对比文件名称:US2017031361A1_Description_20251231_1723
目标专利名称:用于车辆的消息广播
本次调用的模型名称:DeepSeek最新版本模型
特征比对表格
| 技术特征描述及公开性判断结果 | 对比文件原文引用 | 公开性论述 | |
| 技术特征A:一种控制车辆的方法。《直接公开》 | [0105] “As used herein, the adverb “substantially” means that a shape, structure, measurement, quantity, time, etc. may deviate from an exact described geometry, distance, measurement, quantity, time, etc., because of imperfections in materials, machining, manufacturing, etc.” <br> [0095] “FIG. 4 is a diagram of an exemplary process 400 for controlling a vehicle using multipolicy decision-making.” <br> [0096] “In the block 405, a host vehicle 14 computer 20 receives data representing a current environment in which the host vehicle 14 is currently operating.” <br> [0104] “In the block 435, the computer 20 sends the selected policy 46 to the vehicle control unit 34. The vehicle control unit 34, based at least in part on the policy 46, issues commands to one or more controllers 26 to control the driving behavior of the vehicle 14.” | 对比文件明确公开了一种控制车辆的方法。其详细描述了通过计算机20执行过程400来控制主机车辆14(参见段落[0095], [0096])。该方法包括接收环境数据、分析、决策(选择策略46),并最终向车辆控制单元34发送命令以控制车辆的驾驶行为(参见段落[0104])。这与目标专利权利要求中的“一种控制车辆的方法”这一上位概念完全相同,属于方法的直接公开。 | |
| 技术特征B:接收包括用于发送车辆的运动规划的意图消息。《隐含公开》 | [0096] “The computer 20 may further receive data regarding one or more non-host vehicles 16 travelling within a predefined distance to the host vehicle 14. The data may include the current position of each non-host vehicle 16, and other data such as speed, direction of travel etc. as described above.” <br> [0032] “Collected data may further include, e.g., information such as a type (e.g., light truck, passenger car, minivan, etc.), dimensions, make, model, etc. of the one or more of the non-host vehicles 16.” <br> [0022] “The host vehicle 14 is programmed to define a traffic environment 10 which includes the host vehicle 14 and one or more non-host vehicles 16 within a predefined distance to the host vehicle 14.” | 对比文件未明确使用“意图消息”这一术语。然而,对比文件反复提及主机车辆14接收关于非主机车辆16的数据(参见段落[0096]),这些数据包括位置、速度、行驶方向、车辆类型、尺寸等信息(参见段落[0032])。在自动驾驶领域,车辆通过V2X或其他传感器感知方式获取周围车辆的状态和动态信息是常见技术手段。接收到的关于非主机车辆(即发送车辆)的这类数据集合,实质上隐含了其“运动规划”或意图信息,因为位置、速度、方向等是构成运动规划(轨迹)的基础要素。因此,本领域技术人员可以从对比文件中明确,主机车辆接收了包含发送车辆(非主机车辆)运动相关信息的数据,这隐含公开了“接收包括用于发送车辆的运动规划的意图消息”这一特征。 | |
