下载文档原格式
人工智能生态相关的专业名词解释下载温馨提示：该文档是我店铺精心编制而成，希望大家下载以后，能够帮助大家解决实际的问题。
文档下载后可定制随意修改，请根据实际需要进行相应的调整和使用，谢谢!并且，本店铺为大家提供各种各样类型的实用资料，如教育随笔、日记赏析、句子摘抄、古诗大全、经典美文、话题作文、工作总结、词语解析、文案摘录、其他资料等等，如想了解不同资料格式和写法，敬请关注!Download tips: This document is carefully compiled by the editor. I hope that after you download them, they can help yousolve practical problems. The document can be customized and modified after downloading, please adjust and use it according to actual needs, thank you!In addition, our shop provides you with various types of practical materials, such as educational essays, diary appreciation, sentence excerpts, ancient poems, classic articles, topic composition, work summary, word parsing, copy excerpts,other materials and so on, want to know different data formats and writing methods, please pay attention!人工智能技术的快速发展，推动着人工智能生态系统的形成和完善。
VTS-E航海-1：引言1.1 编写目的本文档旨在详细介绍VTS-E航海系统中的部分，包括相关技术、算法和应用。
通过该文档，读者可以了解到在VTS-E航海系统中的作用以及实施细节。
1.2 读者对象本文档的主要读者对象为VTS-E航海系统的开发人员、技术服务人员以及相关部门的管理人员。
读者需具备一定的航海和相关背景知识。
2：术语定义2.1 （Artificial Intelligence, ）是指以机器智能为主体，研究、开发和应用用于模拟、延伸和扩展人的智能的理论、方法、技术和应用系统的学科。
2.2 VTS-E航海系统VTS-E航海系统是一种基于技术的航海管理系统，主要用于保障船舶安全、提高航运效率和减少事故风险。
3：系统架构3.1 硬件架构VTS-E航海系统的硬件架构由多个核心组件组成，包括船舶传感器、雷达系统、图像识别设备、通信设备等。
3.2 软件架构VTS-E航海系统的软件架构采用分布式架构，包括多个子系统的集成，如车辆跟踪系统、危险物品监测系统、航线规划系统、预警系统等。
4：技术4.1 机器学习机器学习是的核心技术之一，通过从大量数据中学习规律和模式，实现对未知数据的预测和决策。
4.2 深度学习深度学习是机器学习的一种方法，通过构建多层神经网络模型，实现对复杂问题的自动学习和分析。
4.3 图像识别图像识别是的一个重要应用领域，通过对图像的分析和处理，实现对物体、场景和人脸等的识别和分类。
5：应用场景5.1 船舶目标检测利用图像识别和深度学习技术，对船舶进行目标检测和跟踪，实现对船舶的自动识别和位置追踪。
5.2 航线规划结合历史数据和机器学习技术，对航线进行预测和规划，使船舶能够选择最佳航线以提高航运效率。
5.3 事故预警通过实时监测船舶状态和环境数据，结合机器学习模型，实现事故预警和风险评估，提前采取措施避免事故发生。
6：附件本文档涉及的附件包括系统设计图、算法流程图、示例数据等。
7：法律名词及注释7.1 智能合同智能合同是基于区块链和智能合约技术实现的自动执行的合同，无需人为干预。
a r X i v :0806.1256v 1 [p h y s i c s .s o c -p h ] 7 J u n 2008Understanding individual human mobility patternsMarta C.Gonz´a lez,1,2C´e sar A.Hidalgo,1and Albert-L´a szl´o Barab´a si 1,2,31Center for Complex Network Research and Department of Physics and Computer Science,University of Notre Dame,Notre Dame IN 46556.2Center for Complex Network Research and Department of Physics,Biology and Computer Science,Northeastern University,Boston MA 02115.3Center for Cancer Systems Biology,Dana Farber Cancer Institute,Boston,MA 02115.(Dated:June 7,2008)Despite their importance for urban planning [1],traffic forecasting [2],and the spread of biological [3,4,5]and mobile viruses [6],our understanding of the basic laws govern-ing human motion remains limited thanks to the lack of tools to monitor the time resolved location of individuals.Here we study the trajectory of 100,000anonymized mobile phone users whose position is tracked for a six month period.We find that in contrast with the random trajectories predicted by the prevailing L´e vy flight and random walk models [7],human trajectories show a high degree of temporal and spatial regularity,each individual being characterized by a time independent characteristic length scale and a significant prob-ability to return to a few highly frequented locations.After correcting for differences in travel distances and the inherent anisotropy of each trajectory,the individual travel patterns collapse into a single spatial probability distribution,indicating that despite the diversity of their travel history,humans follow simple reproducible patterns.This inherent similarity in travel patterns could impact all phenomena driven by human mobility,from epidemic prevention to emergency response,urban planning and agent based modeling.Given the many unknown factors that influence a population’s mobility patterns,ranging from means of transportation to job and family imposed restrictions and priorities,human trajectories are often approximated with various random walk or diffusion models [7,8].Indeed,early mea-surements on albatrosses,bumblebees,deer and monkeys [9,10]and more recent ones on marine predators [11]suggested that animal trajectory is approximated by a L´e vy flight [12,13],a random walk whose step size ∆r follows a power-law distribution P (∆r )∼∆r −(1+β)with β<2.While the L´e vy statistics for some animals require further study [14],Brockmann et al.[7]generalized this finding to humans,documenting that the distribution of distances between consecutive sight-ings of nearly half-million bank notes is fat tailed.Given that money is carried by individuals, bank note dispersal is a proxy for human movement,suggesting that human trajectories are best modeled as a continuous time random walk with fat tailed displacements and waiting time dis-tributions[7].A particle following a L´e vyflight has a significant probability to travel very long distances in a single step[12,13],which appears to be consistent with human travel patterns:most of the time we travel only over short distances,between home and work,while occasionally we take longer trips.Each consecutive sightings of a bank note reflects the composite motion of two or more indi-viduals,who owned the bill between two reported sightings.Thus it is not clear if the observed distribution reflects the motion of individual users,or some hitero unknown convolution between population based heterogeneities and individual human trajectories.Contrary to bank notes,mo-bile phones are carried by the same individual during his/her daily routine,offering the best proxy to capture individual human trajectories[15,16,17,18,19].We used two data sets to explore the mobility pattern of individuals.Thefirst(D1)consists of the mobility patterns recorded over a six month period for100,000individuals selected randomly from a sample of over6million anonymized mobile phone users.Each time a user initiates or receives a call or SMS,the location of the tower routing the communication is recorded,allowing us to reconstruct the user’s time resolved trajectory(Figs.1a and b).The time between consecutive calls follows a bursty pattern[20](see Fig.S1in the SM),indicating that while most consecutive calls are placed soon after a previous call,occasionally there are long periods without any call activity.To make sure that the obtained results are not affected by the irregular call pattern,we also study a data set(D2)that captures the location of206mobile phone users,recorded every two hours for an entire week.In both datasets the spatial resolution is determined by the local density of the more than104mobile towers,registering movement only when the user moves between areas serviced by different towers.The average service area of each tower is approximately3km2 and over30%of the towers cover an area of1km2or less.To explore the statistical properties of the population’s mobility patterns we measured the dis-tance between user’s positions at consecutive calls,capturing16,264,308displacements for the D1and10,407displacements for the D2datasets.Wefind that the distribution of displacements over all users is well approximated by a truncated power-lawP(∆r)=(∆r+∆r0)−βexp(−∆r/κ),(1)withβ=1.75±0.15,∆r0=1.5km and cutoff valuesκ|D1=400km,andκ|D2=80km(Fig.1c,see the SM for statistical validation).Note that the observed scaling exponent is not far fromβB=1.59observed in Ref.