Precise control method and system for complex environment operating robot

Abstract

The present application relates to the technical field of robot precision control, and particularly relates to a precision control method and system for a complex environment operation robot. Laser radar point cloud data is collected, and static point cloud data is screened out; inertial navigation is solved, and a robot predicted state is obtained; the static point cloud is matched with a preset map, and the effective point number ratio of laser observation and the average matching residual error are calculated; a weight adjustment factor of laser observation is dynamically calculated based on a quality index; the factor is used to adaptively adjust the weight of laser observation in fusion, and visual observation is fused, and an optimized robot state estimation is obtained through a filtering algorithm; a control instruction is generated based on the deviation of the estimation and a preset trajectory. The present application can evaluate and adapt to the change of sensor data quality in real time, and significantly improves the trajectory tracking precision and operation robustness of the robot in a harsh environment such as high temperature and high dust.

Description

Technical Field

[0001] This invention relates to the technical field of precise robot control, specifically to a precise control method and system for robots operating in complex environments. Background Technology

[0002] Traditional industrial robots typically operate in structured, known environments, relying on accurate environmental models and stable perception conditions. However, in complex industrial scenarios such as cleaning the interior of metal ore casting ladles and overhauling high-temperature furnaces, the environment presents extreme conditions including high temperatures, high concentrations of suspended dust, extremely confined spaces, and the presence of potentially hazardous substances and flammable / explosive risks. These factors pose significant challenges to robot operations.

[0003] Diffuse dust severely scatters and absorbs light, causing blurred images from visual cameras and numerous false noise points generated by suspended dust in LiDAR point clouds. Environmental features are severely obscured, resulting in distorted point cloud data. Furthermore, localization and mapping based on distorted perception data will suffer significant errors. Traditional multi-sensor fusion algorithms typically use fixed noise covariance or weights. When the data quality of a particular sensor drops sharply due to dust interference, low-quality data continuously "contaminates" the fusion results, leading to incorrect robot state estimation, subsequent trajectory tracking failure, and ultimately, the robot's inability to accurately perform cleaning or inspection tasks. This can result in collisions, missed tasks, or repetitive work. In flammable and explosive environments, communication or control failures could even cause serious safety accidents. Therefore, there is an urgent need for a control method and system that can adapt to dynamic environmental changes, especially under conditions of real-time fluctuations in sensor data quality, while maintaining high accuracy and robustness in state estimation and trajectory tracking. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a precise control method and system for robots operating in complex environments. The specific technical solution adopted is as follows: A precise control method for robots operating in complex environments, comprising: acquiring environmental point cloud data collected by a lidar at the current sampling time; identifying and filtering out noise point cloud data generated by suspended dust in the environment based on the distribution characteristics of all environmental point cloud data in three-dimensional space, thereby obtaining all static point cloud data at the current sampling time; performing inertial navigation calculations on the robot based on motion measurement data from the previous sampling time to obtain the robot's predicted state at the current sampling time; transforming all the static point cloud data to a global map coordinate system, matching it with a preset environmental map, and calculating the effective point ratio and average matching residual at the current sampling time; obtaining a weight adjustment factor for laser observation based on the effective point ratio and the average matching residual; adjusting the uncertainty measure of laser observation in state estimation based on the weight adjustment factor, and fusing the robot's predicted state, laser observation, and visual observation at the current sampling time, and obtaining an optimized state estimate of the robot at the current sampling time through a filtering algorithm; and driving the robot based on the deviation between the optimized state estimate and the preset operating trajectory.

