Indoor decoration robot autonomous obstacle avoidance method based on artificial intelligence

Abstract

The invention discloses an indoor decoration robot autonomous obstacle avoidance method based on artificial intelligence, and particularly relates to the technical field of robot autonomous obstacle avoidance. And calculating a perception quality reduction abnormal coefficient and a motion control distortion coefficient to construct an obstacle avoidance hysteresis fatal collision risk assessment model, and accurately identifying a fatal collision high-risk state possibly faced by the robot in a complex and dynamic indoor environment based on an obstacle avoidance hysteresis fatal collision risk assessment index. A risk identification blind area of an existing obstacle avoidance strategy in a scene with coupling defects of sensing pollution and controlling instability is effectively made up; according to the method, control delay cost and path length cost are adaptively introduced in a high-risk state, an updating target function is constructed, path multi-target optimization is executed, dynamic correction of an original path of the robot is achieved, and path smoothness and energy consumption control can be considered while obstacle avoidance safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of robot autonomous obstacle avoidance, and more specifically, to an artificial intelligence-based autonomous obstacle avoidance method for an interior decoration robot. Background Art

[0002] With the development of intelligent interior decoration equipment, AI-powered interior decoration robots are widely used in automated multi-process operations such as painting, paving, and polishing. To autonomously complete tasks in complex and dynamic indoor environments, obstacle avoidance, as a core safety mechanism, generally integrates multimodal perception and intelligent decision-making modules such as deep learning-based object recognition, semantic segmentation, and path replanning. However, current mainstream autonomous obstacle avoidance methods are mostly based on the assumption of ideal sensor input and stable power control, making it difficult to address the coupled defect scenarios of perception anomalies and control instability during robot operation. Specifically, during long-term continuous operation, robot sensors (such as RGB-D cameras and lidar) are susceptible to interference from dust, paint, and reflections, resulting in partial occlusion or contamination of visual or point cloud inputs, significantly degrading perception quality. Furthermore, when the system reaches a low-battery state, the power output and motion response rate of the drive module are significantly reduced, causing the actual motion to deviate significantly from the planned trajectory. When the two types of problems mentioned above occur simultaneously, the obstacle avoidance system will face a series of system-level coupling risks such as path prediction distortion, increased response lag, and frequent recognition misjudgments, which can easily lead to delayed obstacle avoidance collisions between complex furniture and near edge obstacles. What is more serious is that such compound defects are usually at the edge of AI training data and lack effective historical sample support. Existing path planning algorithms and policy networks are difficult to generalize to such states, which ultimately leads to unstable obstacle avoidance model strategy outputs and uncontrolled robot behaviors such as "freezing" or "repeated detection". To address the above challenges, it is urgent to design an intelligent obstacle avoidance method with pollution perception tolerance and low-battery behavior compensation mechanism to improve the robot's safety robustness and path correction capabilities in high-intensity, long-term decoration tasks. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence includes the following steps:

[0006] Step S1, obtaining perception quality degradation information of a robot sensor and motion control distortion information of the robot in a low-battery state, wherein the perception quality degradation information includes a perception quality degradation abnormality coefficient, and the motion control distortion information includes a motion control distortion coefficient;

[0007] Step S2: constructing an obstacle avoidance hysteresis fatal collision risk assessment model based on the perception quality degradation abnormality coefficient and the motion control distortion coefficient, outputting an obstacle avoidance hysteresis fatal collision risk assessment index, and evaluating the risk level of the robot in the current state of an obstacle avoidance hysteresis fatal collision;

[0008] Step S3, comparing the obstacle avoidance hysteresis fatal collision risk assessment index with a preset obstacle avoidance hysteresis fatal collision risk assessment index threshold to identify the high risk level of the obstacle avoidance hysteresis fatal collision occurring in the current state of the robot;

[0009] In step S4, if the robot is at a high risk of a fatal collision due to delayed obstacle avoidance, an updated objective function is constructed based on the control delay cost and the path length cost, and the original path is adaptively optimized with multiple objectives.

[0010] In a preferred embodiment, the logic for obtaining the abnormal coefficient of perceived quality degradation is as follows:

[0011] Perform missing detection on the two-dimensional space of the environmental perception data collected by the robot sensor, and construct the data integrity matrix Mint(t) in the two-dimensional space (x, y) and at time t. The elements in the data integrity matrix are: And calculate the data integrity ratio: Where Fint(t) is the data integrity ratio;

[0012] For noise interference in point cloud data, the noise difference is defined as: Where Tnosie(x,y,t) is the noise difference, XI(x,y,t) is the actual collected signal strength, is the ideal signal strength after filtering; calculate the overall noise interference intensity: Where Fnosie(t) is the overall noise interference intensity;

