A mobile line marking robot with visual odometry

Abstract

The present invention relates to a mobile line marking robot with visual odometry, making it less susceptible to poor GNSS signals.

Description

[0001] A mobile line marking robot with visual odometry

[0002] Technical field of the invention

[0003] The present invention relates to mobile line marking robots.

[0004] Background of the invention

[0005] Today, robots have been developed that provide solutions for e.g., land surveying, sports field marking, and road marking. Some robots are e.g., developed to paint sports fields autonomously with minimal user interaction. This solution reduces the time it takes to paint a sports field compared to the traditional manual marking operation.

[0006] Some mobile line marking robots employ satellite navigation to enhance the mobile line marking robot’s ability for accurate navigation within the work area. Satellite navigation system is a system of satellites that provide autonomous geo-spatial positioning with global coverage. It allows for small electronic receivers to determine their location (longitude, latitude, and altitude) to within a few meters, or even centimeters, using signals transmitted along a line-of-sight by radio from satellites. The electronic receivers calculate the precise time, position, and carrier phase for the apparatus on which they are mounted or integrated. A satellite navigation system with global coverage may be termed a global navigation satellite system or GNSS (Global Navigation Satellite System). Particularly good accuracy of GNSS systems is obtained by using one or more fixed, land-based reference receivers, or beacons, in addition to the satellites. This is called differential GNSS (DGNSS). There are several DGNSS techniques, such as the classical DGNSS (or DGPS), the Real Time Kinematics (RTK), and the Wide Area RTK (WARTK).

[0007] The use of GNSS systems requires good reception of satellite signals to work reliably. For a mobile line marking robot that employs satellite navigation, the satellite signals may become blocked by obstacles, such as buildings, roofs, sheds, trees, and the like, at or near the work area. The mobile line marking robot may therefore lose track and be momentarily prevented from continuing with satellite navigation-based operation in the work area. For DGNSS, the signal from a reference receiver or beacon may likewise be blocked by, for example, any of the obstacles mentioned above, or by radio disturbances, such as interference, and multi-path propagation.

[0008] When the mobile line marking robot has no other navigation means, it will have to suspend its current propulsion / movement and marking operations and wait until the satellite navigation receiver again receives the satellite signals sufficiently well to recalibrate its current position and resume its satellite navigation-based operations in the work area. Otherwise, the poor navigation signal may lead to uneven lines or the robot losing its centimeter tolerance while marking.

[0009] Furthermore, every time a mobile line marking robot is forced to suspend its operations momentarily and then resume the operations after a short time, the electrical motor, and pump unit will be subjected to a certain stress. If such stops occur frequently, the accumulated wear and tear of critical parts will be considerable, and a premature breakdown of any of these critical parts is at risk. Moreover, repeated stops and resumptions of the mobile line marking robot’s movement will consume a small but not negligible additional amount of electric power, hence causing a penalty in battery power consumption and reducing the maximum operational duration until next recharging of the battery.

[0010] Summary of the invention

[0011] It is one object of the present invention to provide a mobile line marking robot that is less susceptible to the above-mentioned problems.

[0012] The present invention aims to leverage computer vision to augment the sensor input for enhancing the positioning capabilities of the mobile line marking robot. This design incorporates the utilization of one camera alone, or two cameras arranged in a stereo vision configuration, with the system processing the respective streams from the camera(s).

[0013] A first aspect relates to a mobile line marking robot comprising:

[0014] - a chassis with one or more wheels operably connected via a motor controller to at least one motor;

[0015] - a spray means comprising a spray nozzle with an outlet;

[0016] - a paint reservoir;

[0017] - a pump unit operably connected to said spray means and said paint reservoir, thereby allowing paint from the paint reservoir to exit said spray nozzle;

[0018] - a positioning system receiver unit configured for receiving a positioning signal from a Global Navigation Satellite System (GNSS); and

[0019] - a first processor configured to receive the positioning information signal from the positioning system receiver unit and in response thereto a) to activate or deactivate the pump unit and / or b) to instruct the motor controller to navigate a predefined or autonomously generated route; wherein said mobile line marking robot further comprises:

[0020] - one camera or two cameras arranged in a stereo vision configuration;

[0021] - a second processor operably connected to said camera(s);

[0022] - a first memory coupled to the second processor, wherein the first memory comprises program instructions executable by the second processor for: i) receiving respective image data streams from said camera(s); ii) processing received image data from the / each camera, or from both cameras simultaneously from two different points in time to identify one or more objects present in all image data from both points in time; iii) utilizing said identified one or more objects to estimate the displacement of the mobile line marking robot from one point in time to another point in time, and determining its relative change in position and / or direction from said first point in time to said second point in time; iv) comparing the determined change in mobile line marking robot position and / or direction with received historic GNSS position data and / or current and / or historic inertia sensor data and / or historic wheel odometry data; and v) informing said first processor and / or said motor controller to compensate for identified unexpected movement of the mobile line marking robot or identified error in position of the robot.

[0023] A second aspect relates to a mobile line marking robot comprising:

[0024] - a chassis with one or more wheels operably connected via a motor controller to at least one motor;

[0025] - a spray means comprising a spray nozzle with an outlet;

[0026] - a paint reservoir;

[0027] - a pump unit operably connected to said spray means and said paint reservoir, thereby allowing paint from the paint reservoir to exit said spray nozzle;

[0028] - a positioning system receiver unit configured for receiving a positioning signal from a Global Navigation Satellite System (GNSS);

[0029] - a first processor configured to receive the positioning information signal from the positioning system receiver unit and in response thereto a) to activate or deactivate the pump unit and / or b) to instruct the motor controller to navigate a predefined or autonomously generated route;

[0030] - two cameras arranged in a stereo vision configuration;

[0031] - a second processor operably connected to said two cameras;

[0032] - a first memory coupled to the second processor, wherein the first memory comprises program instructions executable by the second processor for: i) receiving respective image data streams from said two cameras; ii) processing received image data from each camera individually, or from both cameras simultaneously from two different points in time to identify one or more objects present in all image data from both points in time; iii) utilizing said identified one or more objects to estimate the displacement of the mobile line marking robot from a first point in time to a second point in time, and determining its relative change in position and / or direction from said first point in time to said second point in time; iv) comparing the determined change in mobile line marking robot position and / or direction with received historic GNSS position data and / or current and / or historic inertia sensor data and / or historic wheel odometry data; and v) informing said first processor and / or said motor controller to compensate for identified unexpected movement of the mobile line marking robot or identified error in position of the robot.

[0033] Detailed description of the invention

[0034] A first aspect relates to a mobile line marking robot comprising:

[0035] - a chassis with one or more wheels operably connected via a motor controller to at least one motor;

[0036] - a spray means comprising a spray nozzle with an outlet;

[0037] - a paint reservoir;

[0038] - a pump unit operably connected to said spray means and said paint reservoir, thereby allowing paint from the paint reservoir to exit said spray nozzle;

[0039] - a positioning system receiver unit configured for receiving a positioning signal from a Global Navigation Satellite System (GNSS); and

[0040] - a first processor configured to receive the positioning information signal from the positioning system receiver unit and in response thereto a) to activate or deactivate the pump unit and / or b) to instruct the motor controller to navigate a predefined or autonomously generated route; wherein said mobile line marking robot further comprises:

[0041] - one camera or two cameras arranged in a stereo vision configuration;

[0042] - a second processor operably connected to said camera(s);

[0043] - a first memory coupled to the second processor, wherein the first memory comprises program instructions executable by the second processor for: i) receiving respective image data streams from said camera(s); ii) processing received image data from the / each camera, or from both cameras simultaneously from two different points in time to identify one or more objects present in all image data from both points in time; iii) utilizing said identified one or more objects to estimate the displacement of the mobile line marking robot from one point in time to another point in time, and determining its relative change in position and / or direction from said first point in time to said second point in time; iv) comparing the determined change in mobile line marking robot position and / or direction with received historic GNSS position data and / or current and / or historic inertia sensor data and / or historic wheel odometry data; and v) informing said first processor and / or said motor controller to compensate for identified unexpected movement of the mobile line marking robot or identified error in position of the robot.

