旨在(zai)通(tong)(tong)(tong)(tong)(tong)(tong)過構建一個(ge)多(duo)模(mo)(mo)(mo)態交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)領(ling)域專用(yong)(yong)的(de)大規(gui)模(mo)(mo)(mo)深(shen)度學(xue)習模(mo)(mo)(mo)型，解(jie)(jie)決(jue)(jue)交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)行(xing)業(ye)中(zhong)數據(ju)處理(li)與決(jue)(jue)策(ce)(ce)制定(ding)過程中(zhong)存在(zai)的(de)難題。通(tong)(tong)(tong)(tong)(tong)(tong)過整合知(zhi)識(shi)圖譜、大語言模(mo)(mo)(mo)型、多(duo)模(mo)(mo)(mo)態交(jiao)互等(deng)(deng)先(xian)進技術和(he)海量交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)數據(ju)，實現(xian)從(cong)數據(ju)采集(ji)、處理(li)到智(zhi)能分析的(de)全流(liu)(liu)程智(zhi)能化(hua)，支持交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)專業(ye)知(zhi)識(shi)問答場景(jing)(jing)和(he)企(qi)業(ye)、政務(wu)輔助(zhu)辦(ban)公等(deng)(deng)場景(jing)(jing)，解(jie)(jie)決(jue)(jue)數據(ju)治理(li)、超大城(cheng)市(shi)(shi)治理(li)與規(gui)劃、交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)流(liu)(liu)預測與管控、公共交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)優化(hua)等(deng)(deng)難題，提(ti)(ti)供(gong)更(geng)加豐富和(he)直觀的(de)交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)信息服務(wu)，實現(xian)更(geng)為(wei)(wei)精(jing)準和(he)智(zhi)能的(de)事件發現(xian)與分析。為(wei)(wei)交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)管理(li)提(ti)(ti)供(gong)精(jing)準的(de)交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)預測、實時(shi)交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)監控和(he)應(ying)急響(xiang)應(ying)策(ce)(ce)略，為(wei)(wei)公眾提(ti)(ti)供(gong)個(ge)性化(hua)出行(xing)服務(wu)決(jue)(jue)策(ce)(ce)，為(wei)(wei)企(qi)業(ye)提(ti)(ti)供(gong)高效的(de)物流(liu)(liu)解(jie)(jie)決(jue)(jue)方案，同時(shi)基于大模(mo)(mo)(mo)型智(zhi)能應(ying)用(yong)(yong)平(ping)臺，為(wei)(wei)超大城(cheng)市(shi)(shi)提(ti)(ti)供(gong)城(cheng)市(shi)(shi)級(ji)交(jiao)通(tong)(tong)(tong)(tong)(tong)(tong)管理(li)規(gui)劃。

隨著城(cheng)(cheng)市化(hua)(hua)(hua)(hua)(hua)進程(cheng)的(de)(de)(de)不(bu)斷推進以及(ji)人(ren)民對(dui)城(cheng)(cheng)市品質(zhi)要求的(de)(de)(de)不(bu)斷提升，各行(xing)(xing)業(ye)的(de)(de)(de)管理(li)(li)決(jue)策水平(ping)均在向(xiang)精細化(hua)(hua)(hua)(hua)(hua)方向(xiang)轉變。面(mian)對(dui)當前(qian)智(zhi)慧(hui)交通(tong)(tong)行(xing)(xing)業(ye)數據處理(li)(li)與決(jue)策制定過(guo)程(cheng)中存在的(de)(de)(de)難題，通(tong)(tong)用大模型為(wei)垂(chui)直行(xing)(xing)業(ye)的(de)(de)(de)產(chan)業(ye)創新需(xu)求帶來(lai)了(le)新思路(lu)與新路(lu)徑。交通(tong)(tong)垂(chui)類大模型可以為(wei)城(cheng)(cheng)市管理(li)(li)者提供更(geng)(geng)加(jia)精細化(hua)(hua)(hua)(hua)(hua)、智(zhi)敏(min)化(hua)(hua)(hua)(hua)(hua)的(de)(de)(de)工具，為(wei)數據治理(li)(li)在城(cheng)(cheng)市交通(tong)(tong)行(xing)(xing)業(ye)的(de)(de)(de)深度(du)應用提供支撐，為(wei)城(cheng)(cheng)市居(ju)民的(de)(de)(de)出行(xing)(xing)提供更(geng)(geng)加(jia)智(zhi)慧(hui)化(hua)(hua)(hua)(hua)(hua)、綠色化(hua)(hua)(hua)(hua)(hua)、人(ren)性化(hua)(hua)(hua)(hua)(hua)的(de)(de)(de)服務選擇。

