Space-time evaluation method and system for motion trail of robot
The spatial and temporal characteristics of the robot's motion trajectory are described through signal timing logic, and the problem of lack of quantitative indicators in the existing technology is solved, and the spatial and fusion evaluation of the robot's motion trajectory is realized, which improves the evaluation accuracy and provides a theoretical basis for the optimization of the robot's autonomous navigation performance. .
Patent Information
- Application Number
- CN202410570835.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-09
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-05-09
AI Technical Summary
In the prior art, the evaluation results of robot motion trajectory lack specific level quantification indicators, and it is impossible to distinguish the advantages and disadvantages of the motion trajectory from time sequence characteristics and continuous signals, and it is impossible to fully and objectively judge the advantages and disadvantages of the motion trajectory of the robot when performing tasks.
Signal timing logic (STL) is used to describe the spatiotemporal characteristics of the robot's motion trajectory, and evaluate it from the two dimensions of timing indicators and spatial indicators. The optimal parameter model is fused through logical connectors to obtain the most accurate model, and grade division and description are performed.
The space-time fusion evaluation of robot motion trajectory is realized, which can fully reflect the advantages and disadvantages of the trajectory, improve the evaluation accuracy, and provide theoretical basis and practical means for the optimization of robot's autonomous navigation performance.
Smart Images

Figure CN118893618A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of robot evaluation technology, and in particular to a spatiotemporal evaluation method and system for robot motion trajectory. Background Art
[0002] With the rapid development of autonomous robot technology, the autonomous navigation ability of robots in the process of performing tasks has become a key indicator for evaluating their performance. In the field of autonomous robots, the measurement of their performance involves multiple complex links such as perception, decision-making, operation and movement. Among the many evaluation indicators, the motion trajectory analysis of robots is widely considered to be an important method to reflect the autonomous navigation performance.
[0003] Traditional methods mainly focus on how robots can reach their destination in the most efficient way, such as in the shortest time, the shortest distance, and with the least amount of energy consumed. However, they cannot continuously monitor the comprehensive performance of the robot's trajectory over the entire time. Continuous monitoring can not only help the team discover and correct errors in real time, optimize the robot's behavior and path planning, and improve its environmental perception capabilities, thereby improving the robot's overall performance in various tasks and environments, but also significantly increase its importance in the field of robot evaluation.
[0004] Signal temporal logic can be used to describe the continuous position information and timing information of the robot. Signal temporal logic (STL) is a formal language used to describe and analyze the behavior of a system in the time dimension. It allows users to define the time constraints and conditions of the system behavior in a mathematically rigorous way. In addition, the evaluation system in the prior art is mostly based on the general classification standards of the intelligence level of unmanned systems, but these classification standards do not provide a clear and quantifiable method for evaluating the quality of motion trajectories.
[0005] In the prior art, the invention patent with the patent publication number CN114489055A discloses a method, medium and device for realizing multi-task motion of robots based on temporal logic. Based on temporal logic, the article encodes the complex operation task specification into an interpretable wTLTL formula, which can quickly plan the motion trajectory of complex operation tasks with user preferences, with low computational complexity and high planning efficiency. The prior art is just an application of temporal logic in motion planning of robot systems, which is used for motion trajectory planning. Summary of the invention
[0006] The technical problem to be solved by the present invention is to solve the problem that the evaluation results of existing motion trajectories lack specific level quantitative indicators to describe, and the existing evaluation methods only evaluate through total distance and total duration, and cannot distinguish the quality of the motion trajectory in a certain time period from the timing characteristics and continuous signals, and cannot comprehensively and objectively judge the quality of the motion trajectory of the robot when performing a task.
