Method and system for collaboration of soft robot edge small model and cloud large model

Abstract

The invention relates to the technical field of robots, and discloses a soft robot edge small model and cloud large model collaboration method and system, physical interaction information of a robot body and the inner wall of a pipeline is acquired, and an atypical interaction mode is identified according to the information, so that the robot can sense complex deformation in the pipeline in real time. And in response to the identified atypical interaction mode, the robot can dynamically adjust the local deformation and motion rhythm of the robot, so that the physical interaction state with the inner wall of the pipeline is optimized, and the robot is driven to deform and move according to the physical interaction state. In addition, the method further comprises the steps of continuously obtaining physical interaction information and conducting feedback adjustment, and it is ensured that the robot can continuously adapt to changes of the pipeline environment.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to a method and system for collaboration between a small edge model of a soft robot and a large cloud model. Background Technology

[0002] In the routine maintenance of oil and gas pipelines, traditional rigid inspection robots often struggle to adapt to the complex and ever-changing pipeline environment, frequently experiencing problems such as motion jamming and sensor failure. To address this, the industry has introduced more flexible soft robot technology, aiming to leverage its unique structural advantages to adapt to these demanding working conditions. These soft robots typically incorporate edge computing units for real-time perception and local motion control, transmitting the processed data to a cloud-based central processing system for in-depth analysis and global decision-making.

[0003] However, in practical applications, especially in extreme situations such as complex internal pipe structures, degraded sensory data quality, multiple risks to the robot itself, and limited communication with the cloud, this collaborative working mode faces severe challenges. For example, when a soft robot travels deep into a pipe and encounters a section with a small but locally highly complex deformable internal structure, its edge computing unit's built-in motion planning and local obstacle avoidance logic struggle to extract the optimal motion strategy. The robot's motion efficiency decreases significantly, and in some local areas, the contact pressure distribution between the robot's shell and the pipe wall becomes exceptionally complex and uneven.

[0004] Meanwhile, the edge units transmit this pre-processed geometric data and their own motion state information about the complex deformation region to the remote central cloud processing system. However, due to the subtle and complex nature of these deformations, the edge units inevitably lose some crucial details during data compression and feature extraction. When the cloud system receives this "simplified" data, its powerful 3D reconstruction and structural analysis logic, while able to identify anomalies in the region, struggles to accurately reconstruct the true shape and depth of these complex deformations. The cloud system attempts to compare this ambiguous anomaly information with past data, but due to a lack of sufficiently refined features, its assessment of the structural integrity of the region is highly uncertain, and may even result in misjudgments.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This invention provides a method and system for collaboration between a small edge model of a soft robot and a large cloud model. It aims to solve the problems in the prior art where soft robots are unable to effectively adapt to complex deformation areas under extreme conditions such as complex internal pipe structures, degraded perception data quality, multiple risks to the robot body, and limited communication with the cloud, resulting in decreased motion efficiency, abnormal contact pressure, and the inability of the cloud system to accurately reproduce the true shape and depth of deformation.

[0007] The technical solution of this application is as follows:

[0008] In a first aspect, this application discloses a method for collaboration between a small edge model of a soft robot and a large cloud model, including:

[0009] Obtain information about the physical interaction between the robot body and the inner wall of the pipe;

[0010] Based on physical interaction information, identify atypical interaction patterns between the robot body and the inner wall of the pipe;

[0011] In response to atypical interaction modes, the robot body's local deformation and motion rhythm are adjusted to optimize the physical interaction between the robot body and the inner wall of the pipe, and the robot body is driven to deform and move according to the adjusted local deformation and motion rhythm.

[0012] It continuously acquires physical interaction information and adjusts local deformation and motion rhythm based on the continuously acquired physical interaction information.

[0013] This technical solution enables soft robots to perceive and identify atypical interaction patterns in complex pipe deformation areas in real time, and dynamically adjust their own deformation and movement accordingly, thereby achieving adaptive passage through complex pipe environments. This effectively solves the problems of poor adaptability and low motion efficiency of traditional methods in complex pipe environments.

[0014] Furthermore, based on the above method, in response to atypical interaction modes, the local deformation and motion rhythm of the robot body are adjusted to optimize the physical interaction state between the robot body and the inner wall of the pipe, and the robot body is driven to deform and move according to the adjusted local deformation and motion rhythm, including:

[0015] Continuously acquire the robot's operating status information, including battery level, internal temperature, and the health status of the tactile sensors;

[0016] Continuously acquire potential risk information in the current environment, including flammable gas concentration and communication signal strength with remote processing systems;

[0017] Based on the currently identified atypical interaction patterns and alternative exercise strategies, predict the rate of energy expenditure and the rate of heat accumulation that each alternative exercise strategy may result in;

[0018] By combining operational status information, potential environmental risk information, and alternative motion strategies, assess the remaining safe operating time and the likelihood of mission completion for the alternative motion strategies.

[0019] Based on the assessment results and the preset security protocols, adjust the task priorities and select action strategies.

[0020] This technical solution enables robots to not only adjust based on physical interaction information, but also to comprehensively consider their own operating status, environmental risks, and the energy consumption and heat accumulation of different strategies, thereby making safer and more efficient decisions and significantly improving the robot's autonomous decision-making ability and safety in complex environments.

[0021] More specifically, in some implementation schemes, task priorities are adjusted and action strategies are selected based on assessment results and preset security protocols, including:

[0022] If the risk of flammable gas is high and the internal temperature has reached the high temperature warning level, then the minimum power consumption safe shutdown strategy will be implemented.

[0023] If the risk of flammable gas is medium and the internal temperature continues to rise, then a partial retreat strategy should be implemented.

[0024] If the immediate security threat is zero and the probability of mission completion is low, then choose a downgraded escape strategy that consumes less energy and causes less wear and tear on the robot itself.

[0025] If the probability of completing the task is high and the overall risk is low, then the task continuation strategy should be selected.

[0026] Through this technical solution, robots can flexibly choose the most appropriate action strategy based on different risk levels and the probability of task completion. For example, in extremely dangerous situations, they can prioritize ensuring safe shutdown, and when the risk is controllable, they can choose to downgrade and escape or continue the task, thereby effectively avoiding potential dangers and optimizing task execution efficiency.