[7]for bank note dispersal,suggesting that the two distributions may capture the same fundamental mechanism driving human mobility patterns.Equation(1)suggests that human motion follows a truncated L´e vyflight[7].Yet,the observed shape of P(∆r)could be explained by three distinct hypotheses:A.Each individual follows a L´e vy trajectory with jump size distribution given by(1).B.The observed distribution captures a population based heterogeneity,corresponding to the inherent differences between individuals.C.A population based heterogeneity coexists with individual L´e vy trajectories,hence(1)represents a convolution of hypothesis A and B.To distinguish between hypotheses A,B and C we calculated the radius of gyration for each user(see Methods),interpreted as the typical distance traveled by user a when observed up to time t(Fig.1b).Next,we determined the radius of gyration distribution P(r g)by calculating r g for all users in samples D1and D2,finding that they also can be approximated with a truncated power-lawP(r g)=(r g+r0g)−βr exp(−r g/κ),(2) with r0g=5.8km,βr=1.65±0.15andκ=350km(Fig.1d,see SM for statistical validation). L´e vyflights are characterized by a high degree of intrinsic heterogeneity,raising the possibility that(2)could emerge from an ensemble of identical agents,each following a L´e vy trajectory. Therefore,we determined P(r g)for an ensemble of agents following a Random Walk(RW), L´e vy-Flight(LF)or Truncated L´e vy-Flight(T LF)(Figure1d)[8,12,13].Wefind that an en-semble of L´e vy agents display a significant degree of heterogeneity in r g,yet is not sufficient to explain the truncated power law distribution P(r g)exhibited by the mobile phone users.Taken together,Figs.1c and d suggest that the difference in the range of typical mobility patterns of indi-viduals(r g)has a strong impact on the truncated L´e vy behavior seen in(1),ruling out hypothesis A.If individual trajectories are described by a LF or T LF,then the radius of gyration should increase in time as r g(t)∼t3/(2+β)[21,22]while for a RW r g(t)∼t1/2.That is,the longer we observe a user,the higher the chances that she/he will travel to areas not visited before.To check the validity of these predictions we measured the time dependence of the radius of gyration for users whose gyration radius would be considered small(r g(T)≤3km),medium(20<r g(T)≤30km)or large(r g(T)>100km)at the end of our observation period(T=6months).Theresults indicate that the time dependence of the average radius of gyration of mobile phone users is better approximated by a logarithmic increase,not only a manifestly slower dependence than the one predicted by a power law,but one that may appear similar to a saturation process(Fig.2a and Fig.S4).In Fig.2b,we have chosen users with similar asymptotic r g(T)after T=6months,and measured the jump size distribution P(∆r|r g)for each group.As the inset of Fig.2b shows,users with small r g travel mostly over small distances,whereas those with large r g tend to display a combination of many small and a few larger jump sizes.Once we rescale the distributions with r g(Fig.2b),wefind that the data collapses into a single curve,suggesting that a single jump size distribution characterizes all users,independent of their r g.This indicates that P(∆r|r g)∼r−αg F(∆r/r g),whereα≈1.2±0.1and F(x)is an r g independent function with asymptotic behavior F(x<1)∼x−αand rapidly decreasing for x≫1.Therefore the travel patterns of individual users may be approximated by a L´e vyflight up to a distance characterized by r g. Most important,however,is the fact that the individual trajectories are bounded beyond r g,thus large displacements which are the source of the distinct and anomalous nature of L´e vyflights, are statistically absent.To understand the relationship between the different exponents,we note that the measured probability distributions are related by P(∆r)= ∞0P(∆r|r g)P(r g)dr g,whichsuggests(see SM)that up to the leading order we haveβ=βr+α−1,consistent,within error bars, with the measured exponents.This indicates that the observed jump size distribution P(∆r)is in fact the convolution between the statistics of individual trajectories P(∆r g|r g)and the population heterogeneity P(r g),consistent with hypothesis C.To uncover the mechanism stabilizing r g we measured the return probability for each indi-vidual F pt(t)[22],defined as the probability that a user returns to the position where it was first observed after t hours(Fig.2c).For a two dimensional random walk F pt(t)should follow ∼1/(t ln(t)2)[22].In contrast,wefind that the return probability is characterized by several peaks at24h,48h,and72h,capturing a strong tendency of humans to return to locations they visited before,describing the recurrence and temporal periodicity inherent to human mobility[23,24].To explore if individuals return to the same location over and over,we ranked each location based on the number of times an individual was recorded in its vicinity,such that a location with L=3represents the third most visited location for the selected individual.Wefind that the probability offinding a user at a location with a given rank L is well approximated by P(L)∼1/L, independent of the number of locations visited by the user(Fig.2d).Therefore people devote mostof their time to a few locations,while spending their remaining time in5to50places,visited with diminished regularity.Therefore,the observed logarithmic saturation of r g(t)is rooted in the high degree of regularity in their daily travel patterns,captured by the high return probabilities(Fig.2b) to a few highly frequented locations(Fig.2d).An important quantity for modeling human mobility patterns is the probabilityΦa(x,y)tofind an individual a in a given position(x,y).As it is evident from Fig.1b,individuals live and travel in different regions,yet each user can be assigned to a well defined area,defined by home and workplace,where she or he can be found most of the time.We can compare the trajectories of different users by diagonalizing each trajectory’s inertia tensor,providing the probability offinding a user in a given position(see Fig.3a)in the user’s intrinsic reference frame(see SM for the details).A striking feature ofΦ(x,y)is its prominent spatial anisotropy in this intrinsic reference frame(note the different scales in Fig3a),and wefind that the larger an individual’s r g the more pronounced is this anisotropy.To quantify this effect we defined the anisotropy ratio S≡σy/σx, whereσx andσy represent the standard deviation of the trajectory measured in the user’s intrinsic reference frame(see SM).Wefind that S decreases monotonically with r g(Fig.3c),being well approximated with S∼r−ηg,forη≈0.12.Given the small value of the scaling exponent,other functional forms may offer an equally goodfit,thus mechanistic models are required to identify if this represents a true scaling law,or only a reasonable approximation to the data.To compare the trajectories of different users we remove the individual anisotropies,rescal-ing each user trajectory with its respectiveσx andσy.The rescaled˜Φ(x/σx,y/σy)distribution (Fig.3b)is similar for groups of users with considerably different r g,i.e.,after the anisotropy and the r g dependence is removed all individuals appear to follow the same universal˜Φ(˜x,˜y)prob-ability distribution.This is particularly evident in Fig.3d,where we show the cross section of ˜Φ(x/σ,0)for the three groups of users,finding that apart from the noise in the data the curves xare indistinguishable.Taken together,our results suggest that the L´e vy statistics observed in bank note measurements capture a convolution of the population heterogeneity(2)and the motion of individual users.Indi-viduals display significant regularity,as they return to a few highly frequented locations,like home or work.This regularity does not apply to the bank notes:a bill always follows the trajectory of its current owner,i.e.dollar bills diffuse,but humans do not.The fact that individual trajectories are characterized by the same r g-independent two dimen-sional probability distribution˜Φ(x/σx,y/σy)suggests that key statistical characteristics of indi-vidual trajectories are largely indistinguishable after rescaling.Therefore,our results establish the basic ingredients of realistic agent based models,requiring us to place users in number propor-tional with the population density of a given region and assign each user an r g taken from the observed P(r g)ing the predicted anisotropic rescaling,combined with the density function˜Φ(x,y),whose shape is provided as Table1in the SM,we can obtain the likelihood offinding a user in any location.Given the known correlations between spatial proximity and social links,our results could help quantify the role of space in network development and evolu-tion[25,26,27,28,29]and improve our understanding of diffusion processes[8,30].We thank D.Brockmann,T.Geisel,J.Park,S.Redner,Z.Toroczkai and P.Wang for discus-sions and comments on the manuscript.This work was supported by the James S.McDonnell Foundation21st Century Initiative in Studying Complex Systems,the National Science Founda-tion within the DDDAS(CNS-0540348),ITR(DMR-0426737)and IIS-0513650programs,and the U.S.Office of Naval Research Award N00014-07-C.Data analysis was performed on the Notre Dame Biocomplexity Cluster supported in part by NSF MRI Grant No.DBI-0420980.C.A.Hi-dalgo acknowledges support from the Kellogg Institute at Notre Dame.Supplementary Information is linked to the online version of the paper at /nature.Author Information Correspondence and requests for materials should be addressed to A.