[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0008] A spatiotemporal evaluation method for a robot motion trajectory comprises the following steps:
[0009] S100, dividing the evaluation level of the robot's motion trajectory into different levels using signal timing logic rules, and distinguishing and describing each level;
[0010] S200, evaluating the robot motion trajectory from two dimensions, namely, the timing index and the space index, respectively, to obtain the optimal timing-level STL parameter model for each level under the timing index, and to obtain the optimal space-level STL parameter model for each level under the space index;
[0011] S300, using a logical connector, the optimal time-series level STL parameter model and the optimal space-level STL parameter model at each level are combined and fused to obtain a plurality of level fused STL parameter models at each level;
[0012] S400, selecting the model with the highest accuracy from the multiple level fusion STL parameter models at each level as the optimal level fusion STL parameter model at this level;
[0013] S500, puts the same robot motion trajectory into the optimal level fusion STL parameter model at different levels to evaluate the robot motion trajectory; and classifies the evaluation results according to different levels to comprehensively reflect the advantages and disadvantages of the robot motion trajectory.
[0014] Advantages: The present invention uses STL to describe the continuous position information and timing information of the robot, and then constructs a spatiotemporal fusion evaluation method for the robot's motion trajectory, and classifies the evaluation results according to different levels, which can comprehensively reflect the advantages and disadvantages of the robot's motion trajectory.
[0015] In one embodiment of the present invention, the evaluation level of the robot motion trajectory is expressed by the following formula:
[0016] H i :STL[A i ,B i ](Y j ≤index≤Y k );
[0017] In the formula, H i is the i-th level of the robot motion trajectory, where i = 1, 2, 3, ..., n, and n is the level of the level; STL is the signal timing logic rule symbol, including "□", "◇", "∧" and "∨"; where "□" represents the global property, which requires that the set STL formula constraint must always be satisfied within the set range; "◇" represents the final property, which requires that the set STL formula constraint must be satisfied at a certain moment within the set range; "∧" represents logical AND, which means that two conditions must be satisfied at the same time; "∨" represents logical OR, which requires that at least one of the two conditions be satisfied; [A i ,B i ] is the level area; index is the index value, Y j and Y k They are the upper and lower limit values of the indicator respectively.
[0018] In one embodiment of the present invention, a sensor, a visual system or a simulator is used to obtain the timing index and the spatial index of the robot; wherein the timing index is represented by "projection distance - projection duration"; and the spatial index is represented by "xy" of the Cartesian coordinate system.
[0019] In one embodiment of the present invention, obtaining the optimal timing level STL parameter model of each level under the timing index in step S200 includes the following steps:
[0020] S211, define the timing level STL initial model;
[0021] S212, initializing the parameters of the timing level STL initial model;
[0022] S213, calculating the robustness: applying the robustness to characterize the deviation distance between the trajectory signal point and the trajectory that meets the timing level STL initial model;
[0023] S214, defining an incentive function: for each level, the training objective function of the time-series level STL initial model is a function with the smallest robustness calculation result value as a reward function;
[0024] S215, establishing a reinforcement learning model: using a reinforcement learning algorithm, establishing an initial time series level STL parameter model, taking the time series index as input, and taking the STL model parameters of the initial time series level STL parameter model as output;
[0025] S216, training and acquiring models: Use the data set to train the initial time-series level STL parameter model; in each training iteration, perform corresponding actions according to the current state and the output of the model, and observe the feedback from the environment; update and optimize the model parameters through the reward function to gradually learn and obtain the optimal time-series level STL parameter model for each level.
[0026] In one embodiment of the present invention, under the timing index, multiple robustness levels are obtained by the following formulas:
[0027]
[0028] Where x(t) is the monitoring signal at time t, l(x(t) is the evaluation result of the x signal at time t, l(x) is the timing evaluation result of the x signal, c is a constant, For different signal timing logic, It is expressed as the “delivery” of tasks under the timing indicator. It is expressed as the "and" of tasks under the timing indicator, and [a,b) is the time interval; among them, according to each level, there are multiple different timing level STL initial models, and each timing level STL initial model is calculated corresponding to different robustness; through the signal timing logic rule symbol, each timing level STL initial model is confirmed, and the corresponding robustness is calculated.