[0027] Preferably, by combining operational status information, potential environmental risk information, and alternative motion strategies, the remaining safe operating time and task completion probability of the alternative motion strategies are assessed, including:

[0028] Continuously acquire the instantaneous discharge current and voltage of the battery;

[0029] Calculate the instantaneous output power of the battery based on the instantaneous discharge current and voltage;

[0030] The actual usable capacity of the battery is adjusted based on the instantaneous output power, battery temperature, and number of cycles.

[0031] Estimate the battery's true remaining capacity percentage based on the corrected actual usable capacity and instantaneous discharge current;

[0032] Calculate the remaining safe operating time based on the actual remaining percentage of battery power and the predicted energy consumption rate.

[0033] This technical solution enables robots to more accurately estimate the actual remaining battery power and safe operating time, providing more reliable energy information for decision-making and avoiding unexpected downtime or task interruption due to inaccurate power estimation.

[0034] Building upon the above, this application further proposes continuously acquiring potential risk information in the current environment, including flammable gas concentration and communication signal strength with the remote processing system, including:

[0035] Multiple wireless signal relay units are deployed, and the wireless signal relay units receive communication signals from the robot body;

[0036] When a decline in communication quality with the remote processing system is detected, the robot's posture and the transmission power and direction of the wireless signal relay unit are adjusted to optimize the strength of the communication signal and the data transmission delay.

[0037] Based on the urgency of the current task and the importance of data transmission, the data that needs to be sent to the remote processing system is prioritized, encoded, and fragmented.

[0038] When communication quality is extremely poor or completely interrupted, activate the local emergency data processing and decision-making mode and suspend the sending of non-essential data to the remote processing system.

[0039] This technical solution enables robots to proactively optimize communication quality and perform local emergency handling when communication is restricted, ensuring the transmission of critical information and the continuous execution of tasks, effectively solving the problem of communication interruption or quality degradation in complex pipeline environments.

[0040] As a technological improvement, based on currently identified atypical interaction patterns and alternative exercise strategies, the potential energy expenditure rate and calorie accumulation rate for each alternative exercise strategy are predicted, including:

[0041] Based on the dynamic changes in physical interaction information, identify the rheological behavior of the medium inside the pipe or the dynamic response characteristics of obstacles;

[0042] Based on the identified rheological behavior or dynamic response characteristics, the prediction parameters for energy consumption rate and heat accumulation rate are corrected;

[0043] Based on the revised prediction parameters, estimate the energy consumption rate and heat accumulation rate of each alternative exercise strategy in the current dynamic environment.

[0044] This technical solution enables robots to more accurately predict the energy consumption and heat accumulation of different motion strategies, especially in dynamically changing pipeline environments, thereby selecting more energy-efficient and safer motion strategies.

[0045] To improve the plan, based on operational status information, potential environmental risk information, and alternative movement strategies, the remaining safe operating time and task completion probability of the alternative movement strategies are assessed, including:

[0046] Based on physical interaction information, identify local features of the inner wall of the pipe;

[0047] Based on the characteristics of the local area, the weights of the robot's operating status information, potential environmental risk information, and strategy prediction results are adjusted, and the remaining safe operating time and task completion probability of each alternative motion strategy are evaluated.

[0048] This technical solution enables robots to dynamically adjust the weights of various evaluation parameters based on the specific characteristics of local areas of the pipeline, making the evaluation results more consistent with the actual situation and improving the accuracy and adaptability of decision-making.

[0049] Based on this, and according to the characteristics of the local area, the weights of the robot's operational status information, potential environmental risk information, and strategy prediction results are adjusted, including:

[0050] The identified rheological behaviors or dynamic response features are classified to identify the corresponding rheological modes or dynamic response modes;

[0051] Based on the classified rheological modes or dynamic response modes, analyze the interactions between different modes and determine the coupling effect;

[0052] Based on the classified rheological modes or dynamic response modes and coupling effects, the weights of the robot's operating state information, potential environmental risk information, and strategy prediction results are dynamically adjusted.

[0053] Through this technical solution, robots can deeply analyze the complex modes of the pipeline environment and their coupling effects, thereby adjusting the evaluation weights more precisely and further improving the intelligence and accuracy of decision-making.

[0054] In one implementation, the weights of the robot's operational state information, potential environmental risk information, and strategy prediction results are adjusted based on local region characteristics, including:

[0055] Based on the continuously acquired physical interaction information, the instantaneous rate of change and trend of change of the physical interaction information are analyzed in real time;

[0056] Based on the instantaneous rate of change and the trend of change, identify the rapid switching points or superposition regions of rheological modes or dynamic response modes;

[0057] In response to the rapid switching or superposition of rheological modes or dynamic response modes, the weights of the robot's operating status information, potential environmental risk information, and strategy prediction results are dynamically adjusted according to the type and intensity of the switching or superposition.

[0058] If multiple modalities are identified as superimposed, the weights are adjusted according to the superposition effect of the multiple modalities.

[0059] Through this technical solution, the robot can capture the dynamic changes in the pipeline environment in real time, especially the rapid switching and superposition of modes, and make fine-grained adjustments to the weights accordingly, making the decision-making system respond to environmental changes more quickly and accurately.

[0060] Secondly, this application also discloses a system for collaboration between a small edge model of a soft robot and a large cloud model, comprising:

[0061] The input end is used to acquire physical interaction information between the robot body and the inner wall of the pipe; based on the physical interaction information, it identifies atypical interaction patterns between the robot body and the inner wall of the pipe.

[0062] The adjustment end is used to respond to atypical interaction modes, adjust the local deformation and motion rhythm of the robot body to optimize the physical interaction state between the robot body and the inner wall of the pipe, and drive the robot body to deform and move according to the adjusted local deformation and motion rhythm.

[0063] The feedback end is used to continuously acquire physical interaction information and adjust local deformation and motion rhythm based on the continuously acquired physical interaction information.

