TypeError:不支持+=：“float”和“function”的操作数类型

Question

我只是想用人工势场（APF）来模拟一个简单的路径规划问题。为此，我有一个Point类，每个类的实例表示代理的位置。实现了一系列的函数来移动代理的位置，找到代理到其他点的距离，并计算相应的APF。还定义了一个Obstacle类，每个实例只是障碍物半径之外的一个点。你知道吗

import autograd.numpy as np
from autograd import grad
import math

class Point(object):

    def __init__(self,x,y):
        self.X = x
        self.Y = y

    def distance(self, goal):
        dx = self.X - goal.X
        dy = self.Y - goal.Y
        return math.hypot(dx, dy)

    def move(self, value):
        self.X += value
        self.Y += value

    def APF(self, goal, k, obstacles):
        num = self.distance(goal)**2
        temp = self.distance(goal)**(2*k)
        for item in obstacles:
            temp = temp + (self.distance(item.point)**2 - item.radius**2)
        den = temp**(1/k)
        return num/den

class Obstacle(object):
    def __init__(self, point, radius):
        self.point = point
        self.radius = radius

要测试上面的代码，需要考虑以下主要代码段。特别地，它应该迭代地计算APF的梯度，并将结果添加到代理的位置（例如，移动它），直到梯度消失。你知道吗

if __name__== "__main__" :
    start = Point(0.0, 0.0)
    goal = Point(15.0, 5.0)

    p1 = Point(2.0, 2.0)
    p2 = Point(3.0, 3.0)
    p3 = Point(4.0, 4.0)
    p4 = Point(5.0, 5.0)
    obs1 = Obstacle(p1, 1.0)
    obs2 = Obstacle(p2, 2.0)
    obs3 = Obstacle(p3, 3.0)
    obs4 = Obstacle(p4, 4.0)

    obstacles = [obs1, obs2, obs3, obs4]
    trajectory = [start]

    k = 7.0 
    temp = start

    while grad(temp.APF(goal, k, obstacles)) != 0.0:
        next = temp.move(grad(temp.APF(goal, k, obstacles)))
        trajctory.append(next)
        temp = next

但是，我遇到了以下错误：

Traceback (most recent call last):
  File "test_APF.py", line 53, in <module>
    next = temp.move(grad(temp.APF(goal, k, obstacles)))
  File "test_APF.py", line 17, in move
    self.X += value
TypeError: unsupported operand type(s) for +=: 'float' and 'function'

对我来说APF返回数字，所以grad应该可以接受。然后，move也将接收数字作为它的参数。你能帮我发现这个问题吗？你知道吗

Answer 1

根据the Autograd readme，autograd.grad()返回一个函数。你知道吗

>>> grad_tanh = grad(tanh)       # Obtain its gradient function
>>> grad_tanh(1.0)               # Evaluate the gradient at x = 1.0

将函数而不是其返回值传递给.move()函数。你知道吗

Answer 2

import autograd.numpy as np
from autograd import grad
import math



def APF(k,start,goal, obstacles):
    dist = math.hypot(start[0] - goal[0], start[1] - goal[1])
    num = dist**2
    temp = dist**(2*k)
    for item in obstacles:
        temp = temp + (dist ** 2 - item.radius ** 2)
    den = temp ** (1 / k)
    return num / den

class Point(object):

    def __init__(self,x,y):
        self.X = x
        self.Y = y

    def distance(self, goal):
        dx = self.X - goal.X
        dy = self.Y - goal.Y
        return math.hypot(dx, dy)

    def move(self, value):
        self.X += value
        self.Y += value



class Obstacle(object):
    def __init__(self, point, radius):
        self.point = point
        self.radius = radius

if __name__== "__main__" :
    start = (0.0, 0.0)
    goal = (15.0, 5.0)

    p1 = Point(2.0, 2.0)
    p2 = Point(3.0, 3.0)
    p3 = Point(4.0, 4.0)
    p4 = Point(5.0, 5.0)
    obs1 = Obstacle(p1, 1.0)
    obs2 = Obstacle(p2, 2.0)
    obs3 = Obstacle(p3, 3.0)
    obs4 = Obstacle(p4, 4.0)

    obstacles = [obs1, obs2, obs3, obs4]
    trajectory = [start]

    k = 7.0
    temp = start


    print(grad(APF)(k,start,goal, obstacles))

        # next = temp.move(grad(temp.APF(goal, k, obstacles)))
        # trajctory.append(next)
        # temp = next

你必须创建两个类来处理这个。。有关信息，请参阅我的参考类。你知道吗