[0029] In one embodiment of the present invention, the step of obtaining the optimal spatial level STL parameter model for each level under the spatial index is the same as the step of obtaining the optimal temporal level STL parameter model for each level under the temporal index; the difference is that: multiple robustness under the spatial index is obtained by the following formula:
[0030]
[0031] Wherein, x(T) is the monitoring signal at position T, l(x(T) is the evaluation result of the x signal at position T, l(x) is the spatial evaluation result of the x signal, c is a constant, φ, φ1, φ2 are different signal space logics, φ1∧φ2 represents the "intersection" of the tasks under the spatial indicator, φ1∨φ2 represents the "and" of the tasks under the spatial indicator, and [a, b) is the position point interval; according to each level, there are multiple different spatial level STL initial models, and each spatial level STL initial model corresponds to a different robustness calculation; through the signal temporal logic rule symbol, each spatial level STL initial model is confirmed, and the corresponding robustness is calculated.
[0032] In one embodiment of the present invention, a reinforcement learning model is established under the spatial index, with the spatial index as input and the STL model parameters of the initial spatial level STL parameter model as output.
[0033] In one embodiment of the present invention, under the spatial index, the training set for obtaining the optimal time-level STL parameter model and the optimal space-level STL parameter model at each level includes a time-series training set and a space-series training set; and the time-series training set includes time-series training subsets for training at different levels, and the space-series training set includes space-series training subsets for training at different levels.
[0034] In one embodiment of the present invention, the logical connectors are an “and” connector and an “or” connector.
[0035] The present invention also provides a system based on the above-mentioned spatiotemporal evaluation method of robot motion trajectory, comprising:
[0036] The level division module is used to divide the evaluation level of the robot's motion trajectory into different levels using signal timing logic rules, and distinguish and describe each level;
[0037] The optimal parameter model module is used to evaluate the robot's motion trajectory from the two dimensions of timing indicators and spatial indicators, and obtain the optimal timing level STL parameter model for each level under the timing indicators, and obtain the optimal spatial level STL parameter model for each level under the spatial indicators;
[0038] The level fusion module is used to combine and fuse the optimal time-series level STL parameter model and the optimal space-level STL parameter model at each level using logical connectors to obtain multiple level fusion STL parameter models at each level;
[0039] The optimal level fusion module is used to select the model with the highest accuracy from multiple level fusion STL parameter models at each level as the optimal level fusion STL parameter model at this level;
[0040] The evaluation level classification module is used to put the same robot motion trajectory into the optimal level fusion STL parameter model at different levels to evaluate the robot motion trajectory; and classify the evaluation results according to different levels to comprehensively reflect the advantages and disadvantages of the robot motion trajectory.
[0041] Compared with the prior art, the beneficial effects of the present invention are as follows: a time-space (space-time) evaluation method based on the continuous trajectory signal of the robot is proposed, and the robot motion trajectory is identified by setting a quantifiable grade interval, so that the quality of the robot during continuous operation can be accurately quantified. Then, an STL formula is constructed to extract the robot trajectory characteristics from the two dimensions of timing index and spatial position, respectively, and a reinforcement learning algorithm is used to obtain the optimal spatial and temporal STL formula parameters, and the two formulas are connected to realize the fusion evaluation of time and space. The present invention not only improves the evaluation accuracy, but also provides a new theoretical basis and practical means for optimizing and improving the robot's autonomous navigation performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 This is a flow chart of a spatiotemporal evaluation method for a robot motion trajectory according to an embodiment of the present invention.
[0043] Figure 2 It is a distribution diagram of motion trajectory curves of different levels according to an embodiment of the present invention.
[0044] Figure 3 A schematic diagram of projection distance definition according to an embodiment of the present invention.
[0045] Figure 4a Schematic diagram of timing information evaluation according to an embodiment of the present invention.
[0046] Figure 4b Schematic diagram of position data evaluation according to an embodiment of the present invention.
