交通AI控制——實踐與方法

第 1 章 百年城市交通控制········································································.001
1.1 交通控制起源........................................................................................................001
1.2 城市交通控制的第一次技術革命 ........................................................................002
1.3 城市交通控制的第二次技術革命 ........................................................................004
1.4 城市交通控制的第三次技術革命 ........................................................................006
1.4.1 高效與低效的差別.....................................................................................007
1.4.2 城市交通控制方法的瓶頸 .........................................................................008
1.4.3 AI 將輔助突破當前交通控制低效的瓶頸 ................................................010
第 2 章 交通控制技術演進歷程··································································.013
2.1 無數據時代的交通控制技術................................................................................013
2.2 小數據時代的交通控制技術................................................................................015
2.3 大數據時代的交通控制技術................................................................................016
2.3.1 “互聯網+信號燈”的興起.......................................................................016
2.3.2 “互聯網+信號燈”的局限性...................................................................017
2.4 AI 時代的交通控制技術......................................................................................019
第 3 章 交通控制應用現狀········································································.021
3.1 交通控制業務現狀................................................................................................021
3.1.1 信號配時工程師與交通信號配時中心 .....................................................021
3.1.2 信號配時人員及交通信號配時中心的困境 .............................................023
3.1.3 交通信號配時中心業務面臨的挑戰 .........................................................025
3.2 工程側交通數據現狀............................................................................................026
3.2.1 工程側對交通數據的達標要求 .................................................................026
3.2.2 環形線圈車輛檢測器數據 .........................................................................027
3.2.3 地磁車輛檢測器數據.................................................................................028
3.2.4 雷達檢測器數據.........................................................................................029
3.2.5 視頻車輛檢測器數據.................................................................................030
3.2.6 互聯網數據.................................................................................................032
3.3 工程側網絡環境現狀............................................................................................034
3.4 交通控制安全現狀................................................................................................036
3.4.1 交通控制安全問題.....................................................................................036
3.4.2 交通控制安全問題解決思路 .....................................................................038
第 4 章 交通控制機遇與挑戰·····································································.040
4.1 交通控制業務應如何升級....................................................................................040
4.1.1 交通控制業務的三次變革 .........................................................................040
4.1.2 城市交通管理業務的主要模式 .................................................................042
4.1.3 交通管理產品的疊代歷程 .........................................................................043
4.2 交通控制系統如何解決數據瓶頸 ........................................................................044
4.2.1 交通控制系統的數據瓶頸 .........................................................................044
4.2.2 交通控制系統突破數據瓶頸的路徑 .........................................................046
4.3 交通控制應用如何突破當前桎梏 ........................................................................055
4.3.1 國外經驗參考.............................................................................................055
4.3.2 經驗借鑒轉化.............................................................................................056
4.3.3 國內推進工作.............................................................................................057
4.4 交通控制產業如何走向良性健康 ........................................................................061
4.4.1 SCATS Cornerstone 及未來創新模式........................................................061
4.4.2 我國交通控制產業革新模式 .....................................................................063
第 5 章 視圖 AI 應用於交通視頻分析··························································.066
5.1 視圖 AI 在城市交通領域的應用簡介..................................................................066
5.2 交通視頻解析的技術架構....................................................................................068
5.2.1 視頻采集模塊.............................................................................................068
5.2.2 預處理模塊.................................................................................................069
5.2.3 目標檢測與目標跟蹤模塊 .........................................................................070
5.2.4 場景分割模塊.............................................................................................071
5.2.5 行為分析模塊.............................................................................................071
5.2.6 事件檢測模塊.............................................................................................072
5.2.7 數據融合與統計模塊.................................................................................072
5.2.8 數據存儲與管理模塊.................................................................................073
5.2.9 可視化與報警模塊.....................................................................................073
5.2.10 API 模塊....................................................................................................074
5.3 交通視頻結構化的技術原理................................................................................075
5.3.1 交通視頻接入.............................................................................................075
5.3.2 交通視頻解碼.............................................................................................077
5.3.3 圖像算法推理.............................................................................................078
5.3.4 目標跟蹤.....................................................................................................079
5.3.5 數據中介.....................................................................................................080
5.3.6 OSD 信息疊加............................................................................................082
5.3.7 視頻編碼.....................................................................................................083
5.3.8 視頻推流.....................................................................................................085
5.3.9 視頻結構化硬件平臺.................................................................................086
5.4 視圖 AI 解析的交通數據種類..............................................................................087
5.4.1 視圖 AI 解析的交通數據種類...................................................................087
5.4.2 視圖 AI 解析的交通數據質量...................................................................088
5.4.3 核心問題分析.............................................................................................090
5.5 視圖 AI 在交通 AI 控制中的作用........................................................................091
第 6 章 存量檢測器采集的交通流量····························································.094
6.1 路口交通流運行特性分析....................................................................................095
6.1.1 路口交通流相似性.....................................................................................095
6.1.2 路口交通流不確定性.................................................................................096
6.2 路段交通流運行特性分析....................................................................................097
6.2.1 路段交通流相似性.....................................................................................097
6.2.2 路段交通流不確定性.................................................................................098
6.3 區域交通流空間特性分析....................................................................................099
6.3.1 區域交通流空間分布相似性 .....................................................................099
6.3.2 區域交通流空間分布不均勻性 .................................................................