[0013] For blur measurement in image data, define blurriness: Where Mhda(x,y,t) is the blur, IT(x,y,t) is the image grayscale in two-dimensional space (x,y) and time t, is the Laplace operator output of the image at (x,y) coordinates, is the global maximum value of the IT Laplacian operator output of the current image; the number of pixels whose blurriness at the current time t is greater than the preset blurriness threshold is counted as the blur abnormality factor Fblur(t);

[0014] Calculate the timing anomaly factor: ΔSX(t)

[0015] =max{|Fint(t)-Fint(t-1)|,|Fnosie(t)-Fnosie(t-1)|,|Fblur(t)-Fblur)|};

[0016] (t-1

[0017] Construct the local coupling anomaly matrix Clocal(t): Clocal(t) = Mint(t)⊙Tnosie(t)⊙Mhda(t), where Tnosie(t) represents the noise difference matrix with noise differences Tnosie(x, y, t) as matrix elements, Mhda(t) represents the ambiguity matrix with ambiguities Mhda(x, y, t) as matrix elements, and ⊙ represents the Hadamard product representation of the matrix.

[0018] Calculate the local coupling anomaly coefficient: Where Flocal(t) is the local coupling anomaly coefficient, Clocal(x,y,t) is the element in the local coupling anomaly matrix Clocal(t);

[0019] Calculate the causal dynamic coefficient: Where YG(t) is the dynamic coefficient of causality, Gcau(x,y,t) is the Granger causal factor field generated based on historical multi-time data, representing the causal relationship at different spatial locations;

[0020] Calculate the perceptual quality degradation anomaly coefficient: Wherein θ1 is the preset timing anomaly factor threshold, and θ2 is the preset local coupling anomaly coefficient threshold.

[0021] In a preferred embodiment, the logic for obtaining the motion control distortion coefficient is as follows:

[0022] During the robot's motion, the following key motion data are collected in real time:

[0023] Actual motion state sequence Sact(t): includes position vector Pact(t) = {Xact(t), Yact(t), Zact(t)}, attitude angle θact(t), velocity Vact(t);

[0024] Expected motion state sequence Sref(t): includes the expected position Pref(t) = {Xref(t), Yref(t), Zref(t)}, attitude angle θref(t), and velocity Vref(t) expected by the control system;

[0025] Calculate the local posture deviation: Δθz(t) = |θact(t) - θref(t)|, where Δθz(t) is the local posture deviation;

[0026] Calculate the path deviation distance: Where Dpath(t) is the path deviation distance;

[0027] Calculate the dynamic response distortion: Where Δdyn(t) is the dynamic response distortion;

[0028] Calculate the spatiotemporal cumulative distortion coefficient within the time window [t0,t]: Where Mctrl(t) is the spatiotemporal cumulative distortion coefficient, τ is the time index, τ∈[t0,t], Dmax is the maximum allowable path deviation reference, a1, a2, and a3 represent the preset proportional coefficients of local posture deviation, path deviation distance, and dynamic response distortion, respectively, and a1, a2, and a3 are all greater than 0;

[0029] Calculate the motion control distortion coefficient: Yssz(t)=min(1,Mctrl(t)), where Yssz(t) is the motion control distortion coefficient.

[0030] In a preferred embodiment, an obstacle avoidance hysteresis fatal collision risk assessment model is constructed based on the perception quality degradation abnormal coefficient and the motion control distortion coefficient, and an obstacle avoidance hysteresis fatal collision risk assessment index is output. The obstacle avoidance hysteresis fatal collision risk assessment model is based on the following formula: bzzs(t)=w1*Cpq(t)+w2*Yssz(t), where bzzs(t) is the obstacle avoidance hysteresis fatal collision risk assessment index, Cpq(t) is the perception quality degradation abnormal coefficient, Yssz(t) is the motion control distortion coefficient, w1 and w2 represent the preset proportional coefficients of the perception quality degradation abnormal coefficient and the motion control distortion coefficient, respectively, and w1 and w2 are both greater than 0.

[0031] In a preferred embodiment, the obstacle avoidance hysteresis fatal collision risk assessment index is compared with a preset obstacle avoidance hysteresis fatal collision risk assessment index threshold to identify the high risk level of the robot in the current state of the obstacle avoidance hysteresis fatal collision, as follows:

[0032] If the obstacle avoidance hysteresis fatal collision risk assessment index is greater than the obstacle avoidance hysteresis fatal collision risk assessment index threshold, the risk level of the robot in the current state causing obstacle avoidance hysteresis fatal collision is marked as high risk level;

[0033] If the obstacle avoidance hysteresis fatal collision risk assessment index is less than or equal to the obstacle avoidance hysteresis fatal collision risk assessment index threshold, there is no need to mark the risk level of the obstacle avoidance hysteresis fatal collision of the current state robot.

[0034] In a preferred embodiment, if the robot is at a high risk of a fatal collision due to delayed obstacle avoidance, an updated objective function Jnew is constructed based on the control delay cost and the path length cost: Among them, J ρ To control the delay cost, is the path length cost, w ρ 、 They represent the preset proportional coefficients for controlling the delay cost and the path length cost, respectively, and w ρ 、 Both are greater than 0.