[0044] A second aspect relates to a mobile line marking robot comprising:

[0045] - a chassis with one or more wheels operably connected via a motor controller to at least one motor;

[0046] - a spray means comprising a spray nozzle with an outlet;

[0047] - a paint reservoir;

[0048] - a pump unit operably connected to said spray means and said paint reservoir, thereby allowing paint from the paint reservoir to exit said spray nozzle;

[0049] - a positioning system receiver unit configured for receiving a positioning signal from a Global Navigation Satellite System (GNSS);

[0050] - a first processor configured to receive the positioning information signal from the positioning system receiver unit and in response thereto a) to activate or deactivate the pump unit and / or b) to instruct the motor controller to navigate a predefined or autonomously generated route;

[0051] - two cameras arranged in a stereo vision configuration;

[0052] - a second processor operably connected to said two cameras;

[0053] - a first memory coupled to the second processor, wherein the first memory comprises program instructions executable by the second processor for: i) receiving respective image data streams from said two cameras; ii) processing received image data from each camera individually, or from both cameras simultaneously from two different points in time to identify one or more objects present in all image data from both points in time; iii) utilizing said identified one or more objects to estimate the displacement of the mobile line marking robot from a first point in time to a second point in time, and determining its relative change in position and / or direction from said first point in time to said second point in time; iv) comparing the determined change in mobile line marking robot position and / or direction with received historic GNSS position data and / or current and / or historic inertia sensor data and / or historic wheel odometry data; and v) informing said first processor and / or said motor controller to compensate for identified unexpected movement of the mobile line marking robot or identified error in position of the robot.

[0054] In one or more embodiments, the compensation in step v) includes instructing the motor controller of the mobile line marking robot to correct the position and / or direction of the mobile line marking robot.

[0055] In one or more embodiments, the compensation in step v) includes instructing the first processor to activate or deactivate the pump unit in response thereto.

[0056] The program instructions executable by the second processor may include the use of a computer vision algorithm that facilitates various functionalities, such as object detection and image segmentation by analyzing the pixel output from the cameras. By tracking the motion of these distinctive points, the system can calculate visual odometry, which provides information about the mobile line marking robot’s movement. The system can flexibly process data from each camera individually, both cameras simultaneously, or switch between the two as necessary, thereby improving fault tolerance and overall system reliability.

[0057] When two cameras are present, the two cameras are preferably arranged side by side and pointing in the same direction. The cameras will simultaneously take one image each and send to the second processor, or to a memory coupled to the second processor.

[0058] The second processor will preferably rectify the image(s) to compensate for distortions from e.g., the lens.

[0059] Rectifying an image to compensate for distortions from the lens of a camera may e.g., involve a process called image distortion correction or lens distortion correction. This correction may be done by the aid of software algorithms that manipulate the pixel values in the image to counteract the effects of lens distortion. The process may involve the steps of initial calibration of the camera, distortion mapping, warping, interpolation, image cropping, and final adjustments. All these steps may not necessarily be used for the present invention but may be used. The steps are briefly discussed in the following. The first step is to calibrate the camera(s) and determine the type and degree of distortion introduced by the lens(es). Different lenses may produce different types of distortions, such as barrel distortion (outward bending), or pincushion distortion (inward bending). The calibration process involves capturing images of known calibration patterns or grids. Once the distortion model is determined, a distortion map is created. This map defines how each pixel in the original image should be moved or transformed to correct for the distortion. Each pixel's new position is calculated based on its original position and the distortion model. The distortion map is then used to warp the original image. This involves applying a geometric transformation to each pixel based on the calculated distortion map. This transformation effectively undoes the distortions introduced by the lens. During the warping process, some pixels might end up in non-integer positions, which can lead to gaps or overlaps in the corrected image.