[0047] Figure 5 A block diagram of a spatiotemporal evaluation system for a robot motion trajectory according to an embodiment of the present invention. DETAILED DESCRIPTION
[0048] In order to facilitate those skilled in the art to understand the technical solution of the present invention, the technical solution of the present invention is further described in conjunction with the accompanying drawings of the specification.
[0049] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.
[0050] See also Figure 1 As shown, the present invention provides a spatiotemporal evaluation method for a robot motion trajectory, comprising the following steps:
[0051] S100, the evaluation level of the robot's motion trajectory is divided into different levels using signal timing logic rules, and each level is distinguished and described.
[0052] In one embodiment of the present invention, in the field of unmanned systems, the intelligence level of unmanned systems is generally divided into five levels according to the degree of their autonomy:
[0053] Level 1: Remote control, requires a human operator to operate.
[0054] Level 2: Assisted autonomy, capable of performing some basic automated functions.
[0055] Level 3: Partial autonomy, capable of autonomous navigation and decision-making in specific situations.
[0056] Level 4: Highly autonomous, able to complete most tasks without human intervention.
[0057] Level 5: Full autonomy, capable of performing all tasks and operations without any human intervention.
[0058] Similarly, the evaluation level of the robot motion trajectory is divided into 5 levels from H1 to H5. This embodiment is only explained with 5 levels, and is not limited to 5 levels. The specific level is adjustable. The present invention uses the STL formula to distinguish and describe different levels:
[0059] H i :STL[A i ,B i ](Y j ≤index≤Y k );
[0060] In the formula, H i is the i-th level of the robot motion trajectory, where i=1,2,3,...,n, n is the level, and in this embodiment, n is 5; STL is the signal timing logic rule symbol, including "□", "◇", "∧" and "∨". Among them, "□" represents the global property, which requires that the set STL formula constraint must always be satisfied within the set range; "◇" represents the final property, which requires that the set STL formula constraint must be satisfied at a certain moment within the set range; "∧" represents logical AND, which means that two conditions must be satisfied at the same time; "∨" represents logical OR, which requires that at least one of the two conditions be satisfied; [A i ,B i ] is the level area; index is the index value, Y j and Y k are the upper and lower limit values of the indicator value respectively, j and k are different positive integers, and k>j.
[0061] See also Figure 2As shown, each level formula describes a rectangular area, and each level curve will pass through and only the curve of this level will pass through this rectangular area.
[0062] Because there are multiple symbols for signal timing logic rules, that is, at each level, there are multiple different STL formula constraints. Take the H1 level as an example: It means that within the AB time period, the index value must be greater than or equal to Y1 and less than or equal to Y2 in order to determine that the evaluation result of this motion trajectory is at the H1 level.
[0063] S200 evaluates the robot's motion trajectory from two dimensions: timing index and space index, and obtains the optimal timing-level STL parameter model for each level under the timing index, and obtains the optimal space-level STL parameter model for each level under the space index.
[0064] See also Figure 1 and Figure 3 As shown, in one embodiment of the present invention, the total time consumed by the robot to reach the designated destination and complete the task is T, and the overall trajectory length is L. The quality of the robot's motion trajectory can be comprehensively evaluated from two dimensions: time and space.
[0065] like Figure 3 As shown, let the straight-line distance from the starting point O to the end point E be the optimal path L min , trajectory point P i (i=1,2,3,4) to the optimal path L min The foot of the perpendicular is M, and the length of OM is the projection distance L pro . It can be seen from the figure that when the projection distances of different trajectory points are the same, the shorter the time taken, the closer the motion trajectory is to the optimal path. Based on this, the position information and timing information of the robot under different motion trajectories can be used as the characteristics of the state space, and the acceleration, deceleration or turning behavior of the robot can be used as the characteristics of the action space. Use sensors, visual systems or simulators to obtain the timing information and position data of the robot, where the timing information is expressed as "projection distance-projection duration", the motion space is simplified into a two-dimensional plane, and the position data is expressed as "xy" in a Cartesian coordinate system. That is, in this embodiment, the timing information is used as a timing indicator, and the position data is used as a spatial indicator.