-L.B.(e-mail:alb@)[1]Horner,M.W.&O’Kelly,M.E.S Embedding economies of scale concepts for hub networks design.Journal of Transportation Geography9,255-265(2001).[2]Kitamura,R.,Chen,C.,Pendyala,R.M.&Narayaran,R.Micro-simulation of daily activity-travelpatterns for travel demand forecasting.Transportation27,25-51(2000).[3]Colizza,V.,Barrat,A.,Barth´e l´e my,M.,Valleron,A.-J.&Vespignani,A.Modeling the WorldwideSpread of Pandemic Influenza:Baseline Case and Containment Interventions.PLoS Medicine4,095-0110(2007).[4]Eubank,S.,Guclu,H.,Kumar,V.S.A.,Marathe,M.V.,Srinivasan,A.,Toroczkai,Z.&Wang,N.Controlling Epidemics in Realistic Urban Social Networks.Nature429,180(2004).[5]Hufnagel,L.,Brockmann,D.&Geisel,T.Forecast and control of epidemics in a globalized world.Proceedings of the National Academy of Sciences of the United States of America101,15124-15129 (2004).[6]Kleinberg,J.The wireless epidemic.Nature449,287-288(2007).[7] D.Brockmann,D.,Hufnagel,L.&Geisel,T.The scaling laws of human travel.Nature439,462-465(2006).[8]Havlin,S.&ben-Avraham,D.Diffusion in Disordered Media.Advances in Physics51,187-292(2002).[9]Viswanathan,G.M.,Afanasyev,V.,Buldyrev,S.V.,Murphy,E.J.,Prince,P.A.&Stanley,H.E.L´e vyFlight Search Patterns of Wandering Albatrosses.Nature381,413-415(1996).[10]Ramos-Fernandez,G.,Mateos,J.L.,Miramontes,O.,Cocho,G.,Larralde,H.&Ayala-Orozco,B.,L´e vy walk patterns in the foraging movements of spider monkeys(Ateles geoffroyi).Behavioral ecol-ogy and Sociobiology55,223-230(2004).[11]Sims D.W.et al.Scaling laws of marine predator search behaviour.Nature451,1098-1102(2008).[12]Klafter,J.,Shlesinger,M.F.&Zumofen,G.Beyond Brownian Motion.Physics Today49,33-39(1996).[13]Mantegna,R.N.&Stanley,H.E.Stochastic Process with Ultraslow Convergence to a Gaussian:TheTruncated L´e vy Flight.Physical Review Letters73,2946-2949(1994).[14]Edwards,A.M.,Phillips,R.A.,Watkins,N.W.,Freeman,M.P.,Murphy,E.J.,Afanasyev,V.,Buldyrev,S.V.,da Luz,M.G.E.,Raposo,E.P.,Stanley,H.E.&Viswanathan,G.M.Revisiting L´e vyflightsearch patterns of wandering albatrosses,bumblebees and deer.Nature449,1044-1049(2007). [15]Sohn,T.,Varshavsky,A.,LaMarca,A.,Chen,M.Y.,Choudhury,T.,Smith,I.,Consolvo,S.,High-tower,J.,Griswold,W.G.&de Lara,E.Lecture Notes in Computer Sciences:Proc.8th International Conference UbiComp2006.(Springer,Berlin,2006).[16]Onnela,J.-P.,Saram¨a ki,J.,Hyv¨o nen,J.,Szab´o,G.,Lazer,D.,Kaski,K.,Kert´e sz,K.&Barab´a si A.L.Structure and tie strengths in mobile communication networks.Proceedings of the National Academy of Sciences of the United States of America104,7332-7336(2007).[17]Gonz´a lez,M.C.&Barab´a si,plex networks:From data to models.Nature Physics3,224-225(2007).[18]Palla,G.,Barab´a si,A.-L.&Vicsek,T.Quantifying social group evolution.Nature446,664-667(2007).[19]Hidalgo C.A.&Rodriguez-Sickert C.The dynamics of a mobile phone network.Physica A387,3017-30224.[20]Barab´a si,A.-L.The origin of bursts and heavy tails in human dynamics.Nature435,207-211(2005).[21]Hughes,B.D.Random Walks and Random Environments.(Oxford University Press,USA,1995).[22]Redner,S.A Guide to First-Passage Processes.(Cambridge University Press,UK,2001).[23]Schlich,R.&Axhausen,K.W.Habitual travel behaviour:Evidence from a six-week travel diary.Transportation30,13-36(2003).[24]Eagle,N.&Pentland,A.Eigenbehaviours:Identifying Structure in Routine.submitted to BehavioralEcology and Sociobiology(2007).[25]Yook,S.-H.,Jeong,H.&Barab´a si A.L.Modeling the Internet’s large-scale topology.Proceedings ofthe Nat’l Academy of Sciences99,13382-13386(2002).[26]Caldarelli,G.Scale-Free Networks:Complex Webs in Nature and Technology.(Oxford UniversityPress,USA,2007).[27]Dorogovtsev,S.N.&Mendes,J.F.F.Evolution of Networks:From Biological Nets to the Internet andWWW.(Oxford University Press,USA,2003).[28]Song C.M.,Havlin S.&Makse H.A.Self-similarity of complex networks.Nature433,392-395(2005).[29]Gonz´a lez,M.C.,Lind,P.G.&Herrmann,H.J.A system of mobile agents to model social networks.Physical Review Letters96,088702(2006).[30]Cecconi,F.,Marsili,M.,Banavar,J.R.&Maritan,A.Diffusion,peer pressure,and tailed distributions.Physical Review Letters89,088102(2002).FIG.1:Basic human mobility patterns.a,Week-long trajectory of40mobile phone users indicate that most individuals travel only over short distances,but a few regularly move over hundreds of kilometers. Panel b,displays the detailed trajectory of a single user.The different phone towers are shown as green dots,and the V oronoi lattice in grey marks the approximate reception area of each tower.The dataset studied by us records only the identity of the closest tower to a mobile user,thus we can not identify the position of a user within a V oronoi cell.The trajectory of the user shown in b is constructed from186 two hourly reports,during which the user visited a total of12different locations(tower vicinities).Among these,the user is found96and67occasions in the two most preferred locations,the frequency of visits for each location being shown as a vertical bar.The circle represents the radius of gyration centered in the trajectory’s center of mass.c,Probability density function P(∆r)of travel distances obtained for the two studied datasets D1and D2.The solid line indicates a truncated power law whose parameters are provided in the text(see Eq.1).d,The distribution P(r g)of the radius of gyration measured for the users, where r g(T)was measured after T=6months of observation.The solid line represent a similar truncated power lawfit(see Eq.2).The dotted,dashed and dot-dashed curves show P(r g)obtained from the standard null models(RW,LF and T LF),where for the T LF we used the same step size distribution as the onemeasured for the mobile phone users.FIG.2:The bounded nature of human trajectories.a,Radius of gyration, r g(t) vs time for mobile phone users separated in three groups according to theirfinal r g(T),where T=6months.The black curves correspond to the analytical predictions for the random walk models,increasing in time as r g(t) |LF,T LF∼t3/2+β(solid),and r g(t) |RW∼t0.5(dotted).The dashed curves corresponding to a logarithmicfit of the form A+B ln(t),where A and B depend on r g.b,Probability density function of individual travel distances P(∆r|r g)for users with r g=4,10,40,100and200km.As the inset shows,each group displays a quite different P(∆r|r g)distribution.After rescaling the distance and the distribution with r g(main panel),the different curves collapse.The solid line(power law)is shown as a guide to the eye.c,Return probability distribution,F pt(t).The prominent peaks capture the tendency of humans to regularly return to the locations they visited before,in contrast with the smooth asymptotic behavior∼1/(t ln(t)2)(solid line)predicted for random walks.d,A Zipf plot showing the frequency of visiting different locations.The symbols correspond to users that have been observed to visit n L=5,10,30,and50different locations.Denoting with(L)the rank of the location listed in the order of the visit frequency,the data is well approximated by R(L)∼L−1. The inset is the same plot in linear scale,illustrating that40%of the time individuals are found at theirfirsttwo preferred locations.FIG.3:The shape of human trajectories.a,The probability density functionΦ(x,y)offinding a mobile phone user in a location(x,y)in the user’s intrinsic reference frame(see SM for details).The three plots, from left to right,were generated for10,000users with:r g≤3,20<r g≤30and r g>100km.The trajectories become more anisotropic as r g increases.b,After scaling each position withσx andσy theresulting˜Φ(x/σx,y/σy)has approximately the same shape for each group.c,The change in the shape of Φ(x,y)can be quantified calculating the isotropy ratio S≡σy/σx as a function of r g,which decreases as S∼r−0.12(solid line).Error bars represent the standard error.d,˜Φ(x/σx,0)representing the x-axis cross gsection of the rescaled distribution˜Φ(x/σx,y/σy)shown in b.。
