基于python机械臂深度学习正逆运动学+源码（期末大作业）

# -*- coding: utf-8 -*- """正解和反解.ipynb Automatically generated by zzh. """ from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense,Activation import numpy as np data = np.loadtxt('data.csv',delimiter=',') X = data[:,0:2] y = data[:,2:] # 创建一个顺序模型 model = Sequential() # 第一层 两个输入的数据 一个输出的数据 model.add(Dense(units=8,input_dim=2)) model.add(Activation('elu')) model.add(Dense(units=16)) model.add(Activation('elu')) model.add(Dense(units=16)) model.add(Activation('elu')) model.add(Dense(units=16)) model.add(Activation('elu')) model.add(Dense(units=16)) model.add(Activation('elu')) model.add(Dense(2)) # 第三层 激活函数 model.add(Activation('linear')) model.summary() # 设置模型的训练参数, 误差函数 loss, 优化算法 optimizer model.compile(optimizer='adam',loss='mse',metrics=['accuracy','mse']) model.fit(X,y,epochs=50,batch_size=64) model.predict(np.array([[31.5,31.5]])) model.save("forward.h5") from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense,Activation import numpy as np data = np.loadtxt('data2.csv',delimiter=',') X = data[:,0:2] y = data[:,2:] # 创建一个顺序模型 model = Sequential() # 第一层 两个输入的数据 一个输出的数据 model.add(Dense(units=8,input_dim=3)) model.add(Activation('tanh')) model.add(Dense(units=100)) model.add(Activation('tanh')) model.add(Dense(units=100)) model.add(Dense(2)) # 第三层 激活函数 model.add(Activation('linear')) model.summary() # 设置模型的训练参数, 误差函数 loss, 优化算法 optimizer model.compile(optimizer='adam',loss='mse',metrics=['accuracy','mse']) model.fit(y,X,epochs=150,batch_size=64) model.predict(np.array([[2.34,2.22,70]])) """# 生成数据""" import numpy as np from math import radians, degrees, cos, sin, asin, fabs, atan2 def _clamp(value, x, y): n = y if y < value else value return x if x > n else n def _matrix2euler(m): m11 = m[0, 0] m12 = m[0, 1] m13 = 0 m21 = m[1, 0] m22 = m[1, 1] m23 = 0 m31 = m[2, 0] m32 = m[2, 1] m33 = 0 x = 0 y = 0 z = 0 y = asin(-_clamp(m31, -1, 1)) if fabs(m31) < 0.9999999: x = atan2(m32, m33) z = atan2(m21, m11) else: x = 0 z = atan2(-m12, m22) return x, y, z def _transform(links, i, theta): rz = np.array([ [cos(theta), -sin(theta), 0], [sin(theta), cos(theta), 0], [0, 0, 1] ]) px = 0 if i != 0: px = links[i - 1] tx = np.array([ [1, 0, px], [0, 1, 0], [0, 0, 1] ]) return rz.dot(tx) def forward(links, joints): """ 正解 :param links: 连杆列表，每根连杆的长度 :param joints: 关节角度 :return: """ if not isinstance(links, list) and not isinstance(links, tuple): raise Exception("link error") if not isinstance(joints, list) and not isinstance(joints, tuple): raise Exception("joint error") if len(links) != len(joints): raise Exception("joint and link num error") mtx = np.identity(3) for i in range(len(links) + 1): angle = 0 if i != 0: angle = radians(joints[i - 1]) mtx = mtx.dot(_transform(links, i, angle)) euler = _matrix2euler(mtx) return mtx[0, 2], mtx[1, 2], degrees(euler[2]) f = open('data2.csv','w') print(forward((2, 1.5), (30, 30))) for i in np.linspace(-180,180,1000): for j in np.linspace(-180,180,1000): x,y,theta = forward((2, 1.5), (i, j)) f.write('{},{},{},{},{}\n'.format(i,j,x,y,theta)) f.close() from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense,Activation import numpy as np import time data = np.loadtxt('data2.csv',delimiter=',') X = data[:,0:2] y = data[:,2:] # 创建一个顺序模型 model = Sequential() # 第一层 两个输入的数据 一个输出的数据 model.add(Dense(units=8,input_dim=3)) model.add(Activation('tanh')) model.add(Dense(units=100)) model.add(Activation('tanh')) model.add(Dense(units=100)) model.add(Activation('tanh')) model.add(Dense(units=100)) model.add(Dense(2)) # 第三层 激活函数 model.add(Activation('linear')) model.summary() # 设置模型的训练参数, 误差函数 loss, 优化算法 optimizer model.compile(optimizer='adam',loss='mse',metrics=['accuracy','mse']) model.fit(y,X,epochs=800,batch_size=64) model.save("backward.h5") model.predict(np.array([[2.34,2.22,70]]))