[0060] Interpolation techniques, such as bilinear or bicubic interpolation, are used to estimate the pixel values at these non-integer positions. Depending on the distortion correction method and the desired output, the corrected image may have areas that extend beyond the original image boundaries. These areas can be cropped to obtain the final distortion-corrected image. After distortion correction, further adjustments may be applied, such as color correction, contrast enhancement, and sharpening, to ensure the final image quality meets the desired standards. The second processor, that may comprise multiple processors, play a crucial role in performing these computations efficiently. The steps involved in distortion correction require substantial computational resources, especially for higher-resolution images and complex distortion models.

[0061] In the field of image processing, there exists a diverse array of rectification and correction techniques, each designed to address specific aspects of image quality and alignment. These techniques serve to enhance the utility and visual fidelity of digital images, catering to a range of applications. Exemplary techniques are briefly discussed in the following. These techniques may be used for the present invention.

[0062] Geometric Rectification is used to correct geometric distortions in images that arise due to factors such as camera angle, perspective, and projection. Geometric rectification is commonly used to transform images so that they align with a desired reference frame. It involves applying geometric transformations like rotation, scaling, and translation to the image.

[0063] Perspective distortion occurs when objects in the scene appear distorted due to the camera's viewpoint. Perspective rectification aims to remove this distortion by transforming the image to a new perspective, often making objects appear as if they were captured from a frontal viewpoint. This technique is useful for tasks like object recognition.

[0064] Panorama stitching involves combining multiple images of a scene taken from different viewpoints into a single seamless panorama. This process requires correcting for geometric and photometric differences between the images, aligning their features, and blending them together. It often involves distortion correction, warping, and blending techniques to create a visually an accurate panorama.

[0065] Chromatic aberration is a type of distortion that causes different colors of light to focus at slightly different points, resulting in color fringing around edges in an image. Correction techniques involve aligning color channels and shifting them to compensate for the color shifts caused by the lens.

[0066] Vignetting is the gradual darkening of an image towards the comers due to optical limitations of the lens. Vignetting correction techniques aim to even out the brightness across the image by applying correction factors to each pixel based on its position.

[0067] Histogram equalization is a technique used to adjust the distribution of pixel intensities in an image. It can be used to enhance the contrast and improve the visual quality of an image by redistributing the pixel values across the entire intensity range.

[0068] Noise reduction techniques are often used to improve image quality by reducing unwanted noise or artifacts caused by sensor limitations or other factors. These techniques involve filtering and smoothing the image to reduce noise while preserving important details.

[0069] Color correction techniques are used to adjust and balance colors in an image to achieve a desired look or to correct color casts caused by lighting conditions or camera settings.

[0070] The second processor will then search for distinctive features in both images, compare the features found and match feature pairs. From these pairs a 3D point cloud can be calculated. The point cloud and the respective features are then saved for the next frame. By obtaining two sets of point clouds with identified features, it is now possible to compare those point clouds. The displacement of the matching feature’s 3D points is then the displacement of the robot.

[0071] The identification of distinctive features in both images is a task commonly known as object recognition or image classification and may involve using a combination of hardware and software techniques. The first step is as already described image rectification. The second processor first receives the raw pixel data from an image, usually represented as a matrix of values corresponding to the color or intensity of each pixel. Rectification techniques, e.g., as described above, are applied to enhance the quality of the input data. These might include resizing the image, normalizing pixel values, and applying various filters to highlight important features. Then the feature extraction step is initiated, which involves identifying and selecting relevant characteristics or patterns in the image that are useful for distinguishing different objects. Common techniques include edge detection, corner detection, texture analysis, and local feature extraction (such as SIFT, SURF, or HOG). Once relevant features are extracted, they may in some embodiments be converted into a format suitable for machine learning algorithms. Vector representations, histograms, or other numerical representations are commonly used. Object recognition systems may utilize machine learning algorithms, particularly deep learning models like Convolutional Neural Networks (CNNs). These algorithms are trained on large datasets with labeled examples of various objects. During training, the algorithm learns to recognize patterns and features that are indicative of specific objects. The trained model is composed of layers of interconnected nodes that progressively learn higher-level features from raw pixel values. Once the model is trained, it can be used for inference on new, unseen images. The processor feeds the preprocessed image data through the trained model, and the model computes a probability distribution over possible object classes. The class with the highest probability is considered the model’s prediction for the object present in the image. Post-processing steps might include filtering out low-confidence predictions, refining the localization of detected objects, and handling overlapping objects. Visualization techniques are used to draw bounding boxes, labels, or other annotations on the original image to indicate the identified objects. The second processor may be equipped with specialized hardware accelerators, like Graphics Processing Units (GPUs) or dedicated Al accelerators (such as TPUs), to efficiently perform the computations required by complex machine learning models, especially deep neural networks.