[0066] In this embodiment, an example is given to obtain the optimal timing level STL parameter model of each level under the timing indicator, which includes the following steps:
[0067] S211, define the timing level STL initial model. Each H i The motion trajectory of the level corresponds to an STL formula, as wait.
[0068] S212, initializing the parameters of the STL initial model at the time sequence level. According to the time limit and space threshold of the robot motion process, the parameters a, b, y1, y2 of the STL initial model at different levels are initialized.
[0069] S213, calculating the robustness: applying the robustness to characterize the deviation distance between the trajectory signal point and the trajectory that meets the timing level STL initial model.
[0070] In this embodiment, different timing level STL initial models have different structural expressions, and different timing level STL initial models have different corresponding robustness ρ calculation methods. In this embodiment, under the timing index, six forms are listed, including all possible STL formulas:
[0071] Timing-Robustness Formula 1: ρ(x,(l(x) <c),t)=c-l(x(t));
[0072] Timing-robustness formula 2: ρ(x,(l(x)≥c),t)=l(x(t))-c;
[0073] Timing-Robustness Formula 3:
[0074] Timing-Robustness Formula 4:
[0075] Timing-Robustness Formula 5:
[0076] Timing-Robustness Formula 6:
[0077] In the formula, x(t) is the monitoring signal at time t, l(x(t) is the evaluation result of the x signal at time t, l(x) is the timing evaluation result of the x signal, c is a constant, For different signal timing logic, It is expressed as the “delivery” of tasks under the timing indicator. It is represented as the “and” of tasks under the timing indicator, and [a,b) is the time interval.
[0078] In this embodiment, there are multiple different timing level STL initial models at each level, and each timing level STL initial model is calculated corresponding to a different robustness. Through the signal timing logic rule symbol, each timing level STL initial model is confirmed, and the corresponding robustness is calculated. For example: at the H1 level, there is and ◇[a1,b1](y1≤index≤y2) are used to express the robustness calculation under the timing index. The expression form of adopts the timing-robustness formula 5 for robustness calculation, and the expression form of ◇[a1,b1](y1≤index≤y2) adopts the timing-robustness formula 6 for robustness calculation. If the signal timing logic rule symbol is omitted, the timing-robustness formula 1 or 2 is directly applied for calculation.
[0079] S214, define the incentive function: for each level, the training objective function of the time-series level STL initial model is to use the function with the smallest robustness calculation result value as the reward function.
[0080] S215, establishing a reinforcement learning model: using a reinforcement learning algorithm, establishing an initial timing level STL parameter model, with the timing indicators as input and the STL model parameters of the initial timing level STL parameter model as output.
[0081] In one embodiment of the present invention, a reinforcement learning algorithm, such as the proximal policy optimization algorithm (PPO algorithm), deep reinforcement learning, Q-Learning or policy gradient method, is used to establish a model of optimal STL formula parameters that can be learned. The model uses timing indicators as input and outputs the parameters of the STL parameter model.
[0082] S216, training and acquiring models: Use the data set to train the initial time-series level STL parameter model; in each training iteration, perform corresponding actions according to the current state and the output of the model, and observe the feedback from the environment; update and optimize the model parameters through the reward function to gradually learn and obtain the optimal time-series level STL parameter model for each level.
[0083] In one embodiment of the present invention, the data set includes a temporal training set and a spatial training set, and the temporal training set is used herein. The temporal training set includes temporal training subsets for training at different levels, and the temporal training subsets at each level correspond to the temporal level STL parameter models for training at the same level.