本項(xiang)目旨在通過構(gou)建(jian)交(jiao)通垂類大模型，為交(jiao)通領域(yu)內復雜的業務需求提供(gong)智能化(hua)解決方案。項(xiang)目覆蓋(gai)從基礎設施搭(da)建(jian)、數據治理(li)、大模型構(gou)建(jian)與優化(hua)，到服務提供(gong)和前(qian)端(duan)展(zhan)示的全流程，形(xing)成一個涵蓋(gai)數據處理(li)、模型訓練(lian)、優化(hua)、應用及展(zhan)示的閉環(huan)系統(tong)。

通(tong)達交(jiao)(jiao)通(tong)大模(mo)型將應(ying)用于運(yun)政管(guan)理、城市(shi)交(jiao)(jiao)通(tong)管(guan)理、智(zhi)能(neng)網(wang)聯等交(jiao)(jiao)通(tong)子領(ling)域，解決(jue)交(jiao)(jiao)通(tong)行(xing)業(ye)中(zhong)數據(ju)處理與(yu)決(jue)策制定過程(cheng)中(zhong)存在的(de)難題，提(ti)供包括交(jiao)(jiao)通(tong)領(ling)域的(de)智(zhi)能(neng)知識問答系統(tong)、自(zi)動生成各(ge)類工作報(bao)告的(de)能(neng)力，以及提(ti)出交(jiao)(jiao)通(tong)安全趨勢預警與(yu)應(ying)對方案等智(zhi)能(neng)功能(neng)，為交(jiao)(jiao)通(tong)從(cong)業(ye)者(zhe)、企業(ye)辦公、城市(shi)管(guan)理和公眾(zhong)出行(xing)提(ti)供全面(mian)的(de)智(zhi)能(neng)化(hua)服務。

1、基于我司對智能(neng)交通(tong)(tong)行(xing)(xing)業(ye)的(de)豐富(fu)的(de)行(xing)(xing)業(ye)經驗和數(shu)據積(ji)累，整理(li)(li)海量數(shu)據，提煉數(shu)據要素，整理(li)(li)高質量的(de)數(shu)據集；利(li)用行(xing)(xing)業(ye)領先的(de)通(tong)(tong)用大模型，訓練(lian)交通(tong)(tong)行(xing)(xing)業(ye)的(de)垂(chui)類大模型，通(tong)(tong)過集成語音識別、圖(tu)像識別和自然(ran)語言理(li)(li)解(jie)技(ji)術，以多種方式與用戶互(hu)動，提供(gong)了更直觀(guan)的(de)信息服務(wu)。

2、交(jiao)(jiao)(jiao)通(tong)(tong)(tong)大(da)(da)模(mo)型(xing)與傳統機器視覺模(mo)型(xing)的結(jie)合實(shi)(shi)現了精準的事件(jian)檢測(ce)和交(jiao)(jiao)(jiao)通(tong)(tong)(tong)管(guan)理(li)(li)優化。交(jiao)(jiao)(jiao)通(tong)(tong)(tong)大(da)(da)模(mo)型(xing)可以為機器視覺模(mo)型(xing)提(ti)供(gong)上(shang)下(xia)文信(xin)息，機器視覺模(mo)型(xing)可以為交(jiao)(jiao)(jiao)通(tong)(tong)(tong)大(da)(da)模(mo)型(xing)提(ti)供(gong)實(shi)(shi)時(shi)(shi)檢測(ce)數據，這(zhe)種雙(shuang)向的信(xin)息融合和實(shi)(shi)時(shi)(shi)互動(dong)，使(shi)得系(xi)統不(bu)僅能夠(gou)及時(shi)(shi)發現和響應交(jiao)(jiao)(jiao)通(tong)(tong)(tong)事件(jian)，還能夠(gou)根據實(shi)(shi)時(shi)(shi)情況(kuang)動(dong)態優化交(jiao)(jiao)(jiao)通(tong)(tong)(tong)管(guan)理(li)(li)措(cuo)施，提(ti)升整個交(jiao)(jiao)(jiao)通(tong)(tong)(tong)系(xi)統的智能化水平。