100
6.4 存量檢測器數據情況............................................................................................101
6.4.1 SCATS 檢測器數據....................................................................................101
6.4.2 SCATS 檢測器質量分析方法 ....................................................................102
6.4.3 分析結果.....................................................................................................103
第 7 章 存量檢測器數據質量診斷·······························································.105
7.1 檢測器數據的時空分布........................................................................................105
7.2 檢測器數據的特征及分類....................................................................................107
7.2.1 特征類型.....................................................................................................107
7.2.2 特征分類準則與標簽.................................................................................108
7.3 基於決策樹的檢測器數據質量診斷方法 ............................................................111
7.3.1 決策樹分類原理.........................................................................................111
7.3.2 數據質量診斷決策樹模型 .........................................................................113
7.3.3 對數據質量診斷決策樹進行剪枝優化 .....................................................113
7.4 檢測器數據質量實例分析....................................................................................114
第 8 章 信號控制系統配時參數合理性診斷···················································.116
8.1 交通信號配時參數特性........................................................................................116
8.1.1 綠信比方案與路網運行相關性 .................................................................116
8.1.2 信號周期與交通參數相關性 .....................................................................118
8.2 綠信比方案合理性診斷........................................................................................119
8.2.1 路網運行需求擬合.....................................................................................119
8.2.2 綠信比方案合理性判定.............................................................................120
8.3 信號周期方案合理性診斷....................................................................................121
8.3.1 交通參數映射函數構建.............................................................................121
8.3.2 信號周期方案合理性判定 .........................................................................124
8.4 交通信號配時參數診斷案例................................................................................125
8.4.1 信號配時綠信比方案合理性診斷 .............................................................125
8.4.2 信號周期方案合理性診斷 .........................................................................126
第 9 章 交通網絡重要節點及控制子區判斷···················································.127
9.1 城市交通網絡節點重要度評估指標 ....................................................................127
9.1.1 基於復雜網絡特性的節點重要度評估指標 .............................................127
9.1.2 基於交通流運行特征的節點重要度評估指標 .........................................129
9.2 節點重要度評估模型............................................................................................130
9.2.1 評估指標的確定.........................................................................................130
9.2.2 評估指標權重方法的確定 .........................................................................131
9.2.3 基於 TOPSIS 法的節點重要度評估方法..................................................137
9.3 構建路網關聯度模型............................................................................................139
9.3.1 控制子區劃分原則.....................................................................................139
9.3.2 關聯度模型的選取.....................................................................................140
9.4 基於改進標簽傳播算法的子區劃分方法 ............................................................142
9.4.1 標簽傳播算法介紹.....................................................................................142
9.4.2 標簽傳播算法改進.....................................................................................144
9.4.3 基於改進標簽傳播算法的控制子區動態劃分流程..................................146
9.5 改進標簽傳播算法的控制子區動態劃分演算 ....................................................147
9.6 基於合並指數的控制子區動態劃分方法演算 ....................................................153
9.7 劃分方法的評估對比............................................................................................157
9.7.1 基於模塊度的劃分方法評估 .....................................................................158
9.7.2 基於控制效益的劃分方法評估 .................................................................
