alt = sum(pos[2] for pos in drone_gps_positions) / len(drone_gps_positions)
self.current_origin = (center_lat, center_lon, center_alt)
self.converter = CoordinateConverter(center_lat, center_lon, center_alt)
drone_local = [self.converter.gps_to_local(*pos) for pos in drone_gps_positions]
target_local = self.converter.gps_to_local(*target_gps)
if mission_type == MissionType.M_FORMATION:
s1, s2 = self._calculate_m_formation(drone_local, target_local, params)
waypoints_local = [[s1[i], s2[i]] for i in range(len(drone_local))]
elif mission_type == MissionType.CIRCLE_FORMATION:
s1, s2 = self._calculate_circle_formation(drone_local, target_local, params)
waypoints_local = [[s1[i], s2[i]] for i in range(len(drone_local))]
elif mission_type == MissionType.LEADER_FOLLOWER:
params = params or {}
route_wps_gps = params.get('route_waypoints')
if route_wps_gps is None or len(route_wps_gps) < 2:
raise ValueError("LEADER_FOLLOWER 需要至少 2 個路徑點")
route_wps_local = [
self.converter.gps_to_local(wp[0], wp[1], 0)[:2]
for wp in route_wps_gps
]
waypoints_local = self._calculate_leader_follower(drone_local, route_wps_local, params)
elif mission_type == MissionType.GRID_SWEEP:
params = params or {}
rect_corners_gps = params.get('rect_corners')
if rect_corners_gps is None or len(rect_corners_gps) != 4:
raise ValueError("GRID_SWEEP 需要 4 個 GPS 角點")
rect_corners_local = [
self.converter.gps_to_local(c[0], c[1], 0)[:2]
for c in rect_corners_gps
]
waypoints_local = self._calculate_grid_sweep(drone_local, rect_corners_local, params)
else:
raise ValueError(f"不支援的任務類型: {mission_type}")
waypoints_gps = []
for drone_wps in waypoints_local:
gps_wps = [self.converter.local_to_gps(*wp) for wp in drone_wps]
waypoints_gps.append(gps_wps)
return waypoints_gps, self.current_origin
# ------------------------------------------------------------------ M-Formation
def _calculate_m_formation(self, drone_positions, target_point, params):
params = params or {}
N = len(drone_positions)
spacing = params.get('spacing', self.spacing)
base_altitude = params.get('base_altitude', self.base_altitude)
altitude_diff = params.get('altitude_diff', self.altitude_diff)
C_x = sum(pos[0] for pos in drone_positions) / N
C_y = sum(pos[1] for pos in drone_positions) / N
T_x, T_y, T_z = target_point
V_x, V_y = T_x - C_x, T_y - C_y
P_x, P_y = -V_y, V_x
length = math.sqrt(P_x ** 2 + P_y ** 2)
P_x_unit, P_y_unit = (P_x / length, P_y / length) if length > 0.01 else (1.0, 0.0)
projections = [((pos[0] - C_x) * P_x_unit + (pos[1] - C_y) * P_y_unit, i)
for i, pos in enumerate(drone_positions)]
projections.sort()
stage1_positions = [None] * N
stage2_positions = [None] * N
for rank, (_, original_idx) in enumerate(projections):
offset = (rank - (N - 1) / 2) * spacing
altitude = base_altitude + (altitude_diff if rank % 2 == 0 else -altitude_diff)
stage1_positions[original_idx] = (C_x + P_x_unit * offset, C_y + P_y_unit * offset, altitude)
stage2_positions[original_idx] = (T_x + P_x_unit * offset, T_y + P_y_unit * offset, altitude)
return stage1_positions, stage2_positions
# ------------------------------------------------------------------ Circle
def _calculate_circle_formation(self, drone_positions, target_point, params):
params = params or {}
N = len(drone_positions)
radius = params.get('radius', 10.0)
altitude = params.get('altitude', 10.0)
start_angle = params.get('start_angle', 0.0)