[0064] This system enables automated control of soft robots to adapt and navigate through complex pipeline environments. Its modular design improves maintainability and scalability, providing reliable hardware and software support for the practical application of soft robots.

[0065] Beneficial effects

[0066] This application discloses a method and system for the collaboration between a small edge model and a large cloud model of a soft robot. By acquiring physical interaction information between the robot body and the inner wall of a pipe, and identifying atypical interaction patterns based on this information, the robot can perceive complex deformations inside the pipe in real time. In response to the identified atypical interaction patterns, the robot can dynamically adjust its local deformation and movement rhythm, thereby optimizing its physical interaction with the inner wall of the pipe and driving the robot to deform and move accordingly. Furthermore, the method includes continuously acquiring physical interaction information and making feedback adjustments to ensure that the robot can continuously adapt to changes in the pipe environment.

[0067] Through the above technical solution, this application effectively solves the problems of existing technologies, such as the difficulty of soft robots adapting to complex and ever-changing pipeline environments, motion stagnation, sensor failure, and the inability of cloud systems to accurately reproduce complex deformations. Specifically, this application overcomes the motion obstacles encountered by traditional rigid inspection robots in complex pipelines through adaptive adjustment of the robot body, avoiding decreased motion efficiency and abnormally complex and uneven local contact pressure. Simultaneously, because the robot can make fine adjustments based on local perception and decision-making, it reduces reliance on remote cloud processing systems, avoiding misjudgments by the cloud system due to the loss of key details during data compression and feature extraction. Therefore, the technical solution of this application can significantly improve the adaptability, motion efficiency, and safety of soft robots in complex pipeline environments, achieving effective passage through complex deformation areas of pipelines, and has significant practical value and technological advancement. Attached Figure Description

[0068] Figure 1 This is a flowchart of a method for collaboration between a small edge model and a large cloud model of a soft robot, provided by an embodiment of the present invention.

[0069] Figure 2 This is a flowchart of a method for responding to atypical interaction modes provided by an embodiment of the present invention;

[0070] Figure 3 This is a flowchart of a method for adjusting task priority and selecting action strategy provided by an embodiment of the present invention;

[0071] Figure 4 This is a schematic diagram of the structure of a system for collaboration between a small edge model of a soft robot and a large cloud model, provided in an embodiment of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Reference Figure 1 , Figure 1 This is a flowchart illustrating a method for collaboration between a small edge model and a large cloud model of a soft robot, as provided in an embodiment of the present invention, including:

[0074] S11, Obtain physical interaction information between the robot body and the inner wall of the pipe;

[0075] S12, Based on the physical interaction information, identify atypical interaction patterns between the robot body and the inner wall of the pipe;

[0076] S13, in response to the atypical interaction mode, adjust the local deformation and movement rhythm of the robot body to optimize the physical interaction state between the robot body and the inner wall of the pipe, and drive the robot body to deform and move according to the adjusted local deformation and movement rhythm.

[0077] S14, continuously acquire the physical interaction information and adjust the local deformation and motion rhythm based on the continuously acquired physical interaction information.

[0078] This application introduces a mechanism for recognizing and responding to atypical interaction patterns, enabling soft robots to dynamically adjust their local deformation and movement rhythm when faced with complex pipe deformation. This optimizes the physical interaction with the inner wall of the pipe, significantly improving the robot's adaptability and throughput efficiency, and effectively solving the problems of poor adaptability and easy jamming in complex environments of traditional methods.

[0079] To better understand the technical solutions proposed in this application, it is necessary to explain some key terms involved. A soft robot refers to a robot made of flexible materials that can adapt to complex environments by changing its shape and stiffness. Physical interaction information refers to data such as force, pressure distribution, and friction generated when the robot body comes into contact with the inner wall of the pipe; this information can be acquired through tactile sensors, pressure sensors, etc. Atypical interaction modes refer to abnormal or unexpected physical interaction states between the robot body and the inner wall of the pipe, such as local high pressure, abnormal friction, or jamming; these modes usually indicate that the robot may face difficulties or reduced efficiency. Local deformation refers to shape changes that occur in a specific area of ​​the robot body, such as local expansion, contraction, or bending. Motion rhythm refers to parameters such as speed, frequency, and gait of the robot body during movement.

[0080] In practical implementation, acquiring physical interaction information between the robot body and the inner wall of the pipe is fundamental to the entire adaptive process. For example, this can be achieved by integrating a high-density array of tactile sensors onto the robot's surface. These sensors can monitor pressure, shear force, and other information at the contact points between the robot and the pipe wall in real time. Another approach is to use visual sensors combined with image processing technology to indirectly infer physical interaction information by analyzing the deformation of the robot's surface texture or marked points. Furthermore, acoustic sensors can be used to monitor the sound wave characteristics generated by friction between the robot and the pipe wall, thereby acquiring physical interaction information.

[0081] Identifying atypical interaction patterns between the robot and the pipe wall based on the acquired physical interaction information is a crucial step. This can be achieved, for example, through preset thresholds or pattern matching algorithms. When sensor data (such as local pressure or friction) exceeds the normal range or does not match preset typical interaction patterns, it can be identified as an atypical interaction pattern. Specifically, a machine learning-based model can be established, trained on a large amount of normal and abnormal interaction data, enabling the model to automatically identify various atypical interaction patterns. For example, when the local pressure on the robot in a certain area of ​​the pipe is consistently higher than the average, accompanied by abnormal vibration frequencies, the system can identify this as an atypical interaction pattern of "local jamming."

[0082] In response to atypical interaction patterns, the system adjusts the local deformation and movement rhythm of the robot body to optimize the physical interaction between the robot body and the inner wall of the pipe, and drives the robot body to deform and move according to the adjusted local deformation and movement rhythm. For example, when a "local jamming" pattern is detected, the robot body can locally contract or expand to reduce the contact area with the inner wall of the pipe or change the pressure distribution at the contact point, thereby reducing friction or releasing the jam. Simultaneously, the movement rhythm can also be adjusted accordingly, such as slowing down, changing the gait frequency, or performing small-amplitude reciprocating movements to help the robot body escape the predicament. Specifically, precise control of local deformation can be achieved by controlling the fluid pressure inside the robot body or the drive motor. For example, after identifying an atypical interaction pattern, the control system can select one or more combinations of local deformation and movement rhythms according to a preset strategy library, and drive the robot body to perform corresponding deformation and movement through actuators.