[0072] In some embodiments, the second processor may be configured to save processed images for reference and compare such an image or images with subsequent received images. If an identified feature has moved in the 2D image, then the second processor assumes that the robot has turned. The identified features derived from pixels in the images are thus derived from the surroundings, and may e.g., be a goal post, a corner flag post, grass straw, bushes, trees, a golf green, a boundary of a golf green, or other characteristic objects in the surroundings.

[0073] Thus, it is now possible to compare the calculated robot position, measured from IMU, wheel odometry, and GNSS positions, with the measured visual odometry and compensate for unexpected movement or error in position. This comparison may e.g., be performed utilizing Kalman filtering.

[0074] Kalman filtering is a mathematical technique e.g., used in robotics and motion planning to estimate the state of a system based on uncertain measurements. In the context of robotic motion planning, Kalman filtering can play an important role in improving the accuracy of state estimation, which is essential for making informed decisions about how a robot should move and interact with its environment.

[0075] In robotic motion planning, the state of a robot includes its position, velocity, orientation, and possibly other relevant variables. The actual state of a robot is often not directly observable due to sensor noise and environmental uncertainties. Kalman filtering helps in estimating the true state by fusing measurements from sensors (such as GPS, inertial sensors, encoders, etc.) with the robot's motion model. The present invention provides an extra information for this estimation by including visual odometry. Hence, Kalman filtering allows the fusion of data from multiple sensors to obtain a more precise estimate of the robot's position and orientation.

[0076] Motion planning often involves predicting the future state of a robot based on its current state and control inputs. Kalman filtering incorporates the robot's dynamic model to predict how the state will evolve over time. By continuously updating these predictions with new sensor measurements, the Kalman filter refines its estimate of the robot's state. Again, the present invention provides an extra information for this estimation by including visual odometry.

[0077] Kalman filtering can be integrated into the process of path planning and navigation. As the robot moves along a planned path, the filter helps correct any deviations from the expected trajectory due to unexpected factors like sensor noise, external disturbances, or unmodeled dynamics.

[0078] Kalman filtering can also aid in obstacle avoidance. By estimating the robot's state accurately, it becomes possible to predict potential collisions with obstacles in the environment. This information can be used to adjust the robot's trajectory in real-time, ensuring safe navigation.

[0079] Kalman filtering allows for adaptive control strategies where the control inputs are adjusted based on the estimated state and sensor measurements. This adaptability is especially important in dynamic and uncertain environments.

[0080] Kalman filtering also provides a mechanism for feedback control by continuously comparing the predicted state with the actual measurements. If a significant discrepancy is detected, the filter can update the state estimate and influence the robot's control actions accordingly.

[0081] There are also alternatives to Kalman filtering that could also benefit from visual odometry input. The choice of technique depends on the specific requirements of the mobile robot's task, the nature of its environment, and the available sensor suite. The following techniques are commonly used for mobile robots.

[0082] Extended Kalman Filter (EKF) is often used for mobile robot localization and state estimation when the system dynamics or measurements are moderately nonlinear. It's commonly employed in scenarios where real-time performance is essential, such as wheeled robots.

[0083] Unscented Kalman Filter (UKF) is preferred when dealing with highly nonlinear systems, as it provides more accurate estimates than EKF. Mobile robots operating in complex environments or with non-trivial dynamics may benefit from UKF for state estimation.