csc重点资助专业列表CSC重点资助专业列表作为中国学生资助中心（CSC）的重点资助专业列表，以下是一些受到重点支持和鼓励的专业方向。
这些专业旨在培养具有创新能力和国际竞争力的高级人才，为国家的发展和进步做出贡献。
1. 人工智能与机器学习专业人工智能和机器学习作为当今科技领域的热门方向，正在以惊人的速度改变着我们的生活和工作方式。
在这个专业中，学生将学习人工智能的理论和应用，包括机器学习算法、深度学习、自然语言处理等。
他们将通过项目实践和实验室研究，开发创新的人工智能解决方案，应对现实世界的挑战。
2. 生物医学工程专业生物医学工程是交叉学科，融合了生物学、医学和工程学的知识和技术。
在这个专业中，学生将学习如何设计和开发医疗设备、生物材料以及医学影像技术。
他们将通过临床实践和研究，为改善人类健康和医疗服务做出贡献。
3. 新能源与可持续发展专业新能源和可持续发展是解决能源和环境问题的关键领域。
在这个专业中，学生将学习如何开发和利用可再生能源，包括太阳能、风能、水能等。
他们将研究新能源技术的创新和应用，为减少碳排放、保护环境做出贡献。
4. 数据科学与大数据技术专业数据科学和大数据技术是面向未来的重要领域。
在这个专业中，学生将学习如何处理和分析大规模数据，挖掘其中的价值和见解。
他们将掌握数据挖掘、机器学习、数据可视化等技术，为企业和社会提供智能决策和解决方案。
5. 人文与社会科学研究专业人文与社会科学研究是培养具有人文关怀和社会责任感的学者和研究人员的专业。
在这个专业中，学生将研究人类的文化、社会和心理现象，探索人类的发展和变化。
他们将通过深入的研究和实地调查，为社会问题的解决和人类发展的推动做出贡献。
以上只是CSC重点资助专业列表的一部分，还有许多其他的专业方向值得关注和探索。
CSC将继续支持和鼓励各个领域的创新和发展，为培养具有国际竞争力的高级人才做出贡献。
无论是选择哪个专业，都希望学生们能够充分发挥自己的才能和潜力，为国家和社会的发展做出积极的贡献。
人力资源管理参考文献20211. 引言人力资源管理（HRM）是企业管理中至关重要的一个方面，它涉及到招聘、培训、绩效评估、员工福利等多个领域。
随着时代的变迁和科技的发展，人力资源管理也在不断演变和改进。
本文将介绍一些关于人力资源管理的2021年度参考文献，为相关行业从业者提供借鉴和参考。
2. 资源管理2.1 Yoon, S., Tseng, M. M., Raines, S. C., & Kim, T. Y. (2021). What can we learn from strategic human resources management research in East Asia? Journal of World Business, 56(6), 101189.这篇研究考察了东亚地区人力资源管理的创新和实践，提供了在东亚市场中应用的参考指南。
它分析了人力资源管理活动对企业绩效的影响，以及区域文化和领导风格对人力资源管理的影响。
2.2 Qiao, W., & Wang, H. (2021). Modeling a green human resource management system using artificial intelligence and internet of things. Journal of Cleaner Production, 282, 124570.这篇论文研究了运用人工智能和物联网技术建立绿色人力资源管理系统的模型。
它探索了如何通过智能化的方法提高组织的可持续发展，并减少不必要的资源浪费。
3. 招聘与选拔3.1 Anderson, N., Lievens, F., van Dam, K., & Born, M. (2021). Significant Selection Factors—Evidence for the Validity of the SOSIE-R. European Journal of Psychological Assessment, 37(3), 446-455.这项研究验证了SOSIE-R（意大利语：Se。
闵行区2019 学年度第二学期高三年级质量调研考试高三英语考生注意：1. 考试时间120 分钟，试卷满分140 分。
2. 本考试设试卷和答题纸两部分, 试卷共12 页。
所有答题必须涂（选择题）或写（非选择题）在答题纸上, 做在试卷上一律不得分。
3. 答题前, 务必在答题纸上填写学校、姓名和考生号。
I. Listening Comprehension Section ADirections: In Section A, you will hear ten short conversations between two speakers. At the end of each conversation, a question will be asked about what was said. The conversations and the questions will be spoken only once. After you hear a conversation and the question about it, read the four possible answers on your paper, and decide which one is the best answer to the question you have heard.1. A. A cook. B. A dentist. C. A surgeon. D. A technician .2. A. She ’d like to have the windows open. B. She likes to have the air conditioner on .C. The air is heavily polluted .D. The windows are always open.3. A. Art attraction. B. Painting skills.C. Furniture quality.D. Room decoration.4. A. She appreciates the man’s effort. B. She does enjoy spicy food.C. She is annoyed with the man.D. She doesn’t like the food the man prepared .5. A. They can’t fit into the machine. B. They were sent to the wrong stress.C. They haven’t been delivered yet .D. They were found to be of the wrong type .6. A. The long waiting time. B. The broken down computer.C. The mistakes in her telephone bill.D. The bad telephone service.7. A. Its quality. B. Its price C. Its materials D. Its appearance.8. A. Walk in the countryside. B. Do some exercise.C. Go shopping.D. Wash some dresses.9. A. He is going to visit a photo studio. B. He has just had his picture taken.C. He is one the way to the theatre.D. He has just returned from a job interview.10. A. He doesn’t want Jenny to get into trouble.B. He doesn’t agree with the woman’s remark.C. He thinks Jenny’s workload too heavy at college.D. He believes most college students are running wild.Section BDirections: In Section B, you will hear two short passages and one longer conversation, and you will be asked several questions on each of the passages and the conversation. The passages and the conversation will be read twice, but the questions will be spoken only once. When you hear a question, read the four possible answers on your paper and decide which one would be the best answer to the question you have heard.Questions 11 through 13 are based on the following passage.11. A. Attend expert growers’ lecture. B. Visit fruit-lovig families.C. Plant fruit trees in an orchard(果园).D. Taste many kinds of apples.12. A. It is a new variety. B. It is the best variety for eating quality.C. It is rarely seen now.D. It needs perfect soil to grow.13 A. To show how to grow apples. B. To introduce an apple festival .C. To help people select apples.D. To attract more people to visit BritainQuestions 14 through 16 are based on the following news.14. A. Lack of sleep could lead to heath problem.B. Lack of sleep affects adults more than children.C. Sleeping problems are one of the leading causes of obesity.D. The amount of sleep people need changes with age.15. A. Less sleep is needed when they enter adolescence.B. Ideally, they need 8 hours of sleep a night.C. They may have difficulty in falling asleep before 11:00 pm .D. They always wake up at midnight due to biological changes .16. A. The amount of sleep and academic performance.B. A comparison of different time to start school .C. Students’ sleeping time and social behaviors.D. A beneficial change of school starting time.Questions 17 through 20 are based on the following conversation.17. A. Extreme sports . B. City life.C. Human’s potential.D. Danger and safety.18. A. They like to stay isolated . B. They prefer danger to safety.C. They want to know their potential.D. They are bored with the traditional ones.19. A. Objective. B. Negative. C. Positive. D. Neutral.20. A. It is interesting and challenging . B. It is dangerous and needs no skills.C. It enables people to face fears.D. It only stimulate individualism.II. Grammar and vocabulary Section ADirections: After reading the passage below, fill in the blanks to make the passage coherent and grammatically correct. For the blanks with a given word, fill in each blank with the proper form of the given word; for the other blanks, use one word that best fits each blank.A brief history of Chopsticks.We’ve discussed the story of knife and fork, but there’s another set of utensils(器皿) used by billions of people around the world—and it has a truly ancient past. The Chinese have been wielding chopsticks since at least 1200 B.C., and by A.D. 500 chopsticks ___1___ (sweep) the Asian continent from Vietnam to Japan. From their humble beginnings ___2___ cooking utensils to paper-wrapped bamboo sets at the sushi counter, there's more to chopsticks than meets the eye.Chopsticks ___3___ (develop) about 5000 years ago in China. The ___4___ (early) versions were probably twigs used to get food from cooking pots. When resources became scare, around 400 B. C. ,crafty chefs figured out ___5___ to conserve fuel by chopping food into smaller pieces that required less cooking fuel, and besides, it could be cooked more quickly. This new method of cooking made it unnecessary to have knives at the dinner table—a practice that also matched the non-violent teachings of Confucius ( 孔子)，___6___ expressed in one of his numerous quotations:“ th e honorable and upright man keeps well away from both the slaughterhouse(屠宰房) and the kitchen. And he allows no knives on his table.” As a vegetarian, he believed that knives’ sharp points evoked( 诱发) violence ___7___ (kill) the happy, contented mood that should reign during meals. Thanks in part to his teachings, chopstick use quickly became widespread throughout Asia.Different cultures adopted different chopstick styles. Perhaps in a nod to Confucius, Chinese chopsticks featured a blunt rather than pointed end. In Japan, chopsticks were 8 inches long for men and 7 inches long for women. In 1878 the Japanese became the first ___8___ (create) the now-ubiquitous disposable set, typically made of bamboo or wood. Wealthy diners could eat with ivory, jade, coral, brass or agate versions, while the most privileged used silver sets. It was believed that the silver would corrode and turn black ___9___ it came into contact with poisoned food.Throughout history, chopsticks have enjoyed a symbiotic relationship with another staple of Asian cuisine: rice.At first glance, you'd think that rice wouldn't make the cut, but in Asia most rice is of the short- or medium-grain variety. The starches(淀粉质食品) in these rices create a cooked product that is gummy and clumpy, unlike the fluffy and distinct grains of Western long- grain rice. ___10___ chopsticks come together to lift steaming bundles of sticky rice, it's a match made in heaven.【答案】1. had swept2. as3. were developed4. earliest5. how6. as7. killing8. to create9. if/when10. As/Because/Since【解析】这是一篇说明文。
发展基于“语义检测”的低参数量、多模态预训练电池通用人工智能模型吴思远;李泓【期刊名称】《储能科学与技术》【年(卷),期】2024(13)4【摘要】ChatGPT的出现意味着一种以“预训练+微调”为主的新科研范式诞生,以OpenAI为代表的企业正朝着训练通用人工智能(AGI)模型的路线前进,AGI意味着人工智能具备超越人类智力并解决通用性问题的能力,其是为了解决通用问题并具有强大的自学能力来促进人类社会发展。
然而OpenAI等模型仍然是以文本为主结合图像等作为输入,对于电池体系来说,文本信息是少数的,更多的是温度、电压-电流曲线等的多模态数据,其所关注的结果包括电池荷电态、电池健康度、剩余寿命和是否出现电池性能跳水的拐点,甚至包括无数据情况下电池二次(梯度)利用的健康度评估。
这意味着ChatGPT的路线虽然也可能解决电池体系的问题,但是以文本为主的样式或许有些“杀鸡用牛刀”,即使未来OpenAI的AGI可能解决当前电池存在的问题,但是在模型参数和输入方面与电池本质不符会使得模型参数量巨大而不适合电池离线端评估。
对于电池体系的AGI,应该有自己独特的“文本语言”即理解电池运行过程中所发生的一切物理、化学过程以及其之间的关联,从而实现通用性并为后续全固态电池量产上车做铺垫。
本文展望了在电池体系发展AGI过程中应该重新设计模型架构,特别在特征表示、数据结构设计、预训练方法、预训练过程设计和实际任务微调等需要重新设计。
此外,相较于运行在服务器端的大模型,发展低参数量特别是离线的模型对于实时预测和基于我国国情及国际形势发展是十分必要的。
本文主要讨论了发展基于“语义检测”的低参数量、多模态预训练电池通用人工智能模型所需要经历的几个阶段、可能面临的困难和评价指标,同时给出中国科学院物理研究所(以下简称物理所)在电池大模型在预训练、微调和测评三个方面“三步走”计划中的规划和可能线路。
【总页数】9页(P1216-1224)【作者】吴思远;李泓【作者单位】中国科学院物理研究所北京清洁能源前沿研究中心;中国科学院物理研究所凝聚态物质科学数据中心【正文语种】中文【中图分类】TP391.77;TP391.9【相关文献】1.基于预训练模型和联合调参的改进训练算法2.基于双预训练Transformer和交叉注意力的多模态谣言检测3.基于预训练和多模态融合的假新闻检测4.基于预训练模型和编码器的图文跨模态检索算法因版权原因，仅展示原文概要，查看原文内容请购买。
Hybrid Tracking of Human Operators using IMU/UWB DataFusion by a Kalman FilterJ. A. Corrales, F. A. Candelas and F. TorresAutomatics, Robotics and Computer Vision Group.Physics, Systems Engineering and Signal Theory Department, University of Alicante.03690 San Vicente del Raspeig, Alicante, Spain.+34 96 590 94 91[jcorrales, Francisco.Candelas, Fernando.Torres]@ua.esABSTRACTThe precise localization of human operators in robotic workplaces is an important requirement to be satisfied in order to develop human-robot interaction tasks. Human tracking provides not only safety for human operators, but also context information for intelligent human-robot collaboration. This paper evaluates an inertial motion capture system which registers full-body movements of an user in a robotic manipulator workplace. However, the presence of errors in the global translational measurements returned by this system has led to the need of using another localization system, based on Ultra-WideBand (UWB) technology. A Kalman filter fusion algorithm which combines the measurements of these systems is developed. This algorithm unifies the advantages of both technologies: high data rates from the motion capture system and global translational precision from the UWB localization system. The developed hybrid system not only tracks the movements of all limbs of the user as previous motion capture systems, but is also able to position precisely the user in the environment.Categories and Subject DescriptorsI.2.9 [Artificial Intelligence]: Robotics – operator interfaces, sensors.General TermsAlgorithms, Measurement, Experimentation, Human Factors. KeywordsMotion capture, inertial sensors, UWB, human tracking and monitoring, indoor location, data fusion, Kalman filter. 1.INTRODUCTIONCollaboration between human beings and robots takes advantage of their complementary features. On one hand, robots are able to do repetitive tasks which are dangerous or exhausting for people. On the other hand, humans can perform complex tasks which require intelligence. However, human-robot interaction in industrial workplaces may be dangerous for human operators because of the possibility of collisions with robots or with heavy objects. Therefore, a precise localization of human operators is needed. Nevertheless, most localization systems only register global position and orientation of the person. Motion capture systems don’t have this drawback because they are able to measure full-body movements.Motion Capture (MoCap) is a technique for digitally recording the movements of a person. The user of the system wears markers (or sensors) near each joint of his body and the motion capture system calculates his movements by comparing the positions and angles between the markers. Although the first motion capture systems were used in biomechanics [2] to study and model the movements of the human body, MoCap is applied in a wide variety of application fields nowadays: animation and computer graphics [5], robot teleoperation [6], human-robot interaction [7]…The ability of motion capture systems to track precisely all the limbs of the human body not only guarantees the security of operators in industrial environments, but also provides significant information about their behaviour. This context-awareness provided by motion capture systems is the first requirement to be satisfied in order to apply ‘Ambient Intelligence’ (AmI) [1] in industrial environments where there is human-robot interaction. Context information can be interpreted in order to create ‘intelligent environments’ which adapt their operation to the user’s needs. Thus, the robot workplace becomes context-aware and the robot controller is able to change the robot movements depending on the position of the human operator.This paper analyses an inertial motion capture system which is used to track human operators who interact with a robotic manipulator in an industrial environment. Although the rotational measurements for each joint are very precise, the global translation position returned by this system accumulates high errors through time. Due to this fact, an additional UWB localization system, which provides more precise position measurements, has been used. In section 2 of this paper, differentmotion capture technologies are compared for their use in industrial workplaces and previous work on similar hybrid tracking systems is described. In section 3, both tracking systems are presented and evaluated. The Kalman Filter (KF) algorithm designed to combine the position data of both systems is explained in section 4. Section 5 describes some experimental results of this filter. Finally, the conclusions of this paper and future research are presented in section 6.2.BACKGROUND2.1Motion Capture TechnologiesSeveral motion capture technologies have been developed in the last decade [13]. However, not all of them are appropriate for industrial applications. Each one has different advantages and limitations depending on the physical principles on which it relies.Mechanical MoCap systems (such as Gypsy5 from Animazoo) are composed of a set of articulated rigid segments (exoskeleton) attached to the user’s limbs and interconnected between them through electromechanical transducers (such as potentiometers). User motion is registered through voltage variations