[0084] See also Figures 1 to 3 As shown, in one embodiment of the present invention, the step of obtaining the optimal spatial level STL parameter model for each level under the spatial index is the same as the step of obtaining the optimal temporal level STL parameter model for each level under the temporal index. The difference is that the multiple robustness ρ under the spatial index is obtained by the following formula:
[0085]
[0086] In the formula, x(T) is the monitoring signal at position T, l(x(T) is the evaluation result of the x signal at position T, l(x) is the spatial evaluation result of the x signal, c is a constant, φ, φ1, φ2 are different signal space logics, φ1∧φ2 represents the "intersection" of tasks under spatial indicators, φ1∨φ2 represents the "and" of tasks under spatial indicators, and [a, b) is the position point interval; similarly, according to each level, there are multiple different spatial level STL initial models, and each spatial level STL initial model corresponds to a different robustness calculation. Through the signal temporal logic rule symbol, each spatial level STL initial model is confirmed, and the corresponding robustness is calculated.
[0087] In addition, the reinforcement learning model established under the spatial indicator takes the spatial indicator as input and the STL model parameters of the initial spatial level STL parameter model as output, and the spatial training set includes spatial training subsets for training at different levels, and the spatial training subset of each level corresponds to the training of the spatial level STL parameter model at the same level.
[0088] S300, using logical connectors, the optimal time-level STL parameter model and the optimal space-level STL parameter model at each level are combined and fused to obtain multiple-level fused STL parameter models at each level.
[0089] In one embodiment of the present invention, taking 5 levels as an example, in this step, the optimal timing level STL parameter model of 5 levels and the optimal space level STL parameter model of 5 levels under the timing index are obtained. In this embodiment, the logical connectors are "and" connectors and "or" connectors, wherein the logical connectors can be used alone or in combination. Taking the H1 level as an example, the optimal timing level STL parameter model is: The optimal space-level STL parameter model is: Combination and fusion exist form of expression.
[0090] S400, from the multiple level fusion STL parameter models at each level, select the model with the highest accuracy as the optimal level fusion STL parameter model at this level.
[0091] In this embodiment, the data set also includes a test set, and the test set also includes a timing test set and a space test set, and the timing test set includes timing test subsets of different levels, and the space test set includes space test subsets of different levels. The timing test subsets and space test subsets of the same level are brought into the level fusion STL parameter model of the same level after fusion, and the level is evaluated and compared with the marked level, and the model with the highest accuracy is selected as the optimal level fusion STL parameter model at this level.
[0092] S500, puts the same robot motion trajectory into the optimal level fusion STL parameter model at different levels to evaluate the robot motion trajectory; and classifies the evaluation results according to different levels to comprehensively reflect the advantages and disadvantages of the robot motion trajectory.
[0093] In one embodiment of the present invention, when the motion trajectory of the robot under test passes through a square area of a certain level of optimal level fusion STL parameter model, the evaluation result of the motion trajectory is considered to be H i class.
[0094] like Figure 4a and Figure 4b As shown, in one embodiment of the present invention, the evaluation results of the robot's timing information and position data are obtained after learning and training the STL formula using the PPO algorithm. Figure 4a is the evaluation result of the timing information, Figure 4a is the evaluation result of the position data. When the curve composed of the dotted line represented by the normal signal passes through the gray rectangle, it means that the evaluation result of this trajectory is level H5.
[0095] See also Figures 1 to 5 As shown, the present invention also provides a system according to the spatiotemporal evaluation method of the robot motion trajectory described above, including: a level division module, which is used to divide the evaluation level of the robot motion trajectory into different levels using signal timing logic rules, and distinguish and describe each level. An optimal parameter model module is used to evaluate the robot motion trajectory from the two dimensions of timing indicators and space indicators, respectively, to obtain the optimal timing level STL parameter model of each level under the timing indicators, and to obtain the optimal space level STL parameter model of each level under the space indicators. A level fusion module is used to combine and fuse the optimal timing level STL parameter model and the optimal space level STL parameter model under each level using logical connectors to obtain multiple level fusion STL parameter models under each level. An optimal level fusion module is used to select the model with the highest accuracy from the multiple level fusion STL parameter models under each level as the optimal level fusion STL parameter model under this level. The evaluation level classification module is used to simultaneously pass the robot's motion trajectory through the optimal level fusion STL parameter model at different levels to evaluate the robot's motion trajectory; and classify the evaluation results according to different levels to comprehensively reflect the advantages and disadvantages of the robot's motion trajectory.