[0084] Particle Filters (Monte Carlo Localization) are extensively used in mobile robot localization tasks, particularly in scenarios where there's high uncertainty, nonGaussian noise, and complex motion dynamics. They can handle multimodal distributions and are useful for robustly estimating the robot's pose.

[0085] Graph-Based SLAM (Simultaneous Localization and Mapping) may also be used to build maps of the mobile robot’s environment while simultaneously estimating their pose within the map. This is crucial for autonomous navigation and exploration tasks.

[0086] Iterative Closest Point (ICP) Algorithm is frequently employed in mobile robot mapping and localization, especially when integrating data from sensors like LIDAR to align and update maps.

[0087] Bayesian Networks and Probabilistic Graphical Models can be used for various tasks in mobile robotics, including perception, decision-making, and localization. They provide a way to represent complex probabilistic relationships and uncertainties in the robot's environment.

[0088] Gaussian processes are applied to mobile robot motion planning and sensor fusion tasks, where uncertainty modeling is crucial. They can help make informed decisions while navigating through uncertain or dynamic environments.

[0089] It's important to note that the choice of technique depends on factors like computational resources, the complexity of the problem, the available sensor suite, and the desired level of accuracy. In practice, many mobile robotic systems use a combination of these techniques, along with other methods, to achieve robust and reliable operation.

[0090] The two cameras also enable 3D capabilities to the mobile line marking robot. In this embodiment, the visual odometry now becomes a subsystem to be used along with the 3D matching subsystem. The 3D matching subsystem will capture images from both cameras simultaneously and perform a comparative analysis of the identified features. Utilizing trigonometry, the system can determine the position of each feature and generate a 3D point cloud. This point cloud can be utilized as an obstacle sensor, or to identify specific features, such as a goal post.

[0091] In one or more embodiments, the mobile line marking robot further comprises an obstacle sensor function, wherein the mobile line marking robot comprises a third processor coupled to said first memory, wherein the first memory comprises program instructions executable by the third processor for:

[0092] - utilizing said identified one or more objects to identify obstacles on a planned path for said mobile line marking robot, and

[0093] - if an obstacle is identified, informing said first processor and / or said motor controller to compensate for the presence of said obstacle.

[0094] Non-limiting examples of obstacles could e.g., be persons, animals, holes, fallen tree branches, molehills, sports posts, such as goal posts, or the like. The compensation process may e.g., be to avoid a collision by deviating from the mobile line marking robot’s planned path, momentarily stopping the mobile line marking robot until the obstacle is gone / out of sight, and at the same time deactivating the pump unit.

[0095] The first, second, third, and fourth processors may in some embodiments be the same processor, and in other embodiments be different processors.

[0096] Preferably, the two cameras are placed on the front end of the mobile line marking robot, preferably opposite to the spray nozzle i.e. , the spray nozzle is placed at the rear end of the mobile line marking robot.

[0097] In one or more embodiments, the mobile line marking robot further comprises a target object identification function, wherein the mobile line marking robot comprises a fourth processor coupled to said first memory, wherein the first memory comprises program instructions executable by the fourth processor for:

[0098] - utilizing said identified one or more objects to identify target objects, such as goal posts, old paint lines, or areas with sick grass, and

[0099] - if a target obstacle is identified, informing said first processor and / or said motor controller to compensate for the presence of said target obstacle.

[0100] In one or more embodiments, the identified target object is an old paint line, and wherein the compensation includes following said old paint line, possibly deviating from a predefine route / path. In one or more embodiments, the identified old paint line is followed until a complete structure, such as a sports field, defined by said paint line is identified.

[0101] In one or more embodiments, the identified old paint line is refreshed, either simultaneously or subsequently.

[0102] In one or more embodiments, the identified old paint line is followed until a complete structure, such as a sports field, defined by said paint line is identified, and wherein said identification is subsequently used for calibration of the mobile line marking robot.

[0103] It should be noted that embodiments and features described in the context of one of the aspects of the present invention also apply to the other aspects of the invention.