| 技术特征C:其中,所述运动规划包括所述发送车辆的轨迹以及与所述发送车辆相关联的一个或多个车辆描述符,并且其中,所述一个或多个车辆描述符包括车辆物理能力,其中,所述车辆物理能力包括转弯半径、车辆最高速度和车辆最大加速度中的至少一项。《隐含公开》 | [0010] “In the system, a non-zero is a tuple of the pose, velocity, and acceleration, and an action at v is a tuple of controls for steering, throttle, brake, shifter, and directionals. Pose may include a position of a vehicle and the orientation (roll, pitch, yaw). Pose may further include one or more of velocity, acceleration and rotation rate.” <br> [0020] “Let Π be a discrete set of policies, where each policy captures a specific high-level driving behavior including a planned trajectory.” <br> [0032] “Collected data may further include, e.g., information such as a type (e.g., light truck, passenger car, minivan, etc.), dimensions, make, model, etc. of the one or more of the non-host vehicles 16.” <br> [0044] “Each policy 46 is designed to capture a different high-level behavior and intention, and may include a planned trajectory.” | 对比文件未直接提及“车辆描述符”或明确列举“转弯半径、车辆最高速度、车辆最大加速度”。然而,对比文件明确公开了“策略”包含规划的轨迹(参见段落[0020], [0044])。同时,对比文件描述了系统状态包含位姿(位置和朝向)、速度、加速度(参见段落[0010]),并且收集的数据包括车辆类型、尺寸等信息(参见段落[0032])。车辆的类型、尺寸等信息直接关联并隐含了其物理能力,例如不同车型(如轿车、卡车)具有不同的转弯半径、最高速度和加速度能力,这是本领域技术人员的公知常识。因此,基于对比文件公开的车辆类型、尺寸信息以及包含速度、加速度的运动状态,本领域技术人员可以合理推断出这些信息共同构成了与发送车辆相关联的、反映其车辆物理能力的描述符,并隐含在所述的运动规划(策略)信息中。 | |
| 技术特征D:对所述意图消息进行解析以识别用于所述发送车辆的所述运动规划。《隐含公开》 | [0096] “The computer 20 may further receive data regarding one or more non-host vehicles 16...” <br> [0100] “In the block 420, as described in detail above, the computer 20 analyzes the past behavior of the non-host vehicles 16 using Bayesian change-point analysis. Based on the change-point analysis, the computer 20 identifies a distribution of likely policies 46 for each of the non-host vehicles 16 at the current timestep.” <br> [0058] “After identifying the one or more non-host vehicles 16 in the traffic environment 10, the computer 20 may be programmed to collect data regarding the driving behavior of the non-host vehicles 16 during a preceding time period.” | 对比文件未明确使用“解析意图消息”的表述。但是,对比文件明确描述了主机车辆的计算机20接收关于非主机车辆的数据(参见段落[0096]),并对其过去的行为数据进行分析(例如使用贝叶斯变点分析),以识别出每个非主机车辆在当前时刻可能遵循的策略(政策)分布(参见段落[0100])。这个“分析”过程,实质上就是对接收到的、反映非主机车辆状态和意图的数据进行处理和解读,以识别出用于该发送车辆的未来行为规划(即策略,包含轨迹)。因此,“对数据进行处理分析以识别策略”隐含公开了“对意图消息进行解析以识别运动规划”这一技术特征。 | |
| 技术特征E:以及至少部分地基于所述运动规划来控制所述车辆,包括:至少部分地基于由所述车辆物理能力限制的所述发送车辆的车辆行为模型,来设置针对所述发送车辆的行为预测。《隐含公开》 | [0100] “...the computer 20 identifies a distribution of likely policies 46 for each of the non-host vehicles 16 at the current timestep.” <br> [0077] “The process begins by drawing a set of samples sεS from the distribution over policies 46 of the non-host vehicles 16 via Eq. 16, where each sample assigns a policy 46 πv εΠ to each nearby non-host vehicle 16 v.” <br> [0078] “For each policy 46 π available to the host vehicle 14 and for each sample s, the computer 20 rolls out forward in time until a decision horizon