[0005] Furthermore, the method for acquiring the static point cloud data includes: selecting any environmental point cloud data as the target point cloud data; using the target point cloud data as the center, acquiring a preset number of adjacent point cloud data within a spherical neighborhood that gradually expands according to a preset step size, and calculating a first distance between the target point cloud data and each adjacent point cloud data; and calculating the probability that the target point cloud data is noisy point cloud data based on the first distance, using the following formula: In the formula, This indicates the probability that the target point cloud data is noisy point cloud data. Indicates the number of adjacent point cloud data; This represents the distance between the target point cloud data and its first adjacent point cloud data. Indicates the target point cloud data and the first The distance between adjacent point cloud data; Represents the normalization function; All environmental point cloud data exceeding a preset first threshold are treated as noisy point cloud data; environmental point cloud data other than noisy point cloud data are treated as all static point cloud data.

[0006] Furthermore, the method for obtaining the effective point ratio includes: taking the ratio between the number of static point cloud data successfully matched to the preset environment map and the number of static point cloud data as the effective point ratio at the current sampling time.

[0007] Furthermore, the method for obtaining the average matching residual includes: averaging the distance between the position of each static point cloud data successfully matched to the preset environment map on the preset environment map and its changed position on the preset environment map after matching, to obtain the average matching residual at the current sampling time.

[0008] Furthermore, the method for obtaining the weight adjustment factor includes: obtaining the weight adjustment factor according to the weight adjustment factor calculation formula, the weight adjustment factor calculation formula being as follows: In the formula, Indicates the first Weighting adjustment factor for laser observations at each sampling time; Indicates the first The weighting coefficients of the average matching residuals at each sampling time; Indicates the first The average matching residual at each sampling time; Indicates the first The weighting coefficient of the ratio of effective points at each sampling time; Indicates the first The ratio of effective points at each sampling time; Indicates the first The weighting coefficients for dust noise intensity at each sampling time; Indicates the first Dust noise intensity at each sampling time.

[0009] Furthermore, the method for obtaining the dust noise intensity includes: calculating the first... The ratio of the noisy point cloud data at the sampling time to all environmental point cloud data is used as the ratio of the noise point cloud data at the sampling time to ... Dust noise intensity at each sampling time.

[0010] Furthermore, the filtering algorithm is a Kalman filter; the optimized state estimation includes the robot's three-dimensional position, three-dimensional attitude, and velocity.

[0011] A precision control system for a robot operating in complex environments includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the precision control method for a robot operating in complex environments as described above.

[0012] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the precise control method for a robot operating in a complex environment as described above.

[0013] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the precise control method for a robot operating in a complex environment as described above.

[0014] This invention offers the following advantages: Firstly, it effectively overcomes the direct impact of dust on laser sensing through intelligent point cloud denoising at the front end, improving the quality of environmental point cloud data. Secondly, it innovatively introduces a dynamic weight adjustment mechanism based on real-time data quality. When dust interference causes increased matching residuals or reduced effective data in the lidar, the system automatically reduces its weight in state estimation, preventing low-quality data from compromising overall estimation accuracy and enhancing the system's operational capability under dynamic sensor performance degradation. Thirdly, it uses highly reliable adaptive fusion state estimation for trajectory tracking control, ensuring the robot can still operate with high precision along a preset path even in extremely complex environments, improving task success rate and operational efficiency. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a precise control method for a robot operating in complex environments, provided as an embodiment of the present invention; Figure 2 This is a block diagram of a precision control system for a robot operating in complex environments, provided as an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a precise control method and system for a robot operating in complex environments according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the precise control method and system for a robot operating in complex environments provided by this invention.

[0020] Please see Figure 1 This illustrates a precise control method for a robot operating in a complex environment according to an embodiment of the present invention. The method includes: step S1: acquiring environmental point cloud data collected by lidar at the current sampling time; identifying and filtering out noise point cloud data generated by environmental suspended dust based on the distribution characteristics of all environmental point cloud data in three-dimensional space, and obtaining all static point cloud data at the current sampling time.