[0104] As used in the specification and the appended claims, the singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from "about" or "approximately" one particular value and / or to "about" or "approximately" another particular value. When such a range is expressed, another embodiment includes from the one particular value and / or to the other particular value. Similarly, when values are expressed as approximations, by use of the antecedent "about", it will be understood that the particular value forms another embodiment.

Claims

Claims1 . A mobile line marking robot comprising:- a chassis with one or more wheels operably connected via a motor controller to at least one motor;- a spray means comprising a spray nozzle with an outlet;- a paint reservoir;- a pump unit operably connected to said spray means and said paint reservoir, thereby allowing paint from the paint reservoir to exit said spray nozzle;- a positioning system receiver unit configured for receiving a positioning signal from a Global Navigation Satellite System (GNSS); and- a first processor configured to receive the positioning information signal from the positioning system receiver unit and in response thereto a) to activate or deactivate the pump unit and / or b) to instruct the motor controller to navigate a predefined or autonomously generated route; characterized in that said mobile line marking robot further comprises:- one camera or two cameras arranged in a stereo vision configuration;- a second processor operably connected to said camera(s);- a first memory coupled to the second processor, wherein the first memory comprises program instructions executable by the second processor for: i) receiving respective image data streams from said camera(s); ii) processing received image data from the / each camera, or from both cameras simultaneously from two different points in time to identify one or more objects present in all image data from both points in time; iii) utilizing said identified one or more objects to estimate the displacement of the mobile line marking robot from one point in time to another point in time, and determining its relative change in position and / or direction from said first point in time to said second point in time; iv) comparing the determined change in mobile line marking robot position and / or direction with received historic GNSS position data and / or current and / or historic inertia sensor data and / or historic wheel odometry data; andv) informing said first processor and / or said motor controller to compensate for identified unexpected movement of the mobile line marking robot or identified error in position of the robot.

2. The mobile line marking robot according to claim 1 , wherein the compensation in step v) includes instructing the motor controller of the mobile line marking robot to correct the position of the mobile line marking robot.

3. The mobile line marking robot according to any one of the claims 1-2, wherein the compensation in step v) includes instructing the motor controller of the mobile line marking robot to correct the direction of the mobile line marking robot.

4. The mobile line marking robot according to any one of the claims 1-3, wherein the compensation in step v) includes instructing the first processor to activate or deactivate the pump unit in response thereto.

5. The mobile line marking robot according to any one of the claims 1-4, wherein the mobile line marking robot comprises two cameras arranged in a stereo vision configuration.

6. The mobile line marking robot according to claim 5, further comprising an obstacle sensor function, wherein the mobile line marking robot comprises a third processor coupled to said first memory, wherein the first memory comprises program instructions executable by the third processor for:- utilizing said identified one or more objects to identify obstacles on a planned path for said mobile line marking robot, and- if an obstacle is identified, informing said first processor and / or said motor controller to compensate for the presence of said obstacle.

7. The mobile line marking robot according to any one of the claims 5-6, further comprising a target object identification function, wherein the mobile line markingrobot comprises a fourth processor coupled to said first memory, wherein the first memory comprises program instructions executable by the fourth processor for:- utilizing said identified one or more objects to identify target objects, such as goal posts, old paint lines, or areas with sick grass, and- if a target obstacle is identified, informing said first processor and / or said motor controller to compensate for the presence of said target obstacle.

8. The mobile line marking robot according to claim 7, wherein the identified target object is an old paint line, and wherein the compensation includes following said old paint line, possibly deviating from a predefine route / path.

9. The mobile line marking robot according to claim 8, wherein the identified old paint line is followed until a complete structure, such as a sports field, defined by said paint line is identified.

10. The mobile line marking robot according to any one of the claims 8-9, wherein the identified old paint line is refreshed, either simultaneously or subsequently.11 . The mobile line marking robot according to claim 8, wherein the identified old paint line is followed until a complete structure, such as a sports field, defined by said paint line is identified, and wherein said identification is subsequently used for calibration of the mobile line marking robot.