H the host and non-host vehicles 14, 16 under the policy assignments (π,s) with closed-loop simulation to yield a set Ψ of simulated trajectories v.” <br> [0082] “Each policy 46 reward rπ for the host vehicle 14 is evaluated as the expected reward over all rollout rewards rπ,s...” <br> [0103] “In the block 430, the computer 20 evaluates the results of the multiple simulations based on a set of rewards established for the traffic environment 10. The computer 20 selects a policy 46 to govern the host vehicle 46 in order to maximize the rewards for the traffic environment 10 as described above.” | 对比文件完整公开了“至少部分地基于运动规划来控制车辆”的核心思想。主机车辆基于识别出的非主机车辆的可能策略(运动规划)分布(参见段落[0100]),通过采样(参见段落[0077])和闭环前向模拟(参见段落[0078])来预测交通环境中所有车辆(包括发送车辆)的未来行为,并评估不同主机策略下的结果(参见段落[0082]),最终选择最优策略来控制主机车辆(参见段落[0103])。这个模拟和预测过程,本质就是基于对非主机车辆运动规划(策略)的理解来设置和预测其行为。同时,如前所述,车辆物理能力(如类型、尺寸)隐含在收集的数据中,并且在模拟车辆行为时,本领域技术人员必然需要考虑车辆的物理能力(如最大加速度决定其加速快慢,转弯半径决定其路径曲率)来构建合理的行为模型进行预测。因此,对比文件隐含公开了基于(隐含的)车辆物理能力限制的行为模型来设置针对发送车辆的行为预测,并据此控制主机车辆。 | |
| 技术特征F:所述一个或多个车辆描述符进一步包括传感器可感知属性。《隐含公开》 | [0032] “Collected data may further include, e.g., information such as a type (e.g., light truck, passenger car, minivan, etc.), dimensions, make, model, etc. of the one or more of the non-host vehicles 16.” <br> [0038] “Data collectors 24 may include a variety of devices. For example, data collectors 24 may include lidar, radar, video cameras, ultrasonic sensors, infrared sensors for sensing the environment, and for example, tracking non-host vehicles 16.” | 对比文件明确列举了收集的非主机车辆数据包括类型、尺寸、品牌、型号等(参见段落[0032])。这些属性,如车辆类型(轿车/卡车)、尺寸(大小),正是可以通过车载传感器(如摄像头、激光雷达、雷达,参见段落[0038])直接感知到的物理特征。因此,“类型、尺寸”等信息属于“传感器可感知属性”的具体示例。本领域技术人员可以直接且毫无疑义地从对比文件公开的“类型、尺寸”等内容中确定,其所描述的车辆描述符包含了传感器可感知属性。 | |
| 技术特征G:至少部分地基于所述运动规划来控制所述车辆包括:至少部分地基于所述运动规划来确定用于所述车辆的预期感兴趣区域;以及将检测算法应用于在所述预期感兴趣区域处的接收的传感器数据,以至少部分地基于所述传感器可感知属性来在所述接收的传感器数据中检测所述发送车辆。《隐含公开》 | [0022] “The host vehicle 14 is programmed to define a traffic environment 10 which includes the host vehicle 14 and one or more non-host vehicles 16 within a predefined distance to the host vehicle 14.” <br> [0100] “...the computer 20 identifies a distribution of likely policies 46 for each of the non-host vehicles 16 at the current timestep.” <br> [0078] “...the computer 20 rolls out forward in time... under the policy assignments (π,s) with closed-loop simulation to yield a set Ψ of simulated trajectories v.” <br> [0038] “Data collectors 24 may include... video cameras... for sensing the environment, and for example, tracking non-host vehicles 16.” | 对比文件虽未明确使用“预期感兴趣区域”这一术语,但其技术方案逻辑上隐含了该概念。主机车辆首先定义了一个包含附近非主机车辆的交通环境(即一个广义的感兴趣区域,参见段落[0022])。更重要的是,通过识别非主机车辆的可能策略(运动规划,参见段落[0100])并进行前向模拟(参见段落[0078]),主机车辆能够预测非主机车辆未来的轨迹和位置。在感知层面,为了高效跟踪非主机车辆(参见段落[0038]),本领域技术人员容易想到,可以基于预测的轨迹(运动规划)将传感器关注的重点区域(即“预期感兴趣区域”)缩小到车辆未来最可能出现的空间范围,而不是监控整个环境。同时,结合已知的车辆传感器可感知属性(如类型、尺寸),可以更有针对性地在传感器数据中检测和识别该特定车辆。这是一种合乎逻辑的技术应用。因此,该特征被对比文件隐含公开。 | |