[0083] By continuously acquiring physical interaction information and adjusting local deformation and motion rhythm based on this information, the system ensures the robot's continuous adaptive capability. For example, after adjusting local deformation and motion rhythm, the system continues to monitor physical interaction information. If new physical interaction information indicates that the interaction state is still not optimized, or if new atypical interaction patterns emerge, the system will identify and adjust again. This feedback mechanism enables the robot to respond to environmental changes in real time and continuously optimize its motion strategy. For example, after adjustment, if local pressure is still too high, the system may further adjust local deformation or try different motion rhythms until the physical interaction state reaches its optimum.

[0084] The proposed method for coordinating a small edge model and a large cloud model of a soft robot forms a closed-loop adaptive control system by introducing real-time perception of the physical interaction information between the robot body and the inner wall of the pipe, intelligent identification of atypical interaction patterns, and dynamic adjustment and feedback mechanisms for local deformation and motion rhythm. When the soft robot moves through the pipe, its sensors continuously acquire physical interaction information with the pipe's inner wall. This information is analyzed in real time to identify any atypical interaction patterns, such as local high pressure, abnormal friction, or potential jamming. Once an atypical interaction pattern is identified, the system responds immediately, adjusting the robot body's local deformation (e.g., local expansion or contraction) and motion rhythm (e.g., speed, gait frequency) according to the current situation to optimize the physical interaction state between the robot body and the pipe's inner wall, thereby avoiding jamming or damage. Subsequently, the robot body deforms and moves according to the adjusted parameters. During this process, the system continuously acquires new physical interaction information and adjusts the local deformation and motion rhythm based on this information, ensuring that the robot body can continuously and efficiently adapt to the complex deformation area of ​​the pipe. This method enables soft robots to autonomously cope with minute but complex structural changes inside pipes, significantly improving their ability to navigate in harsh environments and their efficiency in completing tasks.

[0085] Compared to traditional soft robot inspection methods, this application offers significant advantages. Traditional soft robots typically rely on pre-set motion plans or simple local obstacle avoidance logic. In extreme situations such as complex internal pipe structures, degraded perception data quality, multiple risks to the robot itself, and limited communication with the cloud, their adaptive capabilities are severely limited, easily leading to decreased motion efficiency, jamming, or even damage. For example, when a robot encounters a small but locally highly complex deformed pipe segment, traditional methods struggle to extract the optimal motion strategy, resulting in a significant decrease in robot motion efficiency. Furthermore, in some local areas, the contact pressure distribution between the robot's shell and the pipe wall becomes exceptionally complex and uneven. This application introduces a mechanism for recognizing and responding to atypical interaction patterns, enabling the soft robot to dynamically adjust its local deformation and motion rhythm when facing complex pipe deformations, thereby optimizing its physical interaction with the pipe's inner wall. This real-time, adaptive adjustment capability allows the robot to traverse complex deformation areas more flexibly and efficiently, significantly improving its adaptive capabilities and throughput efficiency. It effectively solves the problems of poor adaptability and jamming in complex environments associated with traditional methods, thereby improving the success rate and safety of inspection tasks.

[0086] In some of the embodiments described above in this application, a scheme for adjusting the local deformation and motion rhythm of the robot body in response to atypical interaction modes was proposed. However, in practical applications, if adjustments are made solely based on physical interaction information, it may not fully consider the internal operating state of the robot body, the potential risks of the external environment, and the impact of different adjustment strategies on task continuity and safety, potentially leading to biased or suboptimal decision-making. Therefore, this application further proposes a more refined adjustment strategy that optimizes the local deformation and motion rhythm of the robot body by comprehensively evaluating multi-dimensional information.

[0087] refer to Figure 2 , Figure 2 This is a flowchart of a method for responding to atypical interaction modes provided by an embodiment of the present invention, including:

[0088] S131, continuously acquire the operating status information of the robot body, including battery power, internal temperature and the health status of the tactile sensors;

[0089] S132, continuously acquire potential risk information in the current environment, including flammable gas concentration and communication signal strength with the remote processing system;

[0090] S133, based on the currently identified atypical interaction patterns and alternative exercise strategies, predict the energy consumption rate and heat accumulation rate that each alternative exercise strategy may result in;

[0091] S134, Combining the operating status information, the potential environmental risk information, and the alternative motion strategies, assess the remaining safe operating time and the likelihood of task completion for the alternative motion strategies;

[0092] S135, based on the assessment results and preset security protocols, adjusts the task priority and selects the action strategy.

[0093] Specifically, continuously acquiring operational status information of the robot means that various key physiological indicators of the robot are monitored in real time during task execution. Battery power, which can be understood as the robot's energy reserves, is monitored in real time via a built-in power sensor; internal temperature refers to the temperature of the robot's core internal components, which is detected by a temperature sensor to prevent overheating damage; and the health of the tactile sensors reflects the reliability of the robot's perception of external physical interactions, which is evaluated through self-testing procedures or comparison with historical data. Acquiring this information aims to provide a comprehensive basis for subsequent decision-making regarding the robot's internal status.

[0094] Continuously acquiring potential risk information in the current environment refers to the real-time detection of external factors in the robot's environment that may threaten its safe operation or task execution. The concentration of flammable gases is monitored in real time using gas sensors integrated into the robot to mitigate the risk of explosion; the strength of the communication signal with the remote processing system is monitored through a wireless communication module to ensure that the robot can receive instructions and transmit data in a timely manner. This acquisition of information aims to provide the robot with necessary early warnings about the external environment.

[0095] In practical applications, predicting the potential energy consumption rate and heat accumulation rate for each identified atypical interaction pattern and alternative motion strategies means that after identifying an atypical interaction pattern, the system generates a series of targeted alternative motion strategies. For each alternative strategy, using a pre-set physical model or machine learning model, combined with the robot's structural parameters, material properties, and the characteristics of the pipeline environment, the system estimates the potential energy consumption rate and internal heat accumulation rate during execution. The purpose is to quantify the resource consumption of different strategies and provide data support for subsequent evaluation.