[0021] This invention is primarily applied to scenarios where robots can automatically perform cleaning tasks inside metal ore casting ladles. After the robot enters the casting ladle, a lidar system begins scanning. Because the environment is filled with high-temperature suspended dust, the original point cloud contains a large number of discretely distributed noise points generated by the dust. Therefore, this invention analyzes each sampling moment. First, it acquires the environmental point cloud data collected by the lidar at the current sampling moment, then identifies and filters out the noisy point cloud data, selecting the points for subsequent analysis.

[0022] Preferably, in one embodiment of the present invention, the method for acquiring static point cloud data includes: selecting any environmental point cloud data as the target point cloud data; acquiring a preset number of adjacent point cloud data within a spherical neighborhood that gradually expands according to a preset step size, with the target point cloud data as the center, and calculating a first distance between the target point cloud data and each adjacent point cloud data; and calculating the probability that the target point cloud data is noisy point cloud data based on the first distance, using the following calculation formula: In the formula, This indicates the probability that the target point cloud data is noisy point cloud data. Indicates the number of adjacent point cloud data; This represents the distance between the target point cloud data and its first adjacent point cloud data. Indicates the target point cloud data and the first The distance between adjacent point cloud data; This represents the normalization function.

[0023] In the above calculation formula, the greater the distance between the target point cloud data and its nearest neighboring point cloud data, the more significant the difference between the two points. The greater the distance between adjacent point cloud data points, the more dispersed the distribution of adjacent point cloud data, and the more likely the target point cloud data is noisy point cloud data. right Normalization is performed to obtain the probability that the target point cloud data is noisy point cloud data.

[0024] The preset quantity is set to 10, meaning the number of adjacent point cloud data is 10. It should be noted that the preset quantity can be set manually and is not limited here.

[0025] Will All environmental point cloud data exceeding a preset first threshold are classified as noisy point cloud data; all environmental point cloud data other than the noisy point cloud data are classified as static point cloud data. The preset first threshold is set to 0.7. It should be noted that the preset first threshold can be set independently and is not limited here.

[0026] Step S2: Based on the motion measurement data of the robot at the previous sampling time, perform inertial navigation calculation on the robot to obtain the predicted state of the robot at the current sampling time.

[0027] In practical applications, robots often operate in harsh environments where dust is abundant and visual and laser data may be severely degraded or fail. Conventional LiDAR may fail. However, since IMU-based robot state prediction has the advantages of being completely autonomous, high-frequency, and unaffected by external environmental interference, in order to optimize the robot state estimation later, this embodiment of the invention first uses IMU to predict the robot's motion state at the current sampling moment.

[0028] In one embodiment of the present invention, a robot state prediction method based on IMU is provided, specifically including: firstly, assuming the robot starts from a fixed docking station, and setting its initial three-dimensional coordinates as follows: The attitude level is The initial velocity is 0. During the system warm-up and stationary phase, static initialization calibration is performed to obtain zero bias IMU acceleration. Zero bias of gyroscope .

[0029] definition The robot's state vector at time t is: in, Indicates the first The three-dimensional position of the robot at each sampling time; Indicates the first Three-dimensional pose at each sampling time; express The three-dimensional linear velocity at time t; For the first Zero bias of the IMY accelerometer and gyroscope at each sampling time.

[0030] Inertial Measurement Unit Obtain from the From the sampling time to the... Acceleration at each sampling time point angular velocity And based on the first The state vector at each sampling time Obtain the Zero bias estimation at each sampling time point This is used to analyze and obtain the robot's predicted state at the current moment. The specific steps are as follows:

[0031] a) First, zero-bias compensation is performed on the IMU data based on the zero-bias estimate from the previous time step, and the result is calculated as follows: Compensation acceleration at each sampling time ;as well as Compensated angular velocity at time t .

[0032] b) Integrate using the principle of inertial navigation, and utilize the compensated angular velocity. Accumulate points and update. Attention Quaternion at a given moment ; using the compensated acceleration Based on the current attitude and the velocity at the previous moment, the integral is updated. Speed ​​at any moment Then use the updated speed Accumulate points and update. 3D position at time .