center_x, center_y, center_z = target_point
stage1_positions = []
stage2_positions = []
angle_step = 360.0 / N
for i in range(N):
angle_rad = math.radians(start_angle + angle_step * i)
final_x = center_x + radius * math.cos(angle_rad)
final_y = center_y + radius * math.sin(angle_rad)
final_z = altitude
current_x, current_y, current_z = drone_positions[i]
stage1_positions.append((
current_x + (final_x - current_x) * 0.5,
current_y + (final_y - current_y) * 0.5,
current_z + (final_z - current_z) * 0.5
))
stage2_positions.append((final_x, final_y, final_z))
return stage1_positions, stage2_positions
# ------------------------------------------------------------------ 路徑跟隨 (Bezier 轉彎)
def _calculate_leader_follower(self, drone_positions, route_wps_local, params):
"""
路徑跟隨編隊 Bezier 曲線轉彎版
步驟:
1. _build_center_path: 在轉折 WP 處用二次 Bezier 曲線平滑轉彎
2. 固定排序每架無人機沿中心路徑套用橫向+縱向偏移
隊形 (俯視):
| (前進方向) |
| D1 | 左偏移, 0m
| D2 | 右偏移, 5m
| D3 | 左偏移, 10m
| D4 | 右偏移, 15m
"""
N = len(drone_positions)
lateral_offset = params.get('lateral_offset', 3.0)
longitudinal_spacing = params.get('longitudinal_spacing', 5.0)
altitude = params.get('altitude', self.base_altitude)
turn_margin = params.get('turn_margin', 0.35) # 轉彎切入距離佔段長比例
curve_resolution = params.get('curve_resolution', 8) # 每個彎道的插值點數
# Step 1: 建立帶 Bezier 轉彎的中心路徑
center_path = self._build_center_path(
route_wps_local, turn_margin, curve_resolution
)
# Step 2: 固定排序
first_dir = (center_path[0][2], center_path[0][3])
first_perp = (-first_dir[1], first_dir[0])
C_x = sum(p[0] for p in drone_positions) / N
C_y = sum(p[1] for p in drone_positions) / N
projections = [
((pos[0] - C_x) * first_perp[0] + (pos[1] - C_y) * first_perp[1], i)
for i, pos in enumerate(drone_positions)
]
projections.sort()
# Step 3: 偏移
all_waypoints = [None] * N
for rank, (_, original_idx) in enumerate(projections):
lat_sign = -1 if rank % 2 == 0 else 1
lat = lat_sign * lateral_offset
lon = rank * longitudinal_spacing
waypoints = []
for (cx, cy, dx, dy) in center_path:
perp_x, perp_y = -dy, dx
ox = cx + lat * perp_x - lon * dx
oy = cy + lat * perp_y - lon * dy
waypoints.append((ox, oy, altitude))
all_waypoints[original_idx] = waypoints
return all_waypoints
def _build_center_path(self, waypoints, turn_margin, curve_resolution):
"""
建立帶 Bezier 曲線轉彎的中心路徑
在每個轉折 WP :
1. 計算 pre_turn = WP - d_in × T
2. 計算 post_turn = WP + d_out × T
3. 用二次 Bezier 曲線: B(t) = (1-t)²·pre + 2t(1-t)·WP + ·post
4. 方向從導數得到: B'(t) = 2(1-t)(WP-pre) + 2t(post-WP)
Args:
waypoints: 路徑航點 [(x, y), ...]
turn_margin: 轉彎切入距離佔相鄰段長的比例 (0~0.5)
curve_resolution: 每個彎道的 Bezier 插值點數
Returns:
[(x, y, dir_x, dir_y), ...] 中心路徑點 + 單位方向
"""
num_wps = len(waypoints)
if num_wps < 2:
return [(waypoints[0][0], waypoints[0][1], 0.0, 1.0)]
# 計算每段方向和長度
segments = []
for i in range(num_wps - 1):
dx = waypoints[i + 1][0] - waypoints[i][0]
dy = waypoints[i + 1][1] - waypoints[i][1]
length = math.sqrt(dx ** 2 + dy ** 2)
if length < 0.01:
segments.append((0.0, 1.0, length))
else:
segments.append((dx / length, dy / length, length))
path = []
# 第一個航點
path.append((waypoints[0][0], waypoints[0][1],
segments[0][0], segments[0][1]))
# 中間航點 (轉彎處)
for i in range(1, num_wps - 1):
d_in_x, d_in_y, len_in = segments[i - 1]
d_out_x, d_out_y, len_out = segments[i]
# 切入距離 T: 取相鄰兩段中較短的 × turn_margin
T = min(len_in, len_out) * turn_margin
if T < 0.5:
# 段太短,不插彎,直接加一個平均方向點
avg_dx = d_in_x + d_out_x