| 技术特征H:至少部分地基于所接收的运动规划来选择所述检测算法。《未公开》 | 对比文件未提及根据运动规划选择或切换不同的检测算法。 | 对比文件描述了使用传感器(如摄像头)跟踪车辆(参见段落[0038]),并基于策略进行行为预测和模拟(参见段落[0078]),但全文没有记载或暗示会根据接收到的其他车辆的运动规划信息,来主动选择或调整用于检测该车辆的特定算法(例如,针对不同运动状态使用不同的图像识别算法)。该特征在对比文件中既无文字记载,也无法从已公开的内容中合理推断得出。 | |
| 技术特征I:至少部分地基于所述运动规划来控制所述车辆包括:至少部分地基于所述传感器可感知属性来将车辆检测传感器数据与所述发送车辆进行关联。《隐含公开》 | [0032] “Collected data may further include, e.g., information such as a type (e.g., light truck, passenger car, minivan, etc.), dimensions, make, model, etc. of the one or more of the non-host vehicles 16.” <br> [0077] “The process begins by drawing a set of samples sεS from the distribution over policies 46 of the non-host vehicles 16 via Eq. 16, where each sample assigns a policy 46 πv εΠ to each nearby non-host vehicle 16 v.” <br> [0069] “Specifically, we assume p(zτ m+1:n \ | π,θ) to be a multivariate Gaussian with mean at the trajectory ψπ,θ obtained by simulating forward in time the execution of policy 46 π under parameters θ from timestep τm +1:” | 对比文件公开了收集非主机车辆的属性信息,如类型、尺寸(传感器可感知属性,参见段落[0032])。同时,系统通过分析为每个非主机车辆分配可能的行为策略(运动规划,参见段落[0077]),并将观测到的车辆状态与策略模拟出的轨迹进行比对(参见段落[0069])。在传感器数据关联(即确定哪个传感器检测对应哪辆已知车辆)的过程中,利用已知的车辆属性(如尺寸、类型)作为关联依据,是本领域多目标跟踪和传感器融合中的常规技术手段。结合对比文件已公开的“拥有车辆属性信息”和“为车辆分配运动规划”这两个事实,本领域技术人员能够合理推断,在实现车辆跟踪和数据关联时,必然会利用已知的传感器可感知属性来辅助将检测到的传感器数据与特定的发送车辆(非主机车辆)正确关联起来。因此,该特征被隐含公开。 |
| 技术特征J:确定所述发送车辆的行为是否符合所述行为预测;以及响应于确定所述发送车辆的所述行为不符合所述行为预测,至少部分地基于所述车辆物理能力来更新所述行为预测。《隐含公开》 | [0069] “Specifically, we assume p(zτ m+1:n \ | π,θ) to be a multivariate Gaussian with mean at the trajectory ψπ,θ obtained by simulating forward in time the execution of policy 46 π under parameters θ from timestep τm +1:” <br> [0071] “Anomalous behavior of a non-host vehicle 16 not modeled by available policies 46 may be detected based on the time-series segmentation obtained via change-point detection.” <br> [0076] “This normality measure on the target vehicle's history can then be compared to that of a set of previously recorded trajectories of other non-host vehicles 16.” <br> [0100] “...the computer 20 identifies a distribution of likely policies 46 for each of the non-host vehicles 16 at the current timestep.” | 对比文件明确公开了异常行为检测(参见段落[0071]),其通过比较观测到的车辆历史轨迹与策略模拟轨迹(参见段落[0069])或与正常行为记录(参见段落[0076])来判断行为是否“正常”或符合预期。当检测到异常行为(即不符合基于策略的行为预测)时,系统需要处理。虽然未明确描述“更新行为预测”的步骤,但在一个实时、闭环的决策系统中(如段落[0100]所述在每个时间步识别策略分布),如果发现车辆行为偏离预测,为了后续决策的准确性,更新对该车辆的认知(包括其可能的行为预测)是必然且合理的后续步骤。同时,如前所述,车辆物理能力是隐含的已知信息,在更新行为模型或预测时,将其作为约束条件(例如,车辆不可能以超出其物理极限的方式运动)是本领域技术人员的常规选择。因此,该特征被对比文件隐含公开。 |
