原文:不想本人开车,Python帮你搞定主动驾驶

01  装置环境

gym是用于开发和比拟强化学习算法的工具包,在python中装置gym库和其中子场景都较为简便。装置gym:

pip install gym

装置主动驾驶模块,这里应用Edouard Leurent公布在github上的包highway-env(链接:https://github.com/eleurent/h...):

pip install --user git+https://github.com/eleurent/highway-env

其中蕴含6个场景:
高速公路——“highway-v0”
汇入——“merge-v0”
环岛——“roundabout-v0”
泊车——“parking-v0”
十字路口——“intersection-v0”
赛车道——“racetrack-v0”
具体文档能够参考这里:https://highway-env.readthedo...

02  配置环境

装置好后即可在代码中进行试验(以高速公路场景为例):

import gymimport highway_env%matplotlib inlineenv = gym.make('highway-v0')env.reset()for _ in range(3):    action = env.action_type.actions_indexes["IDLE"]    obs, reward, done, info = env.step(action)    env.render()

运行后会在模拟器中生成如下场景:

绿色为ego vehicle env类有很多参数能够配置,具体能够参考原文档。

03  训练模型

3.1 数据处理

(1)stat
ehighway-env包中没有定义传感器,车辆所有的state (observations) 都从底层代码读取,节俭了许多后期的工作量。依据文档介绍,state (ovservations) 有三种输入形式:Kinematics,Grayscale Image和Occupancy grid。
Kinematics
输入V*F的矩阵,V代表须要观测的车辆数量(包含ego vehicle自身),F代表须要统计的特色数量。例:

数据生成时会默认归一化,取值范畴:[100, 100, 20, 20],也能够设置ego vehicle以外的车辆属性是地图的相对坐标还是对ego vehicle的绝对坐标。
在定义环境时须要对特色的参数进行设定:

config = \    {    "observation":          {        "type": "Kinematics",        #选取5辆车进行察看(包含ego vehicle)        "vehicles_count": 5,          #共7个特色        "features": ["presence", "x", "y", "vx", "vy", "cos_h", "sin_h"],         "features_range":             {            "x": [-100, 100],            "y": [-100, 100],            "vx": [-20, 20],            "vy": [-20, 20]            },        "absolute": False,        "order": "sorted"        },    "simulation_frequency": 8,  # [Hz]    "policy_frequency": 2,  # [Hz]    }

Grayscale Image
生成一张W*H的灰度图像,W代表图像宽度,H代表图像高度
Occupancy grid
生成一个WHF的三维矩阵,用W*H的表格示意ego vehicle四周的车辆状况,每个格子蕴含F个特色。
(2) action
highway-env包中的action分为间断和离散两种。连续型action能够间接定义throttle和steering angle的值,离散型蕴含5个meta actions:

        0: 'LANE_LEFT',        1: 'IDLE',        2: 'LANE_RIGHT',        3: 'FASTER',        4: 'SLOWER'    }

(3) reward
highway-env包中除了泊车场景外都采纳同一个reward function:

这个function只能在其源码中更改,在外层只能调整权重。(泊车场景的reward function原文档里有,懒得打公式了……)

3.2 搭建模型

DQN网络的构造和搭建过程曾经在我另一篇文章中探讨过,所以这里不再具体解释。我采纳第一种state示意形式——Kinematics进行示范。
因为state数据量较小(5辆车*7个特色),能够不思考应用CNN,间接把二维数据的size[5,7]转成[1,35]即可,模型的输出就是35,输入是离散action数量,共5个。