[0035] In a preferred embodiment, the control delay cost J ρ :J ρ =∫ρ(t)*s(t)dt, where ρ(t) is the control delay curve of the current system and s(t) is the path arc length speed;

[0036] Path length cost Where P(t) is the length of the robot's current motion path.

[0037] In a preferred embodiment, the updated objective function Jnew is used as the optimization goal to solve the new path: Where st is the constraint condition, v(t) is the robot's motion speed, and v min is the minimum speed, v max is the maximum velocity, js(t) is the angular velocity, js max is the maximum angular velocity, WZ(0) is the starting position of the current path, WZ(t js ) is the end position of the current path.

[0038] Technical effects and advantages of the present invention:

[0039] 1. The present invention provides an autonomous obstacle avoidance method for an indoor decoration robot based on artificial intelligence. By integrating the perception quality degradation information and the low-power motion control distortion information, the method realizes the dynamic evaluation of the robot's obstacle avoidance hysteresis fatal collision risk and path adaptive optimization. The method can significantly improve the safety robustness and path correction ability of the indoor decoration robot in a high-intensity, long-cycle operation environment. The specific effects are reflected in the following aspects: On the one hand, the present invention can obtain and quantify the robot's sensor perception anomaly and control distortion state in real time and calculate the perception quality degradation anomaly coefficient and motion control distortion coefficient to construct an obstacle avoidance hysteresis fatal collision risk assessment model. Based on the obstacle avoidance hysteresis fatal collision risk assessment index The system accurately identifies the high-risk states of fatal collisions that robots may face in complex and dynamic indoor environments, effectively making up for the risk identification blind spots of existing obstacle avoidance strategies in scenarios with coupling defects of perception pollution and control instability. On the other hand, the present invention adaptively introduces control delay cost and path length cost in high-risk states, constructs an updated objective function and performs multi-objective optimization of the path, thereby realizing dynamic correction of the robot's original path. It can take into account path smoothness and energy consumption control while ensuring obstacle avoidance safety, avoid uncontrolled behaviors such as robot motion freezing and repeated detection caused by instability of obstacle avoidance strategy, and significantly improve the autonomous obstacle avoidance efficiency and stability of interior decoration robots in areas with complex furniture layouts and dense obstacles. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0041] Figure 1 Flowchart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] Example: Figure 1 The present invention provides an autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence, comprising the following steps:

[0044] Step S1, obtaining perception quality degradation information of a robot sensor and motion control distortion information of the robot in a low-battery state, wherein the perception quality degradation information includes a perception quality degradation abnormality coefficient, and the motion control distortion information includes a motion control distortion coefficient;

[0045] Step S2: constructing an obstacle avoidance hysteresis fatal collision risk assessment model based on the perception quality degradation abnormality coefficient and the motion control distortion coefficient, outputting an obstacle avoidance hysteresis fatal collision risk assessment index, and evaluating the risk level of the robot in the current state of an obstacle avoidance hysteresis fatal collision;

[0046] Step S3, comparing the obstacle avoidance hysteresis fatal collision risk assessment index with a preset obstacle avoidance hysteresis fatal collision risk assessment index threshold to identify the high risk level of the obstacle avoidance hysteresis fatal collision occurring in the current state of the robot;

[0047] Step S4: If the robot is at a high risk of a fatal collision due to delayed obstacle avoidance, an updated objective function is constructed based on the control delay cost and the path length cost, and adaptive multi-objective optimization is performed on the original path.

[0048] Step S1, obtaining perception quality degradation information of a robot sensor and motion control distortion information of the robot in a low-battery state, wherein the perception quality degradation information includes a perception quality degradation abnormality coefficient, and the motion control distortion information includes a motion control distortion coefficient;