[0096] Furthermore, by combining the operational status information, the potential environmental risk information, and the alternative motion strategies, assessing the remaining safe operating time and task completion probability of the alternative motion strategies involves comprehensively considering the robot's operational status information (such as remaining battery power and upper temperature limit), potential environmental risk information (such as flammable gas threshold and communication interruption risk), and the predicted energy consumption and heat accumulation rate. The remaining safe operating time refers to the maximum time the robot can operate safely under the current or predicted risk level; the task completion probability refers to the probability that the robot will successfully complete the current task after adopting a specific alternative strategy. This assessment aims to provide a comprehensive risk and benefit analysis for decision-making.

[0097] Specifically, adjusting task priorities and selecting action strategies based on assessment results and preset safety protocols means that the preset safety protocols are a series of action guidelines that define the robot body should take under different risk levels. For example, when the assessment results show a high safety risk, even if the probability of task completion is high, risk avoidance strategies, such as pausing, retreating, or seeking external support, may be prioritized according to the safety protocols. The adjustment of task priorities and the selection of action strategies aim to ensure that the robot body can perform tasks in the safest and most efficient way in complex environments.

[0098] This application's solution, by continuously acquiring information on the robot's operational status and potential environmental risks, enables the robot to go beyond simple physical interaction feedback when responding to atypical interaction modes, gaining a comprehensive understanding of its own health and the safety of the external environment. Based on this, the resource cost of each strategy is quantified by predicting the energy consumption and heat accumulation of different alternative motion strategies. It is precisely this comprehensive consideration of multi-dimensional information that allows the robot to conduct a more in-depth assessment of the remaining safe operating time and task completion probability of alternative strategies. Finally, combined with preset safety protocols, the robot can dynamically adjust task priorities and select the optimal action strategy, thereby avoiding suboptimal decisions or safety hazards caused by insufficient information, ensuring safety, efficiency, and task continuity when adaptively navigating complex pipeline environments.

[0099] Through the aforementioned technical solution, this application enables soft robots to adjust themselves when facing atypical interaction patterns in complex and deformable pipeline regions, no longer relying solely on immediate physical feedback. Instead, they can comprehensively consider their own operational status, potential environmental risks, and the long-term impact of different strategies. This significantly enhances the intelligence and robustness of the robot's decision-making, effectively preventing task failures or equipment losses due to energy depletion, overheating, or environmental hazards. Furthermore, by dynamically adjusting task priorities and optimizing action strategies, this application ensures that the robot can operate continuously with higher safety and a greater probability of task completion in complex and ever-changing environments, thereby greatly enhancing the practicality and reliability of soft robots in pipeline inspection, maintenance, and other fields.

[0100] In some preferred embodiments, suppose a soft robot, while inspecting a pipe, enters a locally narrow area with a highly viscous fluid adhering to its inner wall, resulting in a continuous, high-friction, atypical interaction pattern between the robot and the pipe wall. If adjustments are made solely based on this basic physical interaction information, the robot might attempt to increase deformation and movement frequency to force its way through, potentially leading to a sharp increase in energy consumption and a rapid rise in internal temperature. However, according to the solution of this application, the robot continuously acquires its operational status information, such as detecting that the battery level is at a low to medium level, the internal temperature is approaching a warning threshold, and that gas sensors in the environment detect an increase in flammable gas concentration and a decrease in the communication signal strength with the remote processing system. In this situation, the system predicts that a forced passage strategy will result in an extremely high rate of energy consumption and heat accumulation, potentially triggering a safety shutdown or explosion risk. Through comprehensive evaluation, the system determines that the remaining safe operating time for the forced passage strategy is extremely short, the probability of task completion is low, and the risk is extremely high. According to preset safety protocols, such as "when the risk of flammable gas is high and the internal temperature has reached the high-temperature warning level, a minimum power consumption safe shutdown strategy will be implemented," the robot will adjust its task priorities, prioritizing either a partial retreat strategy or a safe shutdown strategy, rather than continuing to force its way through. This allows the robot to avoid potential equipment damage or safety accidents, and, when conditions permit, to reassess or await external support, thus ensuring the overall safety and sustainability of the mission.

[0101] In some embodiments described above, this application proposes a scheme to adjust task priorities and select action strategies based on assessment results and preset safety protocols. However, in practical applications, the complexity of pipeline environments and the diversity of potential risks mean that simple strategy selection may not be sufficient to cope with emergencies or optimize resource utilization. For example, when a robot faces a high risk of flammable gases or excessively high internal temperatures, failure to take timely and targeted safety measures may lead to equipment damage or even mission failure. Without addressing these issues, the robustness and safety of the robot's autonomous decision-making in complex pipeline environments will be limited. Therefore, this application further proposes a specific implementation method for adjusting task priorities and selecting action strategies based on assessment results and preset safety protocols. This aims to improve the robot's adaptive decision-making ability under different risk levels and mission completion probabilities through preset conditional judgments and corresponding action strategies.

[0102] refer to Figure 3 , Figure 3 This is a flowchart of a method for adjusting task priority and selecting action strategies according to an embodiment of the present invention, including:

[0103] S1351, if the risk of flammable gas is high and the internal temperature has reached the high temperature warning level, then execute the minimum power consumption safe shutdown strategy.

[0104] S1352, if the risk of flammable gas is medium and the internal temperature continues to rise, then a partial retreat strategy is executed;

[0105] S1353, if the immediate security threat is 0 and the probability of mission completion is low, then select a downgraded escape strategy that consumes less energy and causes less wear to the robot body;

[0106] S1354. If the probability of task completion is high and the overall risk is low, then choose the task continuation strategy.