[0033] c) thereby obtaining the first Predicted state of the robot at each sampling time point It should be noted that the zero bias remains unchanged temporarily in this step (it is predicted to be the same as the previous time step).

[0034] Step S3: Transform all static point cloud data to the global map coordinate system, match it with the preset environment map, and calculate the effective point ratio and average matching residual at the current sampling time; obtain the weight adjustment factor for laser observation based on the effective point ratio and average matching residual.

[0035] In order to quantify and diagnose the reliability of the point cloud data observed by the robot's laser at the current sampling time, in this embodiment of the invention, all static point cloud data are transformed to the global map coordinate system and matched with the preset environment map to calculate the effective point ratio and the average matching residual at the current sampling time; based on the effective point ratio and the average matching residual, the weight adjustment factor of the laser observation is obtained, and the quality of the laser observation is evaluated through the weight adjustment factor.

[0036] The static point cloud data is transformed into a pre-constructed global environment map coordinate system using preliminary robot pose estimation from IMU predictions. Next, the Iterative Closest Point (ICP) algorithm is used to find the optimal match between the static point cloud data at the current sampling time and the corresponding structure in the map. After matching, the effective point ratio, average matching residual, and dust noise intensity are analyzed.

[0037] Preferably, in one embodiment of the present invention, the method for obtaining the effective point ratio includes: taking the ratio between the number of static point cloud data successfully matched to the preset environmental map and the number of static point cloud data as the effective point ratio at the current sampling time. The effective point ratio reflects the proportion of usable information in the environmental point cloud data at the current sampling time. In the subsequent weight adjustment factor calculation formula, its importance is greater than the dust noise intensity and less than the average matching residual.

[0038] Preferably, in one embodiment of the present invention, the method for obtaining the average matching residual includes: averaging the distance between the position of each static point cloud data successfully matched to the preset environment map on the preset environment map and its changed position on the preset environment map after matching, to obtain the average matching residual at the current sampling time. The average matching residual reflects the degree of consistency between the laser observation and the internal prediction of the system at the current sampling time. The larger the average matching residual, the more it indicates that the pose estimation is biased or that the environment has undergone a drastic change that is not modeled. Therefore, it occupies the highest weight in the subsequent weight adjustment factor calculation formula.

[0039] Preferably, in one embodiment of the present invention, the method for obtaining dust noise intensity includes: calculating the first... The ratio of the noisy point cloud data at the sampling time to all environmental point cloud data is used as the ratio of the noise point cloud data at the sampling time to ... The dust noise intensity at each sampling time indicates that the environmental conditions for laser observation at the current sampling time are more severe. Among them, noise intensity is an important early warning signal and auxiliary judgment, but not a decisive criterion. Therefore, it occupies the lowest weight in the subsequent weight adjustment factor calculation formula.

[0040] Preferably, in one embodiment of the present invention, the method for obtaining the weight adjustment factor includes: obtaining the weight adjustment factor according to the weight adjustment factor calculation formula, the weight adjustment factor calculation formula is as follows: In the formula, Indicates the first Weighting adjustment factor for laser observations at each sampling time; Indicates the first The weighting coefficient of the average matching residual at each sampling time; Indicates the first The average matching residual at each sampling time; Indicates the first The weighting coefficient of the ratio of effective points at each sampling time; Indicates the first The ratio of effective points at each sampling time; Indicates the first The weighting coefficients for dust noise intensity at each sampling time; Indicates the first Dust noise intensity at each sampling time.

[0041] In one embodiment of the present invention, the first The weighting coefficient for the average matching residual at the sampling time step is set to 0.5. The weighting coefficient for the ratio of effective points at the sampling time is set to 0.3. The weighting coefficient for the dust noise intensity at each sampling time point is set to 0.2. It should be noted that the weighting coefficient can be set manually, as long as it meets the following requirements. That's all.