[0049] The perception quality degradation anomaly coefficient in the present invention is an important indicator for measuring the degree of degradation of the quality of the perception data of the robot's multimodal sensors (such as RGB-D cameras, lidars, ultrasonic sensors, etc.) in a dynamic and complex interior decoration environment. It can accurately reflect the effective working state of the robot's perception system under the current environmental conditions. The larger the perception quality degradation anomaly coefficient is, the more serious the degree of contamination or interference of the sensor input in the robot's current environment. For example, due to factors such as dust, paint mist, reflective surfaces, and obstructions, there are a lot of distortions, information missing, or mismatches in the visual or point cloud data. This will directly affect the accuracy and response speed of core functions such as target detection, boundary extraction, and path replanning in the obstacle avoidance model, thereby greatly increasing the risk of path deviation, delayed obstacle avoidance, etc. caused by misjudgment, missed judgment, or delayed response of the robot in a complex environment, which may eventually lead to fatal collision accidents. Conversely, a smaller perception quality degradation anomaly coefficient indicates that the quality of the environmental perception data currently being acquired by the robot's sensors is high. The multimodal input data is clear, has minimal occlusion and interference, and exhibits low noise levels. This allows the robot to accurately identify various obstacles, operational boundaries, and dynamic interference sources in the environment, and to perform path planning and real-time corrections based on this high-quality data. In this state, the robot can stably perform obstacle avoidance tasks in complex operating environments, significantly reducing path prediction distortion and control response deviations caused by perception errors or processing delays, effectively improving operational safety and path planning robustness. The perception quality degradation anomaly coefficient is used to assess the risk of the robot's current state. This coefficient can serve as one of the core inputs in the robot's obstacle avoidance delayed fatal collision risk model, forming a joint assessment system with the motion control distortion coefficient to quantify the dynamic risk of the robot under dual perception and control degradation states. By monitoring the perception quality degradation anomaly coefficient in real time, the present invention provides early warning when the robot enters a marginal perception quality state, preventing the robot from entering an uncontrollable state as perception distortion gradually increases, thereby reducing the blind spots created by traditional obstacle avoidance strategies due to the lack of real-time perception degradation monitoring. On the other hand, the dynamic risk assessment and adaptive path optimization process driven by this coefficient can significantly improve the adaptability of the robot system to environmental changes. The introduction of the perception quality degradation anomaly coefficient greatly enhances the safety robustness of the robot system in the face of complex decoration environments and long-term high-load operation scenarios. Through the dynamic calculation and risk feedback of this coefficient, the present invention realizes the transformation of the obstacle avoidance strategy from the traditional "passive emergency response" to the "active prevention" mode, and provides a new intelligent obstacle avoidance mechanism that can be quantified, intervened, and optimized for interior decoration robots in high-risk processes such as painting, paving, and polishing. It fundamentally improves the safety and autonomous adaptability of the robot obstacle avoidance system under complex working conditions, significantly reduces the risk of fatal collisions at the system level, and has significant technical application value and industrial promotion prospects.

[0050] The logic for obtaining the abnormal coefficient of perceived quality degradation is as follows:

[0051] Perform missing value detection in two-dimensional space on the environmental perception data (RGB images, point clouds, etc.) collected by the robot sensor (data missing value detection methods such as hole pixel recognition, point cloud sparsity analysis, abnormal confidence threshold screening, etc., which are not described here), and construct the data integrity matrix Mint(t) in two-dimensional space (x, y) and at time t. The elements in the data integrity matrix are: It is used to measure the spatial coverage integrity of the robot's environmental perception information at the current time t; and calculate the data integrity ratio: Where Fint(t) is the data integrity ratio;

[0052] For noise interference in point cloud data, the noise difference is defined as: Where Tnosie(x,y,t) is the noise difference, XI(x,y,t) is the actual collected signal strength, is the ideal signal strength after filtering; calculate the overall noise interference intensity: Where Fnosie(t) is the overall noise interference intensity;

[0053] For blur measurement in image data, define blurriness: Where Mhda(x,y,t) is the blur, IT(x,y,t) is the image grayscale in two-dimensional space (x,y) and time t, is the Laplace operator output of the image at (x, y) coordinates, which is used to measure the local texture intensity. is the global maximum value of the IT Laplacian operator output of the current image; the number of pixels whose blurriness at the current time t is greater than the preset blurriness threshold is counted as the blur abnormality factor Fblur(t);

[0054] Calculate the timing anomaly factor: ΔSX(t) = max{|Fint(t) - Fint(t-1)|,|Fnosie(t) - Fnosie(t-1)|,|Fblur(t) - Fblur(t-1)|};

[0055] Construct the local coupling anomaly matrix Clocal(t): Clocal(t) = Mint(t)⊙Tnosie(t)⊙Mhda(t), where Tnoise(t) represents the noise difference matrix with noise differences Tnosie(x, y, t) as matrix elements, Mhda(t) represents the ambiguity matrix with ambiguities Mhda(x, y, t) as matrix elements, and ⊙ represents the Hadamard product of the matrix (the Hadamard product refers to the element-by-element product in the matrix, that is, the product of the elements at corresponding positions);

[0056] Calculate the local coupling anomaly coefficient: Where Flocal(t) is the local coupling anomaly coefficient, Clocal(x,y,t) is the element in the local coupling anomaly matrix Clocal(t);

[0057] Calculate the causal dynamic coefficient: Where YG(t) is the dynamic coefficient of causality, Gcau(x,y,t) is the Granger causal factor field generated based on historical multi-time data, representing the causal relationship at different spatial locations;

[0058] It should be noted that the Granger causal factor field refers to the spatial causal weight distribution field generated by Granger causality analysis based on historical perception data (such as RGB images, point cloud data, obstacle avoidance and collision event sequences, etc.). This factor field is used to reflect the causal contribution of each spatial position in the historical data to obstacle avoidance decisions and collision occurrences. It will not be elaborated here.

