【正文】 in uncertain environment, with limited sensor range, and multivehicle operations.翻譯中文 因?yàn)樗赡苋菀妆蛔R(shí)別， 算法略述在上面由來自均衡的跳躍組成指向均衡點(diǎn)， 并且因此不可能提供令人滿意的性能，就費(fèi)用而言(2)。 the final (equilibrium) conditions are characterized by a desired final position xf and zero velocity. In order to provide The minimum time maneuver from origin to destination for any of the degrees of freedom (assuming a general maximum control intensity ,is a bangbang control law ⒙ given by u(t)=U for 0 t t1u(t) =－U for t1 t t1+t2 (4)the sign of the initial control value U can be determined though the switching function (5)If the initial conditions are such that △0≧0thenU = －, and U = otherwise. The time length of the two bangbang segmentsdetermined as follows: (6) with The policy π used to control the vehicle described by (3) is then defined as follows: Considering the two degrees of freedom xl and x2, the slowest39。 again, this is done using the cost functional expressed in eq.(2). Realtime considerationsA significant issue arising from the usage of randomized algorithms for path planning is the distinct possibility of driving the system towards a deadend due to finite putation times. The notion of safety is introduced to prevent such situations to develop:Definition I (safety) A milestone is said to be safe if (x, t) is feasible for all.Primary milestones are added to the tree only if safe. If possible, and practical, primary milestones can be checked for the absence of collisions over an infinite horizon(). This is possible in many cases of interest (including all static environments), and results in hard safety guarantees on the resulting motion plan if the initial condition is itself safe. In cases in which safety cannot be ensured over an infinite time horizon,safety only ensures that the algorithm will always have at least r seconds to pute a new solution.Thesafety of the generated plan derives from the fact that all the primary milestones are by construction safe, and all secondary milestones have at least one primary milestones in their subtree. Maintaining safety guarantees in the face of finite putation times is particularly important, since the algorithm itself has no deterministic guarantees of success. In the sense outlined above the algorithm will always produce safe motion plans, even in the case in which a feasible trajectory to the target set has not been found. The time available for putation is bounded by either 0, or by the duration of the current trajectory segment. When the time is up, a new tree must be selected from the children of the current root. If there are none, since every primary milestone is τsafe, the system has at least τ seconds of guaranteed safety, available for puting a new tree (secondary milestones always have at least one child). If the current root has children, then two cases arise: At least one of the children leads to the destination through an already puted feasible solution. If there is more than one such feasible solution, the solution with the least, upper bound on the costtogo is chosen。在此，對(duì)崔世鋼老師在設(shè)計(jì)的整個(gè)過程中的精心指導(dǎo)和幫助，致以衷心的感謝！參 考 文 獻(xiàn)[1]、王文貴 編 深圳智能交通技術(shù)應(yīng)用與研究文集 [M]北京 人民交通出版社 2000 [2]、 (意) 王武宏等編譯 智能車輛 智能交通系統(tǒng)的關(guān)鍵技術(shù) [M]北京 人民交通出版社 2002[3]、中國公路學(xué)會(huì)《交通工程手冊(cè)》編委會(huì) 交通工程手冊(cè) [M] 北京 人民交通出版社 1996[4]、劉智勇. 智能交通控制理論及其應(yīng)用 [M]. 北京: 科學(xué)出版社 2003[5]、 常斗南 李全利 張學(xué)武 可編程序控制器原理、應(yīng)用、實(shí)驗(yàn)[M]機(jī)械工業(yè)出版社 1998年[6]、朱善君　可編程序控制系統(tǒng)原理．應(yīng)用．維護(hù)[M]　清華大學(xué)出版社　1992年[7]、 劉敏 可編程序控制器技術(shù)[M]機(jī)械工業(yè)出版社 2000年[8]、 李乃夫 可編程序控制器原理、應(yīng)用、實(shí)驗(yàn)[M]中國輕工業(yè)出版社 1998年[9]、 萬太福 可編程序控制器原理及應(yīng)用[M]重慶大學(xué)出版社 1994年[10]、 陳春雨 可編程序控制器應(yīng)用軟件設(shè)計(jì)方法與技巧[M]電子工業(yè)出版社 1992年附錄1：英文資料翻譯 Improving performanceAs it can be easily recognized, the algorithm outlined above consists of jumps from equilibrium point to equilibrium point, and as such is unlikely to provide satisfactory performance, in terms of the cost (2).However, performance may be restored by realizing that the available guidance policy may not only steer the vehicle from equilibrium state to equilibrium state, but from any state to an equilibrium state. This suggests introducing the following step: consider the tree at some point in time and a newly added milestone to the tree. A secondary milestone is defined to be any state of the system (continuous or hybrid) along the path leading from the parent node in the tree to the newly added milestone. Pick n ≥ such secondary milestones at random along that path. Because the vehicle is in motion along the path, these secondary milestones are likely to be at points in the state space that are far from the equilibrium manifold.These secondary milestones are added to the tree, and, as for all newly generated milestones, feasibility is checked for the resulting trajectory to the destination xf Moreover secondary milestones can be selected as the tree node to be expanded in later iterations. Note that all secondary milestones, by construction, have a primary milestone in a child subtree. Data structureThe roadmap is constructed as a tree, consisting of nudes and edges. At the tree nodes all the information concerning each milestone is stored, including: 1) The propagated state of the vehicle (. state , time ). 2) The cumulative cost and upper and lower bounds on the costtogo. The lower bound on the costtogo coincides with the value of the cost function , that is, it corresponds to the cost to go to the target state assuming the presence of no obstacle. The upper bound on the costtogo is initialized to +∞, meaning that a feasible path from the particular node has not been found yet. 3) A counter of the total number of milestones in the children trees. The (state time) couple is initialized through propagation of the system dynamics, and the cumulative cost is updated, according to eq.(2).At