[0042] Step S4: Adjust the uncertainty measure of laser observation in state estimation according to the weight adjustment factor, and fuse the robot's predicted state, laser observation and visual observation at the current sampling time, and obtain the optimized state estimate of the robot at the current sampling time through the filtering algorithm.

[0043] In this embodiment of the invention, the first prediction obtained by IMU is... Predicted state at each sampling time As the prediction step of the Kalman filter. In the update step, the noise covariance matrix of the laser observation is... It is not fixed, but dynamically adjusted according to its quality: the adjustment formula is: ;in, This represents the baseline noise covariance of the lidar in a clean environment; simultaneously, the visual front end enhances the image and extracts feature points using a dehazing algorithm. The noise covariance matrix of the laser observation is then used. The visual observations are input into an extended Kalman filter and fused with IMU predictions to output the optimized global pose and velocity of the robot at the current moment. Specifically, when... When it increases, As the Kalman gain increases, the laser component in the Kalman gain decreases, thus reducing the weight of laser observation in this fusion.

[0044] Step S5: Drive the robot based on the deviation between the optimized state estimate and the preset work trajectory.

[0045] The optimized global pose and velocity of the robot are input into the robot's model predictive controller (MPC). The MPC compares the optimized state estimate with the preset trajectory of the casting ladle cleaning operation, calculates the optimal control sequence (such as wheel speed or joint torque) for the next few steps, and sends the first control variable to the robot's underlying actuator to drive the robot to move precisely along the preset trajectory to complete the cleaning task.

[0046] In summary, the following steps are taken: First, environmental point cloud data from the LiDAR at the current sampling time is acquired. Second, based on the distribution characteristics of all environmental point cloud data in three-dimensional space, noise point cloud data generated by suspended dust in the environment is identified and filtered out to obtain all static point cloud data at the current sampling time. Third, based on the robot's motion measurement data from the previous sampling time, inertial navigation calculations are performed on the robot to obtain the predicted state of the robot at the current sampling time. Fourth, all static point cloud data are transformed to the global map coordinate system and matched with a preset environmental map to calculate the effective point ratio and average matching residual at the current sampling time. Fifth, based on the effective point ratio and average matching residual, a weighting adjustment factor for the LiDAR observation is obtained. Sixth, the uncertainty measure of the LiDAR observation in the state estimation is adjusted according to the weighting adjustment factor, and the predicted state of the robot at the current sampling time, LiDAR observation, and visual observation are fused together. An optimized state estimate of the robot at the current sampling time is obtained through a filtering algorithm. Finally, the robot is driven based on the deviation between the optimized state estimate and the preset work trajectory.

[0047] One embodiment of the present invention provides a precision control system for a robot operating in complex environments. The system includes a memory, a processor, and a computer program. The memory stores the corresponding computer program, and the processor runs the corresponding computer program. When the computer program runs in the processor, it can implement the methods described in steps S1-S5, specifically including: a point cloud data acquisition module 101, used to acquire environmental point cloud data collected by a lidar at the current sampling time; based on the distribution characteristics of all environmental point cloud data in three-dimensional space, identifying and filtering out noise point cloud data generated by suspended dust in the environment, obtaining all static point cloud data at the current sampling time; and a robot state prediction module 102, used to predict the robot's state based on the robot's motion measurement data at the previous sampling time. The inertial navigation system calculates the robot's predicted state at the current sampling time. A laser observation quality assessment module 103 transforms all static point cloud data to a global map coordinate system, matches it with a preset environment map, and calculates the effective point ratio and average matching residual at the current sampling time. Based on the effective point ratio and average matching residual, a weight adjustment factor for laser observation is obtained. A state estimation optimization module 104 adjusts the uncertainty measure of laser observation in state estimation according to the weight adjustment factor, and integrates the robot's predicted state, laser observation, and visual observation at the current sampling time. An optimized state estimate of the robot at the current sampling time is obtained through a filtering algorithm. A drive module 105 drives the robot based on the deviation between the optimized state estimate and the preset work trajectory.