3、大模(mo)(mo)型的(de)插件化(hua)應(ying)(ying)用。通過將大模(mo)(mo)型的(de)核心(xin)功(gong)能(neng)封裝成獨立的(de)插件，可(ke)以輕松(song)地(di)將其集成到各種(zhong)已落地(di)的(de)應(ying)(ying)用平臺(tai)中，提高了模(mo)(mo)型的(de)靈活性和(he)易用性，便于集成和(he)維護(hu)。

以(yi)(yi)大模型平臺為基礎(chu)，打通(tong)交通(tong)管理(li)數據屏障，充分挖掘交通(tong)各類數據價(jia)值，提(ti)高城市交通(tong)管理(li)的決(jue)策(ce)能(neng)力與決(jue)策(ce)水平,以(yi)(yi)及(ji)工作效率，減(jian)少部門人員(yuan)投入的冗余度。通(tong)過(guo)將區(qu)域交通(tong)運行態(tai)(tai)勢(shi)指標化(hua)來反饋運行的實(shi)時動態(tai)(tai)，并根據路面發生的交通(tong)事件來激活各種(zhong)處置方案(an)，實(shi)現道路交通(tong)信息(xi)聯(lian)動應用。實(shi)現對客流高峰的精(jing)準識(shi)別與快速響(xiang)應，優(you)化(hua)疏散路徑，確(que)保(bao)疏散過(guo)程(cheng)高效且(qie)安全。

四、項目應用落地情(qing)況

通(tong)達交通(tong)大模型相關產品將會在上(shang)海、廣東、長沙、寧波等(deng)地進(jin)行試點(dian)應(ying)用，通(tong)過路側設(she)備和先進(jin)的(de)視覺識別技術(shu)，收集“車”、“路”、“事”、“物”等(deng)多維度數(shu)據，結(jie)合歷史交通(tong)數(shu)據和完善的(de)交通(tong)知識庫，對(dui)事件進(jin)行綜(zong)合研判、對(dui)交通(tong)信息進(jin)行審核分析，為交通(tong)管理部(bu)門(men)提供(gong)宏觀層面的(de)規(gui)劃和管理建議。

通達交通大模(mo)型能(neng)夠(gou)提供(gong)智能(neng)交通客(ke)服、設施設備智能(neng)運(yun)維等生成(cheng)式人(ren)工智能(neng)服務，全(quan)面賦能(neng)升(sheng)級交通場景應用和用戶(hu)體(ti)驗，滿足交通規劃、交通流量預測(ce)、自動(dong)駕駛、智能(neng)調度(du)、個性化出行等多場景需求，提升(sheng)交通規劃和管(guan)(guan)理效率(lv)，降低能(neng)源消耗，促(cu)進(jin)交通綠(lv)色低碳發展，同時促(cu)進(jin)智慧交通建設，助(zhu)力(li)城市精細化管(guan)(guan)理。同時推動(dong)大模(mo)型走向產業鏈縱深，助(zhu)力(li)建設大模(mo)型生態圈(quan)。

通(tong)(tong)達交(jiao)(jiao)通(tong)(tong)大模型主要通(tong)(tong)過(guo)服(fu)務(wu)訂閱、產品使用(yong)和平(ping)(ping)臺服(fu)務(wu)三種(zhong)方式盈利，此外，還可能(neng)通(tong)(tong)過(guo)與政企建立長期(qi)合作關系，提供持續的技術支持和優(you)化服(fu)務(wu)，來獲得(de)穩定的收入。已在公司智(zhi)慧高速、智(zhi)慧停車、智(zhi)慧交(jiao)(jiao)管等業務(wu)領域(yu)中得(de)到應用(yong)，如臨(lin)港公安交(jiao)(jiao)通(tong)(tong)一(yi)體化平(ping)(ping)臺、集運MaaS1.0港域(yu)數字孿生試點(dian)等項目利用(yong)大模型進行交(jiao)(jiao)通(tong)(tong)流量監測(ce)和預測(ce)、交(jiao)(jiao)通(tong)(tong)規劃等。