[0096] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting from any point of view, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any reference numerals in the claims should not be regarded as limiting the claims involved.
[0097] The above-described embodiments merely represent implementation methods of the invention. The protection scope of the present invention is not limited to the above-described embodiments. For those skilled in the art, several modifications and improvements may be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.
Claims
1. A spatiotemporal evaluation method for robot motion trajectory, characterized in that: The following steps are involved: S100, dividing the evaluation level of the robot's motion trajectory into different levels using signal timing logic rules, and distinguishing and describing each level; S200, evaluating the robot motion trajectory from two dimensions, namely, the timing index and the space index, respectively, to obtain the optimal timing-level STL parameter model for each level under the timing index, and to obtain the optimal space-level STL parameter model for each level under the space index; S300, using a logical connector, the optimal time-series level STL parameter model and the optimal space-level STL parameter model at each level are combined and fused to obtain a plurality of level fused STL parameter models at each level; S400, selecting the model with the highest accuracy from the multiple level fusion STL parameter models at each level as the optimal level fusion STL parameter model at this level; S500, puts the same robot motion trajectory into the optimal level fusion STL parameter model at different levels to evaluate the robot motion trajectory; and classifies the evaluation results according to different levels to comprehensively reflect the advantages and disadvantages of the robot motion trajectory.
2. The spatiotemporal evaluation method of robot motion trajectory according to claim 1, characterized in that: The evaluation level of the robot's motion trajectory is expressed by the following formula: H i :STL[A i ,B i ](Y j ≤index≤Y k ); In the formula, H i is the i-th level of the robot motion trajectory, where i = 1, 2, 3, ..., n, and n is the level of the level; STL is the signal timing logic rule symbol, including "□", "◇", "∧" and "∨"; where "□" represents the global property, which requires that the set STL formula constraint must always be satisfied within the set range; "◇" represents the final property, which requires that the set STL formula constraint must be satisfied at a certain moment within the set range; "∧" represents logical AND, which means that two conditions must be satisfied at the same time; "∨" represents logical OR, which requires that at least one of the two conditions be satisfied; [A i ,B i ] is the level area; index is the index value, Y j and Y k They are the upper and lower limit values of the indicator respectively.
3. The spatiotemporal evaluation method of robot motion trajectory according to claim 1, characterized in that: Use sensors, visual systems or simulators to obtain the robot's timing and spatial indicators; the timing indicators are expressed as "projection distance-projection time"; the spatial indicators are expressed as "xy" in the Cartesian coordinate system.
4. The spatiotemporal evaluation method of robot motion trajectory according to claim 1, characterized in that: Acquiring the optimal timing level STL parameter model of each level under the timing indicator in step S200 includes the following steps: S211, define the timing level STL initial model; S212, initializing the parameters of the timing level STL initial model; S213, calculating the robustness: applying the robustness to characterize the deviation distance between the trajectory signal point and the trajectory that meets the timing level STL initial model; S214, defining an incentive function: for each level, the training objective function of the time-series level STL initial model is a function with the smallest robustness calculation result value as a reward function; S215, establishing a reinforcement learning model: using a reinforcement learning algorithm, establishing an initial time series level STL parameter model, taking the time series index as input, and taking the STL model parameters of the initial time series level STL parameter model as output; S216, training and acquiring models: Use the data set to train the initial time-series level STL parameter model; in each training iteration, perform corresponding actions according to the current state and the output of the model, and observe the feedback from the environment; update and optimize the model parameters through the reward function to gradually learn and obtain the optimal time-series level STL parameter model for each level.