160
第 10 章 基於時空混合圖卷積的車速預測····················································.163
10.1 交通預測問題描述..............................................................................................163
10.1.1 問題分析和數據.......................................................................................163
10.1.2 交通事件矩陣獨熱編碼 ...........................................................................163
10.2 深度時空混合圖卷積網絡交通預測任務 ..........................................................164
10.2.1 DHSGC 模型整體結構.............................................................................164
10.2.2 DHSGC 模型混合圖卷積層.....................................................................166
10.2.3 DHSGC 模型時間特征提取層.................................................................
167
10.2.4 路段並行預測處理模塊 ...........................................................................170
10.2.5 路網靜態和動態因素編碼 .......................................................................171
10.3 車速預測案例分析..............................................................................................172
10.3.1 數據集介紹和預處理...............................................................................172
10.3.2 DHSGC 模型對比算法.............................................................................172
10.3.3 實驗環境與參數設置...............................................................................172
10.3.4 實驗結果與算法性能分析 .......................................................................173
10.3.5 外部因素消融實驗...................................................................................176
10.4 以速度估計排隊車輛數的可行性分析 ..............................................................176
10.5 基於 LSTM 的排隊車輛數估計..........................................................................177
10.5.1 循環神經網絡與 LSTM............................................................................177
10.5.2 路口建模...................................................................................................181
10.6 實驗與結果分析..................................................................................................183
10.6.1 仿真搭建與數據準備...............................................................................183
10.6.2 實驗結果分析...........................................................................................184
第 11 章 基於改進 ACGAN 的排隊估計 ·····················································.186
11.1 GAN 機制 ............................................................................................................186
11.1.1 GAN 算法原理與學習方法......................................................................186
11.1.2 ACGAN 算法原理....................................................................................187
11.2 Bi-LSTM 網絡機制 .............................................................................................189
11.3 排隊車輛數估計算法 ..........................................................................................190
11.3.1 模型結構...................................................................................................190
11.3.2 目標函數...................................................................................................192
11.3.3 訓練優化...................................................................................................192
11.4 實驗與結果分析..................................................................................................193
11.4.1 訓練細節...................................................................................................193
11.4.2 實驗結果分析 ...........................................................................................193
第 12 章 車輛排隊微觀仿真框架設計··························································.197
12.1 仿真框架需求分析..............................................................................................197
12.2 仿真框架功能設計..............................................................................................198
12.2.1 數據收集模塊...........................................................................................198
12.2.2 多源數據庫...............................................................................................199
12.2.3 算法模塊...................................................................................................199
12.2.4 估計模塊...................................................................................................201
12.2.5 仿真模塊...................................................................................................201
12.2.6 交互可視化模塊.......................................................................................201
12.3 車輛排隊仿真思路..............................................................................................202
12.3.1 相位時刻確定...........................................................................................202
12.3.2 仿真情況分析...........................................................................................203
12.4 車輛排隊仿真整體流程......................................................................................204
第 13 章 智組織交通控制方法···································································.206
13.1 自適應信號控制失效的原因分析 ......................................................................206
13.2 智組織交通控制的基本原理..............................................................................209
第 14 章 單點智組織控制方法···································································.211
14.1 TSS-CII 配時方法的計算基礎............................................................................211
14.1.1 候選信號燈組設計...................................................................................211
14.1.2 各流向權重的確定...................................................................................213
14.1.3 等待時間上限及其確定方法 ...................................................................213
14.1.4 基於地點交通參數的 TSS-CII 配時方法................................................214
14.2 各車道排隊車輛數的計算方法 ..........................................................................215
14.3 關鍵流向的確定方法..........................................................................................216
14.4 擬放行信號燈組綠燈時間的確定方法 ..............................................................217
14.5 放行信號燈組的確定方法..................................................................................217
14.6 基於路段交通參數的 TSS-CII 配時方法...........................................................218
14.7 模擬分析..............................................................................................................219
14.7.1 模擬方案設計...........................................................................................219
14.7.2 實驗結果初步分析...................................................................................222
14.7.3 檢測器布設位置方案分析 .......................................................................223
14.7.4 飽和車頭時距分析...................................................................................224