import torchimport torch.nn as nnfrom torch.autograd import Variableimport torch.nn.functional as Fimport torch.optim as optimimport torchvision.transforms as Tfrom torch import FloatTensor, LongTensor, ByteTensorfrom collections import namedtupleimport random Tensor = FloatTensorEPSILON = 0    # epsilon used for epsilon greedy approachGAMMA = 0.9TARGET_NETWORK_REPLACE_FREQ = 40       # How frequently target netowrk updatesMEMORY_CAPACITY = 100BATCH_SIZE = 80LR = 0.01         # learning rateclass DQNNet(nn.Module):    def __init__(self):        super(DQNNet,self).__init__()        self.linear1 = nn.Linear(35,35)        self.linear2 = nn.Linear(35,5)    def forward(self,s):        s=torch.FloatTensor(s)        s = s.view(s.size(0),1,35)        s = self.linear1(s)        s = self.linear2(s)        return sclass DQN(object):    def __init__(self):        self.net,self.target_net = DQNNet(),DQNNet()        self.learn_step_counter = 0        self.memory = []        self.position = 0        self.capacity = MEMORY_CAPACITY        self.optimizer = torch.optim.Adam(self.net.parameters(), lr=LR)        self.loss_func = nn.MSELoss()    def choose_action(self,s,e):        x=np.expand_dims(s, axis=0)        if np.random.uniform() < 1-e:              actions_value = self.net.forward(x)              action = torch.max(actions_value,-1)[1].data.numpy()              action = action.max()        else:              action = np.random.randint(0, 5)        return action    def push_memory(self, s, a, r, s_):        if len(self.memory) < self.capacity:            self.memory.append(None)        self.memory[self.position] = Transition(torch.unsqueeze(torch.FloatTensor(s), 0),torch.unsqueeze(torch.FloatTensor(s_), 0),\                                                torch.from_numpy(np.array([a])),torch.from_numpy(np.array([r],dtype='float32')))#        self.position = (self.position + 1) % self.capacity    def get_sample(self,batch_size):        sample = random.sample(self.memory,batch_size)        return sample    def learn(self):        if self.learn_step_counter % TARGET_NETWORK_REPLACE_FREQ == 0:            self.target_net.load_state_dict(self.net.state_dict())        self.learn_step_counter += 1        transitions = self.get_sample(BATCH_SIZE)        batch = Transition(*zip(*transitions))        b_s = Variable(torch.cat(batch.state))        b_s_ = Variable(torch.cat(batch.next_state))        b_a = Variable(torch.cat(batch.action))        b_r = Variable(torch.cat(batch.reward))        q_eval = self.net.forward(b_s).squeeze(1).gather(1,b_a.unsqueeze(1).to(torch.int64))        q_next = self.target_net.forward(b_s_).detach() #        q_target = b_r + GAMMA * q_next.squeeze(1).max(1)[0].view(BATCH_SIZE, 1).t()        loss = self.loss_func(q_eval, q_target.t())        self.optimizer.zero_grad() # reset the gradient to zero        loss.backward()        self.optimizer.step() # execute back propagation for one step        return lossTransition = namedtuple('Transition',('state', 'next_state','action', 'reward'))

3.3 运行后果

各个局部都实现之后就能够组合在一起训练模型了,流程和用CARLA差不多,就不细说了。
初始化环境(DQN的类加进去就行了):

import gymimport highway_envfrom matplotlib import pyplot as pltimport numpy as npimport timeconfig = \    {    "observation":        {        "type": "Kinematics",        "vehicles_count": 5,        "features": ["presence", "x", "y", "vx", "vy", "cos_h", "sin_h"],        "features_range":            {            "x": [-100, 100],            "y": [-100, 100],            "vx": [-20, 20],            "vy": [-20, 20]            },        "absolute": False,        "order": "sorted"        },    "simulation_frequency": 8,  # [Hz]    "policy_frequency": 2,  # [Hz]    }    env = gym.make("highway-v0")env.configure(config)

训练模型:

dqn=DQN()count=0reward=[]avg_reward=0all_reward=[]time_=[]all_time=[]collision_his=[]all_collision=[]while True:    done = False    start_time=time.time()    s = env.reset()    while not done:        e = np.exp(-count/300)  #随机抉择action的概率,随着训练次数增多逐步升高        a = dqn.choose_action(s,e)        s_, r, done, info = env.step(a)        env.render()        dqn.push_memory(s, a, r, s_)        if ((dqn.position !=0)&(dqn.position % 99==0)):            loss_=dqn.learn()            count+=1            print('trained times:',count)            if (count%40==0):                avg_reward=np.mean(reward)                avg_time=np.mean(time_)                collision_rate=np.mean(collision_his)                all_reward.append(avg_reward)                all_time.append(avg_time)                all_collision.append(collision_rate)                plt.plot(all_reward)                plt.show()                plt.plot(all_time)                plt.show()                plt.plot(all_collision)                plt.show()                reward=[]                time_=[]                collision_his=[]        s = s_        reward.append(r)            end_time=time.time()    episode_time=end_time-start_time    time_.append(episode_time)           is_collision=1 if info['crashed']==True else 0    collision_his.append(is_collision)

我在代码中增加了一些画图的函数,在运行过程中就能够把握一些要害的指标,每训练40次统计一次平均值。
均匀碰撞发生率:

epoch均匀时长(s):

均匀reward:

能够看出均匀碰撞发生率会随训练次数增多逐步升高,每个epoch继续的工夫会逐步缩短(如果产生碰撞epoch会立即完结)。

一文尽览 | 轨迹预测二十年倒退全面回顾!
Tesla技术计划深度分析:主动标注/感知定位/决策布局/场景重建/场景仿真/数据引擎
PythonRobotics | 基于python的机器人自主导航
IROS2022 | 雪天环境的激光点云解决
全局门路布局 | RRT算法原理及实现
两万字 | 视觉SLAM钻研综述与将来趋势探讨
移动机器人视觉SLAM回环检测原理、现状及趋势