in the potentiometers. Although these systems have accurate measurements and low latencies, they are very uncomfortable for industrial workers who have to wear the exoskeleton for many hours a day.Magnetic MoCap systems (such as MotionStar from Ascension) use a permanent transmitter (a set of three coils) that induces magnetic fields in the environment. These magnetic fields are measured by small receivers attached to the user’s body and so user location is estimated. These systems are accurate and don’t have light-of-sight restrictions. Nevertheless, they are not convenient for industrial workplaces because electronic devices and ferrous metals can change the electromagnetic fields induced by the transmitter and thus distort receiver measurements.Optical MoCap systems (such as Vicon MX from Vicon) are based on the installation of a set of calibrated cameras which record images of the markers attached to the actor’s body. The position of each marker is triangulated by using three or more images that contain the marker. Orientation is deduced from the relative orientation between three or more markers. These systems have high accuracy and high sampling rate, which enables quick movement registration. However, these systems are very complex to install in an industrial environment because they require many calibrated cameras in order to avoid marker occlusion (line-of-sight restrictions).Inertial MoCap systems (such as GypsyGyro-18 from Animazoo) use inertial sensors: accelerometers and gyroscopes. These sensors are tied to the actor’s body. Actor’s limb positions are calculated by double integrating the accelerations obtained from accelerometers and orientations are calculated by integrating the angular rates obtained from gyroscopes. These systems are the most appropriate for industrial environments because of their numerous advantages. Inertial sensors have low latencies, can be sampled at high rates, are self-contained and easy to install (no emitters are needed in the environment). Furthermore, they don’t have line-of-sight restrictions as optical systems. However, the main disadvantage of these systems is the accumulation of errors through time (drift). This problem has been solved by developing hybrid systems which combine inertial measurements with other sensors.2.2Previous Work on Hybrid TrackingPose (position and orientation) estimation by inertial sensors is a well-studied field with applications in vehicle navigation, augmented reality (AR) and robotics. Most systems use additional sensors in order to correct drift accumulation of inertial sensors. Kalman filtering is the most frequently used technique for combining measurements of different sensors in this hybrid tracking systems.Foxlin [4] develops a small device for head-tracking in virtual environments. It is composed by three orthogonal angular rate gyroscopes, a two-axis inclinometer and a two-axis compass. This system implements a complementary Kalman filter which estimates errors in orientation (from the inclinometer and the compass) and angular rate (from the gyros).This device is an example of an inertial measurement unit (IMU):a package of inertial sensors (gyroscopes and accelerometers) which are used for tracking purposes by dead-reckoning estimation. Recent advances in sensor miniaturization have enabled the creation of small-sized compact MEMS (Micro-Electro-Mechanical Systems) IMUs, which are usually composedby gyroscopes, accelerometers and magnetometers. Accelerometers and magnetometers measurements are used to correct drift of orientation measurements from gyroscopes. MTx from Xsens and InertiaCube3 from Intersense are two examplesof commercial MEMS IMUs.Recent inertial tracking systems use these commercial MEMS IMUs because they return precise orientation and don’t have the typical drift problems of stand-alone gyros. However, these IMUs only obtain orientation measurements and no position informationis supplied. An additional localization system is needed in orderto obtain position measurements.Caron et al. [3] propose a multisensor Kalman filter which combines GPS and IMU data to localize an autonomous land vehicle (ALV). This Kalman filter has two different measurements models (one for each sensor type) which are weighted according to fuzzy context variables that define sensors data reliability.Ribo et al. [8, 9] present a wearable AR system that is mounted ona helmet. It consists of a real-time 3D visualization subsystem (composed by a stereo see-through HMD) and a real-time tracking subsystem (composed by a camera and an IMU). Sensor fusion is accomplished by an extended Kalman filter (EKF) which uses inertial measurements during the prediction step and vision-based measurements during the correction step.Roetenberg et al. [10] have designed a wearable human motion tracking system consisting of a magnetic tracker (composed by a magnetic source and three magnetic sensors) and an inertial tracker (composed by three IMUs). The magnetic tracker is ableto calculate relative distances and orientations between body segments while the inertial tracker registers accelerations and angular rates. A complementary Kalman filter is developed to correct inertial measurements with the magnetic ones.Finally, Vlasic et al. [11] present a motion capture system which fuses accelerometers, gyroscopes and acoustic sensors by an EKF. The ultrasonic subsystem provides relative distances betweensensors. However, this MoCap system and the previous one don’t obtain absolute localization of users in the environment because they only return relative measurements. The MoCap system presented in this paper not only tracks all the limbs of the body but also gets the global position of the user in the environment.3. SYSTEM ARCHITECTUREThe industrial environment built in this work project has three main devices (Figure 1): a Mitsubishi PA-10 robotic arm, an Animazoo GypsyGyro-18 MoCap system and an UWB localization system from Ubisense.Figure 1. Main components of the industrial workplace: a PA-10 robotic arm (left) and a human operator (right), who wearsa GypsyGyro-18 MoCap suit and an Ubisense tag. The PA-10 is an industrial robotic arm (or manipulator) normally used for pick-and-place applications and component assembly. The robot controller is connected to a PC and can be controlled with a software library. The two tracking systems are described in the following sections.3.1 Inertial Motion Capture SystemThe GypsyGyro-18 is an inertial motion capture system composed of 18 small IMUs attached to a lycra suit which is worn by a human operator (Figure 1). Each IMU is an InertiaCube3 from Intersense which measures the orientation (roll, pitch and yaw) of the operator’s limb to which it is attached. This orientation data is transmitted through wireless link to a controller PC where global position of the operator is estimated with a footstep extrapolation algorithm.All this movement data (limbs orientations and global body translation) is represented on a 3D hierarchical skeleton structure (Figure 2) whose size corresponds to the limbs’ lengths. The hips node is the root node of the skeleton and represents global translation and rotation of the whole body in the environment. The other nodes of the skeleton represent limbs’ rotations registered by IMUs. The rotation of each node is relative to the coordinate system of the parent node in the hierarchy of theskeleton.Figure 2. GypsyGyro-18 hierarchical skeleton.The GypsyGyro-18 datasheet points out that orientation measurements calculated from each IMU have a resolution lower than 1º. Some experiments have verified these accuracy values, which are sufficient for general industrial manipulation tasks. Nevertheless, the accuracy of the global translation measurements estimated by the footstep algorithm is not specified. A set of experiments have been developed in order to quantify this accuracy. These experiments