[0048] A third objective of this invention is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the methods described in steps S1-S5.

[0049] The fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in steps S1-S5.

[0050] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0051] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A precise control method for a robot operating in complex environments, characterized in that, The method includes: acquiring environmental point cloud data collected by lidar at the current sampling time; identifying and filtering out noise point cloud data generated by environmental suspended dust based on the distribution characteristics of all environmental point cloud data in three-dimensional space to obtain all static point cloud data at the current sampling time; performing inertial navigation calculation on the robot based on the robot's motion measurement data at the previous sampling time to obtain the robot's predicted state at the current sampling time; transforming all the static point cloud data to a global map coordinate system, matching it with a preset environmental map, and calculating the effective point ratio and average matching residual at the current sampling time; obtaining a weight adjustment factor for laser observation based on the effective point ratio and the average matching residual; and adjusting the laser observation based on the weight adjustment factor. The uncertainty measure in state estimation is integrated with the robot's predicted state at the current sampling moment, laser observation, and visual observation. An optimized state estimate of the robot at the current sampling moment is obtained through a filtering algorithm. The robot is driven based on the deviation between the optimized state estimate and the preset work trajectory. The method for acquiring static point cloud data includes: selecting any environmental point cloud data as the target point cloud data; acquiring a preset number of neighboring point cloud data within a spherical neighborhood that gradually expands according to a preset step size, centered on the target point cloud data, and calculating a first distance between the target point cloud data and each neighboring point cloud data; calculating the probability that the target point cloud data is noisy point cloud data based on the first distance, using the following formula: In the formula, This indicates the probability that the target point cloud data is noisy point cloud data. Indicates the number of adjacent point cloud data; This represents the distance between the target point cloud data and its first adjacent point cloud data. Indicates the target point cloud data and the first The distance between adjacent point cloud data; Represents the normalization function; All environmental point cloud data exceeding a preset first threshold are considered noisy point cloud data; environmental point cloud data other than noisy point cloud data are considered all static point cloud data; the method for obtaining the effective point ratio includes: taking the ratio between the number of static point cloud data successfully matched to the preset environmental map and the total number of static point cloud data as the effective point ratio at the current sampling time; the method for obtaining the average matching residual includes: averaging the distance between the position of each static point cloud data successfully matched to the preset environmental map on the preset environmental map and its changed position on the preset environmental map after matching, to obtain the average matching residual at the current sampling time; the method for obtaining the weight adjustment factor includes: obtaining the weight adjustment factor according to the weight adjustment factor calculation formula, the weight adjustment factor calculation formula is as follows: In the formula, Indicates the first Weighting adjustment factor for laser observations at each sampling time; Indicates the first The weighting coefficient of the average matching residual at each sampling time; Indicates the first The average matching residual at each sampling time; Indicates the first The weighting coefficient of the ratio of effective points at each sampling time; Indicates the first The ratio of effective points at each sampling time; Indicates the first The weighting coefficients for dust noise intensity at each sampling time; Indicates the first The dust noise intensity at the sampling time; the method for obtaining the dust noise intensity includes: calculating the dust noise intensity at the sampling time. The ratio of the noisy point cloud data at the sampling time to all environmental point cloud data is used as the ratio of the noise point cloud data at the sampling time to ... Dust noise intensity at each sampling time.

2. The precise control method for a robot operating in complex environments according to claim 1, characterized in that, The filtering algorithm is a Kalman filter; the optimized state estimation includes the robot's three-dimensional position, three-dimensional attitude, and velocity.

3. A precision control system for a robot operating in complex environments, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the precise control method for a robot operating in a complex environment as described in any one of claims 1 to 2.

4. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the precise control method for a robot operating in a complex environment as described in any one of claims 1 to 2.

5. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the precise control method for a robot operating in a complex environment as described in any one of claims 1 to 2.

Images