[0107] Specifically, "high risk of flammable gas" means that the concentration of flammable gas detected by the gas sensor exceeds a preset danger threshold, indicating an immediate high risk of explosion or combustion. Simultaneously, "internal temperature has reached a high temperature warning" means that the temperature sensor inside the robot has detected that the temperature has reached or exceeded a critical value that could damage internal electronic components or actuators. Under these dual high-risk conditions, the "minimum power consumption safe shutdown strategy" aims to minimize energy consumption, avoid exacerbating risks due to continuous operation, and quickly bring the robot to a safe state to await external rescue or risk resolution. This strategy typically includes shutting down non-essential sensors and actuators, maintaining only core communication and safety monitoring functions.

[0108] The phrase "medium risk of flammable gas" can be understood as the concentration of flammable gas at a moderate risk level, not yet reaching an immediate danger, but posing a potential threat. Meanwhile, "the internal temperature continues to rise" indicates a significant trend of heat accumulation inside the robot, which, without intervention, could quickly reach a high-temperature warning level. In this situation, implementing a "partial retreat strategy" aims to allow the robot to maintain a certain level of mobility while attempting to move from the current high-risk area to a relatively safer area, thus avoiding further escalation of the risk. For example, the robot can adjust its deformation mode, utilizing the support of the pipe's inner wall, to slowly move backward or laterally while monitoring environmental changes.

[0109] In practical applications, "zero immediate safety threat" means that there are no factors in the current environment that directly endanger the robot's safety, such as flammable gases, high temperatures, or structural damage. However, "low probability of task completion" indicates that, based on the current operating status, energy reserves, and environmental conditions, the probability of the robot completing the predetermined task is low. In this case, choosing a "low-energy-consumption, minimal-wearing-to-the-robot degraded escape strategy" aims to prioritize the protection of the robot, avoid unnecessary energy consumption and mechanical wear, and attempt to escape the predicament at the lowest cost. This strategy might include adopting a gentler creeping pattern, reducing deformation, or finding a spacious area in the pipe to rest.

[0110] Furthermore, "high probability of task completion" means that the robot has sufficient energy, good operating condition, and suitable environmental conditions, enabling it to successfully complete the predetermined task. Meanwhile, "low overall risk" indicates that all potential risk factors are controllable or at a low level. Under these ideal conditions, selecting the "task continuation strategy" means that the robot will continue to execute the task according to the original plan or the optimized path to ensure its successful completion.

[0111] This application's solution effectively addresses the limitations of the aforementioned basic solutions in strategy selection within complex and ever-changing environments by establishing a decision-making logic based on multi-dimensional risk assessment and task completion probability. Specifically, when the robot faces a high risk of flammable gas and an internal high-temperature warning, the system can quickly identify the highest priority safety threat and immediately trigger a minimum power consumption safety shutdown strategy. The principle behind this strategy is to minimize energy input and heat generation by cutting off unnecessary functions, thereby avoiding potential explosions or overheating damage and ensuring the safety of the robot and the environment. When the risk level is slightly lower but a potential threat still exists (such as a moderate risk of flammable gas and a continuously rising internal temperature), the system chooses a partial retreat strategy. The principle behind this strategy is to actively change the robot's position, moving it away from the risk accumulation area, thus buying time for further risk assessment and response, and preventing risk escalation. When the immediate safety threat is eliminated but the task completion prospect is poor, the system prioritizes the robot's long-term health and resource optimization, selecting a low-energy-consumption, low-wear degraded escape strategy. The principle behind this strategy is to extend the robot's lifespan by reducing operational intensity and optimizing movement patterns, while conserving energy for subsequent tasks or maintenance. Finally, when the probability of task completion is high and the overall risk is low, the system chooses a task continuation strategy. This strategy aims to ensure the smooth progress of the task and maximize its benefits. This hierarchical, contextualized decision-making mechanism allows the robot to dynamically adjust its behavior based on the real-time environment and its own state, thereby maximizing the success rate and efficiency of the task while ensuring safety.

[0112] Through the aforementioned technical solution, the soft robot can make more refined and intelligent decisions based on the real-time risk status and task progress within the complex deformation area of ​​the pipeline. This solution significantly improves the robot's survivability and task robustness in extreme environments, avoiding potential dangers and resource waste caused by single or rigid strategies. Specifically, by introducing multiple safety thresholds and corresponding emergency strategies, the robot can effectively avoid serious safety accidents such as flammable gas explosions and overheating damage, ensuring the safety of equipment and operators. Simultaneously, in non-urgent but task-impeded situations, the introduction of a degradation escape strategy allows the robot to seek the optimal escape path while protecting itself, extending equipment lifespan and reducing operating costs. This adaptive strategy selection mechanism enables the robot not only to "pass through" complex areas but also to "pass through" complex areas "safely and efficiently," thereby greatly enhancing the practical value and reliability of soft robots in high-risk application scenarios such as industrial inspection and rescue.

[0113] In some preferred embodiments, a specific example is given below. Suppose a soft robot, while performing a pipeline inspection task, suddenly enters an unknown area. At this moment, its onboard sensors detect a rapid increase in the concentration of flammable gas, reaching a preset high-risk threshold. Simultaneously, the robot's internal temperature sensor issues a high-temperature warning, indicating that the internal temperature is approaching a dangerous level. According to the above scheme, the system will immediately determine that the current situation meets the conditions of "high risk of flammable gas and high temperature warning," and quickly execute a minimum power consumption safety shutdown strategy. The robot will immediately stop all unnecessary movement and data acquisition, maintaining only minimal communication and safety monitoring functions to minimize energy consumption and heat generation, avoid potential explosion risks, and await external instructions or risk resolution.

[0114] For example, in another scenario, a robot is traversing a narrow and winding pipe. Its tactile sensors indicate that the friction between the robot and the pipe's inner wall is continuously increasing, and the internal temperature is also showing a continuous upward trend, though not yet reaching a high-temperature warning level. Simultaneously, gas sensors detect a moderate risk level of flammable gas concentration. At this point, the system determines that the current situation meets the criteria of "moderate flammable gas risk and continuously rising internal temperature" and selects to execute a partial retreat strategy. The robot will adjust its deformation pattern, for example, by contracting some body segments and using the peristalsis of other segments, slowly moving backward a short distance to attempt to escape the area of ​​excessive friction or rapid heat accumulation, while continuously monitoring environmental changes to buy time and space for further decision-making.