[0059] Calculate the perceptual quality degradation anomaly coefficient: Where θ1 is the preset timing anomaly factor threshold, and θ2 is the preset local coupling anomaly coefficient threshold;

[0060] It should be noted that the above formulas are all dimensionless and numerical calculations. Common dimensionless methods include Min-Max normalization and Z-Score normalization, which will not be described here.

[0061] The motion control distortion coefficient (MCDC) in this paper is a key indicator used to measure the degree of distortion between a robot's actual motion state and its intended motion trajectory or control instructions. It aims to reflect the comprehensive manifestation of phenomena such as trajectory deviation, attitude control deviation, and dynamic response anomalies caused by factors such as sensor anomalies, actuator response lag, path planning anomalies, or external interference in specific environments or working conditions. A large MCDC indicates a significant discrepancy between the robot's current actual motion state and the desired control instructions. This indicates that the robot's motion control chain may be experiencing significant out-of-control trends, such as severe response lag, insufficient attitude adjustment, or path deviation. This significantly reduces the effectiveness of the robot's obstacle avoidance function, preventing it from completing avoidance maneuvers in a timely manner, and significantly increases the probability of delayed, fatal collisions. Especially in complex dynamic environments, a large MCDC often indicates that the robot is unable to effectively respond to challenges posed by unexpected obstacles or environmental changes. Its obstacle avoidance response time is significantly prolonged, and the stability of the obstacle avoidance path planning and execution chain is compromised, potentially leading to irreversible collisions in high-risk areas or at high speeds. On the contrary, when the motion control distortion coefficient is small, it indicates that the robot's current motion state is basically consistent with the ideal control instructions, and its motion posture, path tracking, acceleration and deceleration response, etc. are all within the expected design range. The robot can timely complete obstacle perception, planning, and obstacle avoidance actions, and the overall motion control link remains efficient, stable, and reliable. At this time, the probability of a fatal collision caused by delayed obstacle avoidance is significantly reduced. Therefore, by constructing a motion control distortion coefficient, the present invention not only achieves quantitative monitoring of robot motion control anomalies, but also dynamically assesses the high risk of a fatal collision caused by delayed obstacle avoidance in the current state based on this coefficient, thereby driving the triggering of active risk avoidance strategies, dynamic path correction, and emergency control strategies, effectively improving the robot's operational safety and robustness in dynamic and complex environments, and fully demonstrating the application value and beneficial effects of the motion control distortion coefficient in the high-safety operation of robots.

[0062] The logic for obtaining the motion control distortion coefficient is as follows:

[0063] During the robot's motion, the following key motion data are collected in real time:

[0064] Actual motion state sequence Sact(t): includes position vector Pact(t) = {Xact(t), Yact(t), Zact(t)} (where Xact(t) represents the x-axis coordinate, Yact(t) represents the y-axis coordinate, and Zact(t) represents the z-axis coordinate), attitude angle θact(t), and velocity Vact(t);

[0065] The expected motion state sequence Sref(t): includes the expected position Pref(t) = {Xref(t), Yref(t), Zref(t)} (where Xref(t) represents the x-axis coordinate, Yref(t) represents the y-axis coordinate, and Zref(t) represents the z-axis coordinate) expected by the control system, the attitude angle θref(t), and the velocity Vref(t);

[0066] Calculate the local posture deviation, which reflects the deviation between the current posture and the expected posture: Δθz(t) = |θact(t) - θref(t)|, where Δθz(t) is the local posture deviation;

[0067] Calculate the path deviation distance: Where Dpath(t) is the path deviation distance;

[0068] Calculate the dynamic response distortion, which is used to measure the degree of distortion of the acceleration response: Where Δdyn(t) is the dynamic response distortion;

[0069] Calculate the spatiotemporal cumulative distortion coefficient within the time window [t0,t]: Where Mctrl(t) is the spatiotemporal cumulative distortion coefficient, τ is the time index, τ∈[t0,t], Dmax is the maximum allowable path deviation reference, a1, a2, and a3 represent the preset proportional coefficients of local posture deviation, path deviation distance, and dynamic response distortion, respectively, and a1, a2, and a3 are all greater than 0;

[0070] It should be noted that the above formulas are all dimensionless and calculated numerically. Common dimensionless methods include Min-Max normalization and Z-Score standardization, which are not described in detail here. a1, a2, and a3 are set according to actual conditions. For example, an expert weighting method is used, that is, experts in related fields are invited to determine the preset proportional coefficients of various indicators through professional opinion surveys and comprehensive evaluations. For example, a1, a2, and a3 can be 0.3, 0.4, and 0.3 respectively.