involve comparing the actual displacement of a person at different distances (200, 300 and 400 cm) with the displacement obtained from the MoCap system. Six trials have been executed for each distance and their results are shown in Table 1:Table 1. GypsyGyro-18 translational error statistics (in cm). Distance (cm)Minimum errorMaximum errorMean errorStandard Deviation200 16.70 66.04 40.10 17.92 300 15.33 69.54 37.92 20.97 400 35.43 64.23 51.09 10.67These errors are very high for industrial purposes. In some cases, they represent more than 30% of the actual distance. For this reason, an additional localization system (based on UWB technology) is needed in order to correct the translational errors of the GypsyGyro-18.3.2 UWB Localization SystemAn Ultra-WideBand (UWB) radio positioning system from Ubisense is used to obtain more accurate translation information of the human operator.The Ubisense system consists of two kinds of hardware devices: sensors and tags (Figure 3). Four sensors are situated at fixes positions on the localization area. Tags are small devices, of similar size to a credit card, which are carried by the users. A tag sends UWB pulses to the sensors, which use a combination of TDOA (Time-Difference of Arrival) and AOA (Angle of Arrival)PA-10Ubisense TagGypsyGyrosuittechniques to estimate the global position (3D coordinates) of the user carrying the tag.Sensors are connected to an Ethernet switch and send the location information to a controller PC which can access to this data through a software library. Slave sensors are also connected to amaster sensor for synchronization in the TDOA algorithm.Figure 3. Ubisense architecture.UWB is an appropriate technology for human positioning because of the following advantages:- Immunity to multipath fading: UWB signals are much less susceptible to this effect than conventional RF (Radio-Frequency) technologies because UWB receivers are able to differentiate the original pulses from the reflected/refracted ones owning to their small time duration.- Immunity to RF interferences: UWB signals have low power values which allow them to coexist with other RF signals despite their large bandwidth. Thereby, UWB systems have higher accuracy (15-30cm) than other RF technologies (1-3m).- Reduced infrastructure: The number of sensors to be installed in the workplace is small. Other technologies (e.g. ultrasound) require denser sensor installations.- No line-of-sight restrictions: In optical localization systems, there shouldn’t be any obstacle (occlusion) between the emitter and the receiver. UWB technology doesn’t have this limitation.4. FUSION OF POSITION MEASURES4.1 Algorithm MotivationThe first strategy for correcting the translational errors of the GypsyGyro-18 would be to substitute its location data with the coordinates returned by the Ubisense system. However, this strategy is not suitable because the Ubisense has small data frequency (5-9fps) which will cause extremely high latencies for industrial environments. In addition, the Ubisense software applies a filter to the data which reduces the number of measurements, involving a variable data frequency. On the other hand, the GypsyGyro-18 supplies constant data rates (30-120fps), but with accumulated errors in global translational measurements. For these reasons, the best solution is to combine the global translational measurements from both systems. This fusion will correct the defects of one system with the advantages of the other system: The GypsyGyro-18 will supply a high data frequencywhile the Ubisense will correct the accumulated errors with its location measurements. The GypsyGyro-18 rotational measurements for each joint (obtained from IMUs) will remain unchanged because they are accurate relative rotation transformations in the skeleton (Figure 2).4.2 Coordinate TransformationThe first step to combine the global position measurements of both tracking systems is to represent them in the same coordinate system (Figure 4). The Ubisense frame U is a fixed coordinate system in the workplace while the GypsyGyro-18 frame G is determined every time the system is initialized. The Ubisense frame is selected as the reference frame and thus allmeasurements will be completely located in the environment.Figure 4. Ubisense and GypsyGyro-18 coordinate frames. As shown in Figure 4, between the GypsyGyro-18 frame and the Ubisense frame there is only a translation and a rotation about the Z axis by α. Therefore, the following equation will be used to transform a point p from the GypsyGyro coordinate system G p to the Ubisense coordinate system U p :()(),,,U U G U U UU G G G G G x y z α=⋅=⋅⋅p T p Trans Rot z p (1)Expanding the previous expression, the following equation will beobtained:()()()()cos sin 0sin cos 0001100011U UG GU U G GU U G G x x x y y y z z z αααα⎡⎤⎡⎤⎡⎤−⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥=⋅⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎢⎥⎣⎦⎣⎦⎣⎦(2) The angle α will be a known constant parameter (angle between the north direction and the Y axis of the Ubisense system). The only unknown variables of the transformation matrix U G T are the three coordinates of the translation vector from the Ubisense frame to the GypsyGyro frame. They can be calculated from equation 1 by substituting two corresponding measurements: ()()cos sin U U G G G x x x y αα=−+ (3) ()()sin cos U U G G G y y x y αα=−− (4)UU G G z z z =− (5)After obtaining the transformation matrix U G T , all thetranslational measurements from the GypsyGyro-18 (3D position of the hips node) will be transformed from the GypsyGyro-18 frame to the Ubisense frame by equation 1.4.3 Kalman Filter Fusion AlgorithmGlobal translational measurements from both trackers are combined by a fusion algorithm based on a standard Kalman filter (Figure 5). First of all, the transformation matrix U G T will be initialized with equations 3-5 and the first two measurements. Thereby, the following measurements form the GypsyGyro-18 will be transformed to the Ubisense coordinate system with equation 1.Figure 5. Fusion algorithm diagram.After representing all measurements in the same coordinate system, a Kalman filter will be applied to them. The Kalman filter is a recursive stochastic technique which estimates the state n ∈ℜx of a dynamic system from a set of incomplete and noisy measurements [12]. The dynamic system is modeled by the following state-transition equation at time k (state model):111k k k k −−−=++x Ax Bu w (6)In equation 6, A is a n n × state-transition matrix, B is a n p ×matrix, u is a 1p × vector with system inputs and w is a 1n ×process noise vector (zero mean multivariate normal distribution with covariance matrix k Q ). In the current work, the state vector x is composed by the coordinates (),,x y z =p of the global position of the user in the environment. A is a 3x3 identity matrix in order to incorporate directly the GypsyGyro-18 measurements and B is a null matrix because there are no control inputs. The process noise covariance matrix Q is a diagonal matrix because state vector variables are not correlated. The diagonal terms of this matrix correspond to the mean error of the GypsyGyro-18 measurements.Sensor measurements m ∈ℜz at time k are modeled in a KF by the following equation (measurement model):k k k =+z Hx v (7)In equation 7, H is a m n × observation matrix which represents how the state of the system is registered by sensors and v is a 1m × measurement noise vector (zero mean multivariate normal distribution with covariance matrix k R ). In this paper, H is a 3x3 identity matrix and R is a diagonal matrix whose diagonal terms correspond to the mean error of the Ubisense measurements.The Kalman filter algorithm is composed of two steps: prediction step and correction step. The prediction step obtains a-prioriestimate ˆk −p (equation 8) of the global position of the user and apriori estimate of the error covariance