[0115] For example, when a robot encounters a difficult obstacle in a pipe, after several attempts, the system assesses that the probability of completing the current task is low, but environmental sensors show an immediate safety threat of 0, meaning there are no direct dangers such as flammable gases or high temperatures. In this situation, the system will choose a degraded escape strategy that consumes less energy and causes less wear and tear on the robot. The robot might try to use a gentler, slower crawling motion, or change the frequency and amplitude of its deformation waves, to try to bypass the obstacle or find other escape paths with minimal energy consumption and mechanical wear, rather than blindly making high-intensity, high-energy attempts.

[0116] In some embodiments described above, relying solely on coarse battery level information when assessing the remaining safe operating time and task completion probability of alternative motion strategies may lead to inaccurate judgments about the robot's actual endurance. This inaccuracy may stem from battery performance degradation under different discharge conditions, temperatures, and cycle counts, thus affecting the accurate prediction of task completion probability and safe operating time. Failure to address these issues could result in the robot shutting down due to battery depletion at critical moments, or adopting inappropriate strategies due to misjudgment of safe operating time, thereby increasing task risk or causing task failure. To address this, this application proposes a more refined assessment method that uses real-time monitoring and correction of the battery's instantaneous state to more accurately calculate the remaining safe operating time.

[0117] The above-mentioned assessment of the remaining safe operating time and task completion probability of alternative motion strategies, combining the operational status information, the potential environmental risk information, and the alternative motion strategies, includes:

[0118] Continuously acquire the instantaneous discharge current and voltage of the battery;

[0119] Calculate the instantaneous output power of the battery based on the instantaneous discharge current and voltage;

[0120] The actual usable capacity of the battery is adjusted based on the instantaneous output power, battery temperature, and number of cycles.

[0121] Estimate the actual remaining battery capacity percentage based on the corrected actual usable capacity and the instantaneous discharge current;

[0122] The remaining safe operating time is calculated based on the actual remaining battery percentage and the predicted energy consumption rate.

[0123] Specifically, continuously acquiring the battery's instantaneous discharge current and voltage refers to monitoring the battery's instantaneous current and voltage output values ​​in real time during operation using current and voltage sensors integrated within the robot body. This data forms the basis for evaluating the battery's real-time operating status, aiming to provide accurate raw input for subsequent power calculations and capacity adjustments.

[0124] Calculating the instantaneous output power of the battery based on the instantaneous discharge current and voltage can be understood as multiplying the real-time acquired instantaneous discharge current and voltage using Ohm's law or a power calculation formula to obtain the battery's instantaneous output power at the current moment. The instantaneous output power reflects the immediate level of energy consumption of the robot body, and its purpose is to provide key dynamic parameters for subsequent battery capacity adjustments.

[0125] In practical applications, the actual usable capacity of the battery is corrected based on the instantaneous output power, battery temperature, and cycle count. Specifically, this involves considering that the actual performance of the battery is affected by various factors. For example, battery capacity decreases in low-temperature environments, high temperatures accelerate battery aging, and the maximum usable capacity decays after multiple charge-discharge cycles. Therefore, by using a preset battery performance model or lookup table, combined with the current instantaneous output power, internal temperature sensor data, and the cycle count recorded by the battery management system, the nominal or previously calibrated usable capacity of the battery is dynamically corrected to obtain a more accurate representation of the actual usable capacity. The purpose is to improve the accuracy of battery capacity estimation and avoid misjudgments caused by battery performance degradation.

[0126] Furthermore, estimating the battery's true remaining capacity percentage based on the corrected actual usable capacity and the instantaneous discharge current means, after obtaining the corrected actual usable capacity, combining it with the current instantaneous discharge current, and calculating the percentage of the battery's current remaining capacity relative to the corrected actual usable capacity using the coulomb counting method or a state estimation algorithm. This step aims to provide an accurate indication of remaining capacity that takes into account the actual battery degradation, and its purpose is to provide reliable input for subsequent safe operating time calculations.

[0127] Therefore, calculating the remaining safe operating time based on the actual remaining battery percentage and the predicted energy consumption rate involves converting the estimated actual remaining battery percentage into the actual remaining battery capacity (e.g., by multiplying by the corrected actual usable capacity), and then dividing by the energy consumption rate predicted based on atypical interaction patterns and alternative motion strategies. This calculation result represents the remaining time the battery can support the robot's safe operation under the current operating state and predicted motion strategy. The aim is to provide the decision-making system with a quantitative, high-precision indicator of safe operating time to support more informed strategy selection.

[0128] This application's solution continuously acquires the battery's instantaneous discharge current and voltage, enabling real-time capture of the battery's dynamic operating state. Based on this real-time data, the instantaneous output power can be accurately calculated, providing crucial dynamic data for subsequent battery capacity adjustments. Given that the battery's actual usable capacity is affected by various factors such as temperature, cycle count, and instantaneous load, this solution combines instantaneous output power, battery temperature, and cycle count to correct the battery's actual usable capacity, overcoming the limitations of traditional methods that rely solely on nominal capacity or simple charge counting. This correction ensures that the estimated actual remaining battery charge percentage more closely reflects reality. Finally, by combining this highly accurate actual remaining charge percentage with the predicted energy consumption rate, a more reliable and accurate remaining safe operating time can be calculated. This refined battery state assessment mechanism effectively solves the problem of inaccurate safe operating time prediction caused by dynamic changes in battery performance in complex pipeline environments, providing solid data support for the adaptive decision-making of the soft robot body.

[0129] Through the above technical solution, this application can significantly improve the accuracy and reliability of task planning and risk assessment for soft robots in complex pipeline environments. Specifically, by continuously monitoring the instantaneous discharge current and voltage of the battery, and combining this with instantaneous output power, battery temperature, and cycle count to dynamically correct the actual available capacity, the estimation of the battery's true remaining charge percentage becomes more accurate. This high-precision charge estimation, in turn, makes the calculation of the remaining safe operating time more reliable, avoiding misjudgments caused by battery performance degradation or environmental factors. Compared to assessment methods that rely solely on coarse charge information, this solution can provide the robot with a more accurate prediction of its endurance, thereby optimizing the selection of alternative motion strategies, effectively reducing the risk of robot downtime due to battery depletion during task execution, and improving the likelihood of task completion and overall operational safety.