[0071] Calculate the motion control distortion coefficient: Yssz(t)=min(1,Mctrl(t)), where Yssz(t) is the motion control distortion coefficient;

[0072] Step S2: constructing an obstacle avoidance hysteresis fatal collision risk assessment model based on the perception quality degradation abnormality coefficient and the motion control distortion coefficient, outputting an obstacle avoidance hysteresis fatal collision risk assessment index, and evaluating the risk level of the robot in the current state of an obstacle avoidance hysteresis fatal collision;

[0073] An obstacle avoidance hysteresis fatal collision risk assessment model is constructed based on the perception quality degradation abnormality coefficient and the motion control distortion coefficient, and an obstacle avoidance hysteresis fatal collision risk assessment index is output. The obstacle avoidance hysteresis fatal collision risk assessment model is based on the following formula: bzzs(t) = w1*Cpq(t) + w2*Yssz(t), where bzzs(t) is the obstacle avoidance hysteresis fatal collision risk assessment index, Cpq(t) is the perception quality degradation abnormality coefficient, Yssz(t) is the motion control distortion coefficient, w1 and w2 are preset proportional coefficients of the perception quality degradation abnormality coefficient and the motion control distortion coefficient, respectively, and both w1 and w2 are greater than 0.

[0074] It should be noted that the above formulas are all dimensionless and calculated numerically. Common dimensionless methods include Min-Max normalization and Z-Score standardization, which are not described in detail here. w1 and w2 are set according to actual conditions. For example, the expert weighting method is adopted, that is, experts in related fields are invited to determine the preset proportional coefficients of various indicators through professional opinion surveys and comprehensive evaluations. For example, w1 and w2 can be 0.5 and 0.5 respectively.

[0075] The above calculation expression shows that the greater the perception quality degradation anomaly coefficient and the greater the motion control distortion coefficient, the greater the obstacle avoidance delayed fatal collision risk assessment index. This indicates that the robot's comprehensive obstacle avoidance capability has degraded significantly due to factors such as insufficient perception information integrity, ambiguous environmental perception, path deviation, and posture distortion, and the greater the risk of delayed fatal collision. In this case, when faced with dynamic obstacles or complex environments, the robot may be unable to accurately and promptly identify the obstacle's location and dynamic changes due to the decline in both perception and motion control capabilities, and may also have difficulty in timely correcting its path and posture. This significantly increases the probability of collision, especially in high-speed movement, complex terrain, and dense obstacle scenarios. Conversely, the smaller the perception quality degradation anomaly coefficient and the smaller the motion control distortion coefficient, the lower the obstacle avoidance delayed fatal collision risk assessment index. This indicates that the robot can more stably maintain high-quality environmental perception and precise motion control response, providing high safety and robustness in obstacle avoidance tasks, and significantly reducing the risk of fatal collisions caused by delayed response.

[0076] Step S3 compares the obstacle avoidance hysteresis fatal collision risk assessment index with a preset obstacle avoidance hysteresis fatal collision risk assessment index threshold to identify the high risk level of the robot in the current state of the obstacle avoidance hysteresis fatal collision, as follows:

[0077] If the obstacle avoidance hysteresis fatal collision risk assessment index is greater than the obstacle avoidance hysteresis fatal collision risk assessment index threshold, it means that the robot's current perception and motion control state can no longer fully guarantee the timeliness and effectiveness of its obstacle avoidance response. There is a high probability of fatal collision with obstacles due to obstacle avoidance hysteresis. The high risk flag should be triggered immediately, and the risk level of the robot's current obstacle avoidance hysteresis fatal collision should be marked as high risk.

[0078] If the obstacle avoidance hysteresis fatal collision risk assessment index is less than or equal to the obstacle avoidance hysteresis fatal collision risk assessment index threshold, it means that the robot's perception integrity and motion control response are within the safe range in the current state, and the overall obstacle avoidance capability is relatively reliable. There is no need to mark the current state as high risk, and normal operation can be maintained. There is no need to mark the risk level of the robot in the current state of obstacle avoidance hysteresis fatal collision.

[0079] It should be noted that the threshold value of the obstacle avoidance delayed fatal collision risk assessment index can be set based on the specific application needs and the safety requirements of the actual operating environment through statistical analysis of historical operating data, simulation testing, or expert experience. For example, based on the robot's historical obstacle avoidance success rate, collision event frequency, delayed response time distribution and other data in different environments, the appropriate threshold value can be determined using the quantile method, probability distribution fitting method, or optimal threshold determination method based on the ROC curve. The threshold range can also be determined through expert research and risk assessment in combination with application scenarios (such as high-speed operation scenarios and complex obstacle-intensive environments) to meet the requirements of different safety levels.