avg_dy = d_in_y + d_out_y
avg_len = math.sqrt(avg_dx ** 2 + avg_dy ** 2)
if avg_len > 0.01:
avg_dx /= avg_len
avg_dy /= avg_len
else:
avg_dx, avg_dy = d_in_x, d_in_y
path.append((waypoints[i][0], waypoints[i][1], avg_dx, avg_dy))
continue
# P0 = pre_turn (弧線起始)
p0_x = waypoints[i][0] - d_in_x * T
p0_y = waypoints[i][1] - d_in_y * T
# P1 = WP 本身 (控制點)
p1_x = waypoints[i][0]
p1_y = waypoints[i][1]
# P2 = post_turn (弧線結束)
p2_x = waypoints[i][0] + d_out_x * T
p2_y = waypoints[i][1] + d_out_y * T
# 加入 pre_turn 點 (方向 = incoming)
path.append((p0_x, p0_y, d_in_x, d_in_y))
# 加入 Bezier 插值點
for j in range(1, curve_resolution):
t = j / curve_resolution
# B(t) = (1-t)²·P0 + 2t(1-t)·P1 + t²·P2
one_minus_t = 1.0 - t
bx = one_minus_t * one_minus_t * p0_x + \
2 * t * one_minus_t * p1_x + \
t * t * p2_x
by = one_minus_t * one_minus_t * p0_y + \
2 * t * one_minus_t * p1_y + \
t * t * p2_y
# B'(t) = 2(1-t)(P1-P0) + 2t(P2-P1) → 切線方向
tdx = 2 * one_minus_t * (p1_x - p0_x) + 2 * t * (p2_x - p1_x)
tdy = 2 * one_minus_t * (p1_y - p0_y) + 2 * t * (p2_y - p1_y)
# 正規化
t_len = math.sqrt(tdx ** 2 + tdy ** 2)
if t_len > 0.01:
tdx /= t_len
tdy /= t_len
else:
tdx, tdy = d_in_x, d_in_y
path.append((bx, by, tdx, tdy))
# 加入 post_turn 點 (方向 = outgoing)
path.append((p2_x, p2_y, d_out_x, d_out_y))
# 最後一個航點
path.append((waypoints[-1][0], waypoints[-1][1],
segments[-1][0], segments[-1][1]))
return path
# ------------------------------------------------------------------ 柵狀掃描
def _calculate_grid_sweep(self, drone_positions, rect_corners_local, params):
"""柵狀掃描:掃描方向沿矩形長邊,切割方向沿短邊"""
N = len(drone_positions)
line_spacing = params.get('line_spacing', 5.0)
altitude = params.get('altitude', self.base_altitude)
c0, c1, c2, c3 = rect_corners_local
edge_a = (c1[0] - c0[0], c1[1] - c0[1])
len_a = math.sqrt(edge_a[0] ** 2 + edge_a[1] ** 2)
edge_b = (c3[0] - c0[0], c3[1] - c0[1])
len_b = math.sqrt(edge_b[0] ** 2 + edge_b[1] ** 2)
if len_a >= len_b:
sweep_vec = edge_a
sweep_len = len_a
cut_vec = edge_b
cut_len = len_b
sweep_start_mid = ((c0[0] + c3[0]) / 2, (c0[1] + c3[1]) / 2)
sweep_end_mid = ((c1[0] + c2[0]) / 2, (c1[1] + c2[1]) / 2)
cut_start_corner = c0
else:
sweep_vec = edge_b
sweep_len = len_b
cut_vec = edge_a
cut_len = len_a
sweep_start_mid = ((c0[0] + c1[0]) / 2, (c0[1] + c1[1]) / 2)
sweep_end_mid = ((c3[0] + c2[0]) / 2, (c3[1] + c2[1]) / 2)
cut_start_corner = c0
sweep_dir = (sweep_vec[0] / sweep_len, sweep_vec[1] / sweep_len)
cut_dir = (cut_vec[0] / cut_len, cut_vec[1] / cut_len)
C_x = sum(p[0] for p in drone_positions) / N
C_y = sum(p[1] for p in drone_positions) / N
dist_to_start = math.sqrt(
(C_x - sweep_start_mid[0]) ** 2 + (C_y - sweep_start_mid[1]) ** 2
)
dist_to_end = math.sqrt(
(C_x - sweep_end_mid[0]) ** 2 + (C_y - sweep_end_mid[1]) ** 2
)
if dist_to_start <= dist_to_end:
near_corner_a = cut_start_corner
else:
sweep_dir = (-sweep_dir[0], -sweep_dir[1])
near_corner_a = (cut_start_corner[0] + sweep_vec[0],
cut_start_corner[1] + sweep_vec[1])
projections = [
((pos[0] - C_x) * cut_dir[0] + (pos[1] - C_y) * cut_dir[1], i)
for i, pos in enumerate(drone_positions)
]
projections.sort()
strip_width = cut_len / N
drone_sweep_proj = C_x * sweep_dir[0] + C_y * sweep_dir[1]
near_sweep_proj = near_corner_a[0] * sweep_dir[0] + near_corner_a[1] * sweep_dir[1]
gather_sweep_proj = (drone_sweep_proj + near_sweep_proj) / 2
all_waypoints = [None] * N