5. The spatiotemporal evaluation method of robot motion trajectory according to claim 4, characterized in that: Under the timing index, multiple robustness is obtained by the following formula: Where x(t) is the monitoring signal at time t, l(x(t) is the evaluation result of the x signal at time t, l(x) is the timing evaluation result of the x signal, c is a constant, For different signal timing logic, It is expressed as the "delivery" of tasks under the timing indicator. It is expressed as the "and" of tasks under the timing indicator, and [a,b) is the time interval; there are multiple different timing level STL initial models under each level, and each timing level STL initial model is calculated corresponding to different robustness; through the signal timing logic rule symbol, each timing level STL initial model is confirmed, and the corresponding robustness is calculated.
6. The spatiotemporal evaluation method of robot motion trajectory according to claim 4, characterized in that: The steps for obtaining the optimal spatial-level STL parameter model for each level under the spatial index are the same as the steps for obtaining the optimal temporal-level STL parameter model for each level under the temporal index; the difference is that multiple robustness under the spatial index is obtained by the following formula: Wherein, x(T) is the monitoring signal at position T, l(x(T) is the evaluation result of the x signal at position T, l(x) is the spatial evaluation result of the x signal, c is a constant, φ, φ1, φ2 are different signal space logics, φ1∧φ2 represents the "intersection" of the tasks under the spatial indicator, φ1∨φ2 represents the "and" of the tasks under the spatial indicator, and [a, b) is the position point interval; according to each level, there are multiple different spatial level STL initial models, and each spatial level STL initial model corresponds to a different robustness calculation; through the signal temporal logic rule symbol, each spatial level STL initial model is confirmed, and the corresponding robustness is calculated.
7. The spatiotemporal evaluation method of robot motion trajectory according to claim 6, characterized in that: Under the spatial index, the reinforcement learning model established takes the spatial index as input and the STL model parameters of the initial spatial level STL parameter model as output.
8. The spatiotemporal evaluation method of robot motion trajectory according to claim 1, characterized in that: Under the spatial indicator, the training set for obtaining the optimal time-level STL parameter model and the optimal space-level STL parameter model at each level includes a time-series training set and a space-series training set; and the time-series training set includes time-series training subsets for training at different levels, and the space-series training set includes space-series training subsets for training at different levels.
9. The spatiotemporal evaluation method of robot motion trajectory according to claim 1, characterized in that: The logical connectors are the "and" connector and the "or" connector.
10. A system for the spatiotemporal evaluation method of robot motion trajectory according to any one of claims 1 to 9, characterized in that: include: The level division module is used to divide the evaluation level of the robot's motion trajectory into different levels using signal timing logic rules, and distinguish and describe each level; The optimal parameter model module is used to evaluate the robot's motion trajectory from the two dimensions of timing indicators and spatial indicators, and obtain the optimal timing level STL parameter model for each level under the timing indicators, and obtain the optimal spatial level STL parameter model for each level under the spatial indicators; The level fusion module is used to combine and fuse the optimal time-series level STL parameter model and the optimal space-level STL parameter model at each level using logical connectors to obtain multiple level fusion STL parameter models at each level; The optimal level fusion module is used to select the model with the highest accuracy from multiple level fusion STL parameter models at each level as the optimal level fusion STL parameter model at this level; The evaluation level classification module is used to put the same robot motion trajectory into the optimal level fusion STL parameter model at different levels to evaluate the robot motion trajectory; and classify the evaluation results according to different levels to comprehensively reflect the advantages and disadvantages of the robot motion trajectory.
Citation Information
Patent Citations
Robot trajectory processing method and device, and robot
CN108460494A
Robot complex sequential logic task path planning method based on reinforcement learning
CN114355947A
Mobile robot motion planning algorithm based on linear sequential logic
CN115963835A
Mechanical arm task and motion planning method and system based on linear sequential logic
CN116604550A
Industrial robot motion trajectory planning and evaluation method and device
CN117681213A
Cited By
Cascaded large model robot motion planning method based on sequential logic constraint
CN120839780A