第 15 章 幹線智組織控制方法···································································.226
15.1 傳統幹線協調控制方法失效原因分析 ..............................................................226
15.2 TSS-CAI 的基本原則與概念界定......................................................................229
15.2.1 TSS-CAI 的基本原則...............................................................................229
15.2.2 相關概念的界定.......................................................................................229
15.3 幹線協調控制有效帶寬評價算法 ......................................................................231
15.4 TSS-CAI 算法設計..............................................................................................232
15.4.1 TSS-CAI 控制策略...................................................................................232
15.4.2 TSS-CAI 控制算法的處理過程...............................................................233
15.4.3 上遊路口幹線方向綠燈時間設計 ...........................................................233
15.5 模擬分析..............................................................................................................234
15.5.1 模擬方案設計...........................................................................................234
15.5.2 有效帶寬搜索分析...................................................................................235
15.5.3 模擬結果及對比分析...............................................................................237
15.5.4 有效帶寬閾值分析...................................................................................238
第 16 章 區域智組織控制方法···································································.240
16.1 傳統區域協調控制方法失效原因分析 ..............................................................240
16.2 TSS-CDI 的基本概念界定..................................................................................243
16.3 TSS-CDI 動態控制子區的劃分方法..................................................................245
16.3.1 擁擠條件下的控制子區及其動態劃分定義 ...........................................245
16.3.2 局部擁擠條件下控制子區劃分的方法 ...................................................246
16.3.3 TSS-CDI 截流點和卸載點的確定方法 ...................................................248
16.4 TSS-CDI 算法設計..............................................................................................249
16.4.1 TSS-CDI 控制策略...................................................................................249
16.4.2 TSS-CDI 算法流程...................................................................................249
16.5 模擬分析..............................................................................................................250
16.5.1 模擬方案設計...........................................................................................250
16.5.2 模擬結果及對比分析...............................................................................254
第 17 章 交通控制方案滾動生成系統設計····················································.257
17.1 系統架構..............................................................................................................257
17.2 系統功能..............................................................................................................258
17.3 數據結構..............................................................................................................259
第 18 章 交通 AI 控制算法框架·································································.262
18.1 交通控制要素......................................................................................................262
18.1.1 關鍵參數...................................................................................................262
18.1.2 評價指標...................................................................................................263
18.2 深度強化學習算法..............................................................................................264
18.2.1 DRL 算法..................................................................................................265
18.2.2 RL 算法.....................................................................................................266
18.2.3 卷積神經網絡...........................................................................................267
18.2.4 註意力機制...............................................................................................270
18.3 交通 AI 控制框架設計........................................................................................270
18.3.1 控制對象研究...........................................................................................270
18.3.2 DRL 的交通控制框架設計 ......................................................................272
第 19 章 交通 AI 控制方法·······································································.274
19.1 多智能體深度強化學習模型..............................................................................274
19.2 基於 GCN 速度預測的交通環境........................................................................275
19.2.1 子區圖網絡建模.......................................................................................275
19.2.2 子區速度預測...........................................................................................277
19.3 基於改進型 QMIX 算法的多智能體設計..........................................................277
19.3.1 子區智能體設計.........................................................................................277
19.3.2 QMIX 動作決策 .......................................................................................280
19.4 基於滾動時域的控制模式..................................................................................282
19.5 控制方案滾動生成方法......................................................................................283
第 20 章 交通 AI 控制方案驗證·································································.284
20.1 實驗數據..............................................................................................................284
20.1.1 基礎數據...................................................................................................284
20.1.2 運行數據...................................................................................................285
20.2 模型訓練..............................................................................................................286
20.2.1 系統環境...................................................................................................286
20.2.2 預測模型訓練...........................................................................................286
20.2.3 控制模型訓練...........................................................................................288
20.3 實驗結果分析......................................................................................................289
後記 ······································································································.292
參考文獻 ································································································.295