matrix k −P (equation 9) by incorporating position measurements k p from the GypsyGyro-18:1ˆk k k −−=+p Ap Bu (8)1T k k −−=+P AP A Q (9)On the other hand, the correction step uses Ubisense measurements k z in order to eliminate error accumulation in previous a-priori estimates and thus compute an improved a-posteriori estimate of the global position ˆk p (equation 11) and the error covariance k P (equation 12): ()1T T k k k −−−=+K P H HP H R (10)()ˆˆˆk k k k k −−=+−p p K z Hp (11)()k k k −=−P I K H P (12)Finally, the transformation matrix U G T is re-calculated with the position estimate ˆk p. Thereby, the prediction step will be executed with the GypsyGyro-18 rate and the correction step will be executed with the Ubisense rate.Although the Ubisense system usually returns accurate positions, some measurements from the Ubisense system have big errors and shouldn’t be incorporated to the Kalman filter. These outliers are eliminated by a filter which verifies that the distance between the current position and the previous one does not involve an excessive velocity for a person walking.5. EXPERIMENTAL RESULTS A set of experiments has been performed to verify the developedfusion algorithm. A human operator wearing the GypsyGyro-18 suit and an Ubisense tag has walked along a preestablished linearpath in the industrial workplace described in section 3. Themeasurements from both trackers are registered by a Visual C++ program which is running in the controller PC where the GypsyGyro-18 and Ubisense software libraries (DLLs) are installed. All these data are combined by the Kalman filter fusion algorithm which has been implemented in Matlab in order to obtain graphical representation of the resulting measurements. Figure 6 shows the global translational measurements returned by the GypsyGyro-18 and the Ubisense systems in a linear path in theXY plane. They are represented in the Ubisense coordinate system. It is also represented the predefined linear path that the human operator has followed. The trajectory obtained from the GypsyGyro-18 presents an error of 0.56m with regard to the preestablished path. This error is produced by the GypsyGyro-18 footstep extrapolation algorithm because it sometimes estimates wrongly when the feet come into contact with the floor. This translational error justifies the need of the UWB system. However, the frequency of measurements from the Ubisense system (6Hz) is highly lower than the GypsyGyro-18 data rate (30Hz), as it is shown in Figure 6. Because of this fact, the fusion of both systems is used in order to combine their complementary features.2.852.92.9533.053.13.153.2X (m)Y (m )Figure 6. Comparison between the planned path andmeasurements from the GypsyGyro and the Ubisense systems.Figure 7 shows the trajectory estimated by the Kalman filter fusion algorithm implemented in this paper. The GypsyGyro-18 global translational error has been reduced to 0.14m and the resulting data rate is equal to the GypsyGyro-18 frequency (30Hz).Figure 7. XY trajectory from the fusion algorithm.Figure 8 represents the position estimates of the fusion algorithm that are obtained during the prediction step and the ones that are obtained during the correction step of the developed Kalman filter. GypsyGyro-18 measurements are used in the prediction step in order to fill in the gaps between Ubisense measurements and thus obtain a more detailed trajectory. Ubisense measurements are used in the correction step in order to improve the accuracy of the global translation measurements from the GypsyGyro-18. The current experiment has been performed several times and the obtained results have been very similar to the ones described above.Figure 8. Prediction and correction estimates from the KF.An additional experiment has been developed in order to verify the integration between the global translation data from the fusion algorithm and the limbs orientation measurements from the MoCap system. It is an interaction task between the PA-10 manipulator and a human operator who is wearing the GypsyGyro-18 suit and an Ubisense tag. The operator picks up an object which is out of the robot’s workplace and gives it to the manipulator (Figure 9). Limbs orientations from the GypsyGyro-18 and global displacement in the environment are shown over a skeleton in a 3D rendering software (Figure 9). The obtained animation shows that limbs movements are very accurate and the global localization of the human operator is appropriate.6. CONCLUSIONSIn this paper, a hybrid tracking system for localizing precisely a human operator in a robotic workplace is developed. It consists of two components: an inertial motion capture system (GypsyGyro-18) and an UWB localization system (Ubisense ). The MoCap system is able to register the movements of the operator’s limbs with high precision. Nevertheless, the global position of the operator in the environment is not determined with sufficient accuracy. Thereby, an UWB localization system is used in order to obtain precise position measurements. A fusion algorithm based on a Kalman filter has been implemented in order to combine global position measurements of both systems. This fusion algorithm joins the advantages of the MoCap system (high data rate and accurate rotational data of each limb) and the UWB system (accurate global position estimation).Figure 9. Frame sequence of a human-robot interaction task where a human operator gives an object to the robot.The main advantage of this hybrid tracking system over previous systems (see section 2.2) is the combination of the global position of the operator in the environment with the precise location of all his limbs. Thereby, the operator is completely localized in the workplace. The precision of the system guarantees the security of the operator and allows the development of intelligent interactive tasks with robots.In future work, more complex interaction tasks which take into account spatial relationships between the robot and the skeleton of the operator will be developed.7.ACKNOWLEDGMENTSThis work is funded by the Spanish Ministry of Education and Science through the research project ‘Design, Implementation and Experimentation of Intelligent Manipulation Scenarios for Automatic Assembly and Disassembly Applications’ (DPI2005-06222) and the pre-doctoral grant AP2005-1458. The Valencian Government has also funded this research through the project ‘infraestructura05/053’.8.REFERENCES[1] Alcañiz, M., and Rey, B. 2005. New Technologies ForAmbient Intelligence. In Ambient Intelligence. IOS Press,Amsterdam, 3-15.[2] Andriacchi, T.P., and Alexander, E.J. 2000. Studies ofHuman Locomotion: Past, Present and Future. Journal ofBiomechanics, 33, 10, 1217-1224. [3] Caron, F., Duflos, E., Pomorski, D., and Vanheeghe, P. 2006.GPS/IMU Data Fusion Using Multisensor Kalman Filtering: Introduction of Contextual Aspects. Information Fusion, 7, 2, 221-230.[4] Foxlin, E. 1996. Inertial Head-Tracker Sensor Fusion by aComplementary Separate-Bias Kalman Filter. In Proceedings of the IEEE Virtual Reality Annual InternationalSymposium, (Santa Clara, California, 1996). IEEE Computer Society, 185-194.[5] Izani, M., Eshaq, A.R., Razak, A., and Norhan, N. 2004. AStudy on Practical Approach of Using Motion Capture andKeyframe Animation Techniques. In Proceedings of theInternational Conference on Computer Graphics, Imagingand Visualization, (Malaysia, 2004). IEEE ComputerSociety, 52-55.[6] Miller, N., Jenkins, O.C., Kallmann, M., and Mataric, M.J.2004. Motion Capture from Inertial Sensing for UntetheredHumanoid Teleoperation. In Proceedings of the 4thIEEE/RAS International Conference on Humanoid Robots,(Santa Monica, 2004). IEEE Robotics and AutomationSociety, 547-565.[7] Onishi, M., Odashima, T., Luo, Z., and Hosoe, S. 2006. AnImmersion-type 3D Dynamic Simulation Environment forDeveloping Human Interactive Robot Systems. Systems and Computers in Japan, 37, 10, 47-57.。