[0080] Step S4: If the robot is at a high risk of a fatal collision due to delayed obstacle avoidance, an updated objective function Jnew is constructed based on the control delay cost and the path length cost: Among them, J ρ To control the delay cost, is the path length cost, w ρ 、 They represent the preset proportional coefficients for controlling the delay cost and the path length cost, respectively, and w ρ 、 All greater than 0;

[0081] It should be noted that w ρ 、 Set it according to the actual situation. For example, adopt the expert empowerment method, that is, invite experts in related fields to determine the preset proportion coefficients of various indicators through professional opinion surveys and comprehensive evaluations. For example, w ρ 、 It can be 0.5, 0.5;

[0082] Controlling delay cost J ρ :Jρ =∫ρ(t)*s(t)dt, where ρ(t) is the control delay curve of the current system and s(t) is the path arc length speed;

[0083] Path length cost Where P(t) is the length of the robot's current motion path;

[0084] Taking the updated objective function Jnew as the optimization goal, solve the new path: Where st is the constraint condition, v(t) is the robot's motion speed, and v min is the minimum speed, v max is the maximum velocity, js(t) is the angular velocity, js max is the maximum angular velocity, WZ(0) is the starting position of the current path, WZ(t js ) is the end position of the current path;

[0085] It should be noted that the above formulas are all dimensionless and numerical calculations. Common dimensionless methods include Min-Max normalization and Z-Score normalization, which will not be described here.

[0086] The present invention provides an autonomous obstacle avoidance method for indoor decoration robots based on artificial intelligence. By integrating the perception quality degradation information and the low-power motion control distortion information, the method realizes the dynamic evaluation of the robot's obstacle avoidance hysteresis fatal collision risk and path adaptive optimization. The method can significantly improve the safety robustness and path correction ability of indoor decoration robots in high-intensity and long-cycle working environments. The specific effects are reflected in the following aspects: On the one hand, the present invention can obtain and quantify the robot's sensor perception anomaly and control distortion state in real time and calculate the perception quality degradation anomaly coefficient and motion control distortion coefficient to construct an obstacle avoidance hysteresis fatal collision risk assessment model. Based on the obstacle avoidance hysteresis fatal collision risk assessment index Accurately identify the high-risk states of fatal collisions that robots may face in complex and dynamic indoor environments, and effectively make up for the risk identification blind spots of existing obstacle avoidance strategies in scenarios with coupling defects of perception pollution and control instability; on the other hand, the present invention adaptively introduces control delay cost and path length cost in high-risk states, constructs an updated objective function and performs multi-objective optimization of the path, thereby realizing dynamic correction of the robot's original path. It can take into account path smoothness and energy consumption control while ensuring obstacle avoidance safety, avoid uncontrolled behaviors such as robot motion freezing and repeated detection caused by instability of obstacle avoidance strategies, and significantly improve the autonomous obstacle avoidance efficiency and stability of interior decoration robots in areas with complex furniture layouts and dense obstacles.

[0087] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0088] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0089] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence, characterized by: The steps include: Step S1, obtaining perception quality degradation information of a robot sensor and motion control distortion information of the robot in a low-battery state, wherein the perception quality degradation information includes a perception quality degradation abnormality coefficient, and the motion control distortion information includes a motion control distortion coefficient; Step S2: constructing an obstacle avoidance hysteresis fatal collision risk assessment model based on the perception quality degradation abnormality coefficient and the motion control distortion coefficient, outputting an obstacle avoidance hysteresis fatal collision risk assessment index, and evaluating the risk level of the robot in the current state of an obstacle avoidance hysteresis fatal collision; Step S3, comparing the obstacle avoidance hysteresis fatal collision risk assessment index with a preset obstacle avoidance hysteresis fatal collision risk assessment index threshold to identify the high risk level of the obstacle avoidance hysteresis fatal collision occurring in the current state of the robot; In step S4, if the robot is at a high risk of a fatal collision due to delayed obstacle avoidance, an updated objective function is constructed based on the control delay cost and the path length cost, and the original path is adaptively optimized with multiple objectives.

2. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 1, characterized in that: The logic for obtaining the abnormal coefficient of perceived quality degradation is as follows: Perform missing detection on the two-dimensional space of the environmental perception data collected by the robot sensor, and construct the data integrity matrix Mint(t) in the two-dimensional space (x, y) and at time t. The elements in the data integrity matrix are: And calculate the data integrity ratio: Where Fint(t) is the data integrity ratio; For noise interference in point cloud data, the noise difference is defined as: Where Tnosie(x,y,t) is the noise difference, XI(x,y,t) is the actual collected signal strength, is the ideal signal strength after filtering; Calculate the overall noise interference intensity: Where Fnosie(t) is the overall noise interference intensity; For blur measurement in image data, define blurriness: Where Mhda(x,y,t) is the blur, IT(x,y,t) is the image grayscale in two-dimensional space (x,y) and time t, is the Laplace operator output of the image at (x,y) coordinates, is the global maximum value of the IT Laplacian operator output of the current image; the number of pixels whose blurriness at the current time t is greater than the preset blurriness threshold is counted as the blur abnormality factor Fblur(t); Calculate the timing anomaly factor: ΔSX(t) =max{|Fint(t)-Fint(t-1)|,|Fnosie(t)-Fnosie(t-1)|,|Fblur(t)-Fblur)|}; (t-1 Construct the local coupling anomaly matrix Clocal(t): Clocal(t) = Mint(t)⊙Tnosie(t)⊙Mhda(t), where Tnosie(t) represents the noise difference matrix with noise differences Tnosie(x, y, t) as matrix elements, Mhda(t) represents the ambiguity matrix with ambiguities Mhda(x, y, t) as matrix elements, and ⊙ represents the Hadamard product representation of the matrix. Calculate the local coupling anomaly coefficient: Where Flocal(t) is the local coupling anomaly coefficient, Clocal(x,y,t) is the element in the local coupling anomaly matrix Clocal(t); Calculate the causal dynamic coefficient: Where YG(t) is the dynamic coefficient of causality, Gcau(x,y,t) is the Granger causal factor field generated based on historical multi-time data, representing the causal relationship at different spatial locations; Calculate the perceptual quality degradation anomaly coefficient: Wherein θ1 is the preset timing anomaly factor threshold, and θ2 is the preset local coupling anomaly coefficient threshold.

3. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 1, characterized in that: The logic for obtaining the motion control distortion coefficient is as follows: During the robot's motion, the following key motion data are collected in real time: Actual motion state sequence Sact(t): includes position vector Pact(t) = {Xact(t), Yact(t), Zact(t)}, attitude angle θact(t), velocity Vact(t); Expected motion state sequence Sref(t): includes the expected position Pref(t) = {Xref(t), Yref(t), Zref(t)}, attitude angle θref(t), and velocity Vref(t) expected by the control system; Calculate the local posture deviation: Δθz(t) = |θact(t) - θref(t)|, where Δθz(t) is the local posture deviation; Calculate the path deviation distance: Where Dpath(t) is the path deviation distance; Calculate the dynamic response distortion: Where Δdyn(t) is the dynamic response distortion; Calculate the spatiotemporal cumulative distortion coefficient within the time window [t0,t]: Where Mctrl(t) is the spatiotemporal cumulative distortion coefficient, τ is the time index, τ∈[t0,t], Dmax is the maximum allowable path deviation reference, a1, a2, and a3 represent the preset proportional coefficients of local posture deviation, path deviation distance, and dynamic response distortion, respectively, and a1, a2, and a3 are all greater than 0; Calculate the motion control distortion coefficient: Yssz(t)=min(1,Mctrl(t)), where Yssz(t) is the motion control distortion coefficient.

4. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 1, characterized in that: An obstacle avoidance hysteresis fatal collision risk assessment model is constructed based on the perception quality degradation abnormality coefficient and the motion control distortion coefficient, and an obstacle avoidance hysteresis fatal collision risk assessment index is output. The obstacle avoidance hysteresis fatal collision risk assessment model is based on the following formula: bzzs(t) = w1*Cpq(t) + w2*Yssz(t), where bzzs(t) is the obstacle avoidance hysteresis fatal collision risk assessment index, Cpq(t) is the perception quality degradation abnormality coefficient, Yssz(t) is the motion control distortion coefficient, w1 and w2 represent the preset proportional coefficients of the perception quality degradation abnormality coefficient and the motion control distortion coefficient, respectively, and both w1 and w2 are greater than 0.

5. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 4, characterized in that: The obstacle avoidance hysteresis fatal collision risk assessment index is compared with the preset obstacle avoidance hysteresis fatal collision risk assessment index threshold to identify the high risk level of the robot in the current state of obstacle avoidance hysteresis fatal collision, as follows: If the obstacle avoidance hysteresis fatal collision risk assessment index is greater than the obstacle avoidance hysteresis fatal collision risk assessment index threshold, the risk level of the robot in the current state causing obstacle avoidance hysteresis fatal collision is marked as high risk level; If the obstacle avoidance hysteresis fatal collision risk assessment index is less than or equal to the obstacle avoidance hysteresis fatal collision risk assessment index threshold, there is no need to mark the risk level of the obstacle avoidance hysteresis fatal collision of the current state robot.

6. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 5, characterized in that: If the robot is at a high risk of a fatal collision due to obstacle avoidance hysteresis, the updated objective function is constructed based on the control delay cost and path length cost. Among them, J ρ To control the delay cost, is the path length cost, w ρ 、 They represent the preset proportional coefficients for controlling the delay cost and the path length cost, respectively, and w ρ 、 Both are greater than 0.

7. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 6, characterized in that: Controlling delay cost J ρ :J ρ =∫ρ(t)*s(t)dt, where ρ(t) is the control delay curve of the current system and s(t) is the path arc length speed; Path length cost Where P(t) is the length of the robot's current motion path.

8. The autonomous obstacle avoidance method for an interior decoration robot based on artificial intelligence according to claim 6, characterized in that: Taking the updated objective function Jnew as the optimization goal, solve the new path: Where st is the constraint condition, v(t) is the robot's motion speed, and v min is the minimum speed, v max is the maximum velocity, js(t) is the angular velocity, js max is the maximum angular velocity, WZ(0) is the starting position of the current path, WZ(t js ) is the end position of the current path.

Images