| 技术特征K:所述一个或多个车辆描述符进一步包括车辆位置属性。《直接公开》 | [0096] “The data may include the current position of each non-host vehicle 16...” <br> [0010] “Pose may include a position of a vehicle and the orientation (roll, pitch, yaw).” | 对比文件明确且直接地记载了收集的数据包括非主机车辆的当前位置(参见段落[0096]),并且定义了位姿包含车辆的位置(参见段落[0010])。“位置”是典型的车辆位置属性。因此,对比文件直接公开了车辆描述符包括车辆位置属性。 | |
| 技术特征L:至少部分地基于所述运动规划来控制所述车辆包括:至少部分地基于所述车辆位置属性来确定所述发送车辆的位置;确定在所述车辆的位置与所述发送车辆的位置之间的比较是否指示误差;以及响应于确定在所述车辆的位置与所述发送车辆的位置之间的所述比较指示误差,来触发对所述车辆的位置的重新计算。《隐含公开》 | [0096] “The computer 20 may further receive data regarding one or more non-host vehicles 16... The data may include the current position of each non-host vehicle 16...” <br> [0056] “The computer 20 may further receive map data describing the area where the host vehicle 14 is currently driving.” <br> [0069] “Specifically, we assume p(zτ m+1:n \ | π,θ) to be a multivariate Gaussian with mean at the trajectory ψπ,θ obtained by simulating forward in time the execution of policy 46 π under parameters θ from timestep τm +1:” <br> [0076] “This normality measure on the target vehicle's history can then be compared to that of a set of previously recorded trajectories of other non-host vehicles 16.” | 对比文件公开了主机车辆接收非主机车辆的位置数据(参见段落[0096])和地图数据(参见段落[0056])。系统通过将观测到的车辆轨迹与基于策略模拟的轨迹进行比较来评估行为(参见段落[0069])或与历史正常轨迹比较(参见段落[0076])。虽然未明确记载“触发对主机车辆自身位置的重新计算”,但定位技术中常利用已知地标或外部参考来校正自身位置。当主机车辆基于其自身定位系统对非主机车辆位置的观测值,与接收到的非主机车辆自身报告的位置(基于其运动规划/位置属性)存在系统性误差时,本领域技术人员能够合理推断,这种误差可能源于主机车辆自身定位的偏差。在这种情况下,利用该误差作为触发条件,来启动对主机车辆自身定位的重新计算或优化,是一种合乎逻辑且可行的技术手段。因此,该特征被对比文件隐含公开。 |
| 技术特征M:至少部分地基于所述运动规划来控制所述车辆包括:确定所述运动规划是否是不安全的;以及响应于确定所述运动规划是不安全的,来向所述发送车辆发送安全警告。《未公开》 | [0081] “Typical metrics include... minimum distance to obstacles to evaluate safety...” <br> [0091] “cost-map integral: Given a cost map of obstacles in the environment (both directly observed and “virtual obstacles” denoting areas the host vehicle 14 should avoid, such as oncoming traffic), the cost for this metric is the line integral through the cost map. This cost gets larger as the host vehicle 14 get closer to obstacles, and minimizing it should improve safety.” | 对比文件公开了在评估策略(运动规划)时,会考虑安全性指标,如与障碍物的最小距离(参见段落[0081])和成本地图积分(参见段落[0091]),这隐含了评估运动规划安全性的过程。然而,对比文件完全没有提及,在主机车辆判定其他车辆(发送车辆)的运动规划不安全时,会向该发送车辆发送安全警告。对比文件描述的系统是一个自主决策系统,其核心是基于对环境的评估为“自身”选择安全策略,而没有与其他车辆进行主动安全协同或警告交互的记载或暗示。因此,“向发送车辆发送安全警告”这一技术特征在对比文件中既未直接公开,也无法从已公开的内容中合理推断得出。 |
<<<A>>><<<K>>>