for rank, (_, original_idx) in enumerate(projections):
strip_center_offset = (rank + 0.5) * strip_width
base_x = near_corner_a[0] + cut_dir[0] * strip_center_offset
base_y = near_corner_a[1] + cut_dir[1] * strip_center_offset
waypoints_list = []
gather_offset = gather_sweep_proj - near_sweep_proj
gx = base_x + sweep_dir[0] * gather_offset
gy = base_y + sweep_dir[1] * gather_offset
waypoints_list.append((gx, gy, altitude))
num_lines = max(1, int(strip_width / line_spacing))
actual_spacing = strip_width / num_lines
first_line_offset = (rank * strip_width) + actual_spacing / 2
entry_x = near_corner_a[0] + cut_dir[0] * first_line_offset
entry_y = near_corner_a[1] + cut_dir[1] * first_line_offset
waypoints_list.append((entry_x, entry_y, altitude))
current_cut_offset = first_line_offset
for line_idx in range(num_lines):
line_near_x = near_corner_a[0] + cut_dir[0] * current_cut_offset
line_near_y = near_corner_a[1] + cut_dir[1] * current_cut_offset
line_far_x = line_near_x + sweep_dir[0] * sweep_len
line_far_y = line_near_y + sweep_dir[1] * sweep_len
if line_idx % 2 == 0:
waypoints_list.append((line_far_x, line_far_y, altitude))
else:
waypoints_list.append((line_near_x, line_near_y, altitude))
if line_idx < num_lines - 1:
next_cut_offset = current_cut_offset + actual_spacing
next_near_x = near_corner_a[0] + cut_dir[0] * next_cut_offset
next_near_y = near_corner_a[1] + cut_dir[1] * next_cut_offset
next_far_x = next_near_x + sweep_dir[0] * sweep_len
next_far_y = next_near_y + sweep_dir[1] * sweep_len
if line_idx % 2 == 0:
waypoints_list.append((next_far_x, next_far_y, altitude))
else:
waypoints_list.append((next_near_x, next_near_y, altitude))
current_cut_offset = next_cut_offset
all_waypoints[original_idx] = waypoints_list
return all_waypoints
# ================================================================================
# 測試
# ================================================================================
if __name__ == "__main__":
planner = FormationPlanner()
drones = [
(24.123450, 120.567800, 0.0),
(24.123470, 120.567820, 0.0),
(24.123440, 120.567810, 0.0),
(24.123460, 120.567830, 0.0),
]
target = (24.12400, 120.56795, 10.0)
# M-Formation
wps, origin = planner.plan_formation_mission(drones, target, MissionType.M_FORMATION)
print("M-Formation:")
for i, wp_list in enumerate(wps):
print(f" Drone {i}: {len(wp_list)} waypoints")
# Leader-Follower (Bezier 轉彎)
route = [
(24.12345, 120.56780),
(24.12370, 120.56800),
(24.12390, 120.56820),
(24.12400, 120.56800),
(24.12420, 120.56790),
]
wps_lf, origin_lf = planner.plan_formation_mission(
drones, target, MissionType.LEADER_FOLLOWER,
params={
'route_waypoints': route,
'lateral_offset': 3.0,
'longitudinal_spacing': 5.0,
'turn_margin': 0.35,
'curve_resolution': 8,
'altitude': 10.0
}
)
print(f"\nLeader-Follower (Bezier turns):")
for i, wp_list in enumerate(wps_lf):
print(f" Drone {i}: {len(wp_list)} waypoints")
for j, wp in enumerate(wp_list):
print(f" WP{j}: ({wp[0]:.6f}, {wp[1]:.6f}, {wp[2]:.1f})")
# Grid Sweep
rect = [
(24.1237, 120.5679),
(24.1237, 120.5683),
(24.1240, 120.5683),
(24.1240, 120.5679)
]
wps2, origin2 = planner.plan_formation_mission(
drones, target, MissionType.GRID_SWEEP,
params={'rect_corners': rect, 'line_spacing': 5.0, 'altitude': 10.0}
)
print(f"\nGrid Sweep:")
for i, wp_list in enumerate(wps2):
print(f" Drone {i}: {len(wp_list)} waypoints")