[0130] In some preferred embodiments, assuming a soft robot detects atypical interaction patterns on the inner wall of a pipe (e.g., localized pipe contraction or the presence of viscous fluid), it needs to evaluate several alternative motion strategies. One alternative strategy is "creeping through," with an expected energy consumption rate of X joules / second; another is "rapid sprinting," with an expected energy consumption rate of Y joules / second (Y > X). To accurately assess the remaining safe operating time of these strategies, the robot's battery management system continuously acquires the instantaneous discharge current and voltage of the battery. For example, at a certain moment, the instantaneous discharge current is 2A and the voltage is 12V, resulting in an instantaneous output power of 24W. Simultaneously, an internal temperature sensor displays a battery temperature of 45°C, and the battery management system records 500 charge-discharge cycles completed. Based on a preset battery performance model, the system corrects the actual usable capacity of the battery, for example, from the nominal 10000mAh to 9000mAh. Based on the corrected actual usable capacity and the current instantaneous discharge current, the actual remaining battery capacity percentage is estimated to be 60%. If the predicted energy consumption rate is X joules per second, the remaining safe operating time can be calculated to be Z minutes. If the "rapid sprint" strategy is selected, with an energy consumption rate of Y joules per second, the remaining safe operating time will be shortened to Z' minutes. Through this refined calculation, the decision-making system can more accurately weigh the endurance risks of different strategies. For example, when the battery is low, the lower-energy-consumption "creeping through" strategy can be prioritized, even if its task completion time is slightly longer, ensuring that the robot body can safely operate until the task is completed or returns.

[0131] In some embodiments described above, this application proposes continuously acquiring potential risk information in the current environment, including flammable gas concentration and communication signal strength with a remote processing system. However, in actual pipelines with complex deformation zones, the communication environment can be extremely harsh, with signal attenuation, obstruction, or interference, leading to unstable or even interrupted communication quality with the remote processing system. This affects the accurate acquisition and timely transmission of potential risk information, potentially jeopardizing the safe operation and task execution of the soft robot. Therefore, this application further proposes a method to optimize the acquisition of potential risk information, particularly communication signal strength, to ensure that the soft robot can continuously and reliably acquire and process environmental risk information in complex environments.

[0132] The aforementioned continuous acquisition of potential risk information in the current environment, including flammable gas concentration and communication signal strength with the remote processing system, specifically includes:

[0133] Multiple wireless signal relay units are deployed, and the wireless signal relay units receive communication signals from the robot body;

[0134] When a decrease in communication quality with the remote processing system is detected, the posture of the robot body and the transmission power and direction of the wireless signal relay unit are adjusted to optimize the strength of the communication signal and the data transmission delay.

[0135] Based on the urgency of the current task and the importance of data transmission, the data to be sent to the remote processing system is prioritized, encoded, and fragmented.

[0136] When communication quality is extremely poor or completely interrupted, activate the local emergency data processing and decision-making mode and suspend the sending of non-essential data to the remote processing system.

[0137] Specifically, deploying multiple wireless signal relay units aims to build a redundant and flexible communication network. These relay units can be pre-deployed at specific locations within the pipeline or carried by other auxiliary robots. Their main function is to receive communication signals emitted by the soft robot itself and forward them to a remote processing system, thereby effectively extending the communication distance and overcoming signal attenuation and blockage problems inside the pipeline. The wireless signal relay unit can be understood as a device with signal reception, amplification, and forwarding capabilities, designed to enhance the robustness of the communication link.

[0138] Furthermore, when a decline in communication quality with the remote processing system is detected, such as when the signal strength falls below a preset threshold or the data transmission delay increases significantly, the system will trigger a series of optimization measures. Specifically, the robot's posture adjustment may include changing its body's bending angle, twisting direction, or local expansion state to find the optimal signal transmission path. Simultaneously, the transmission power of the wireless signal relay unit can be dynamically increased to enhance signal penetration, and its transmission direction can be precisely adjusted to align with the robot or the next relay node, thereby maximizing the strength of the communication signal and minimizing data transmission delay.

[0139] Furthermore, to ensure the transmission of critical information under limited or unstable communication bandwidth, this application introduces a data priority encoding and fragmentation mechanism. Specifically, data to be sent to the remote processing system is classified and encoded based on the urgency of the current task (e.g., whether an emergency evacuation situation is underway) and the importance of the data transmission (e.g., security warning data, critical sensor readings, etc.). High-priority data will be transmitted first and may be fragmented into smaller data packets to improve the transmission success rate. The aim is to ensure that the most critical information can be delivered to the remote processing system in a timely manner, even under conditions of limited communication.

[0140] Furthermore, when communication quality is extremely poor or completely interrupted, the soft robot will activate a local emergency data processing and decision-making mode. This means that the robot will no longer rely entirely on instructions from the remote processing system, but will autonomously conduct risk assessments and make action decisions based on its locally stored preset rules, environmental models, and real-time sensor data. In this mode, sending non-essential data (e.g., routine status reports, non-urgent image data) to the remote processing system will be suspended to conserve limited communication resources and prioritize handling local emergencies, ensuring the robot's autonomous and safe operation.

[0141] This application's solution effectively extends communication coverage and enhances signal transmission stability by deploying multiple wireless signal relay units, thereby overcoming the challenges of signal attenuation and obstruction in areas with complex pipeline deformation. When communication quality deteriorates, the communication link can be proactively optimized by dynamically adjusting the robot's posture and the transmission power and direction of the relay units, ensuring that potential risk information (especially the communication signal strength itself) can be continuously and accurately acquired. Furthermore, by prioritizing and fragmenting data, this application ensures that the most critical safety and task-related data are transmitted first when communication bandwidth is limited, avoiding the risk of losing important information due to communication interruptions. Moreover, when communication is completely interrupted, a local emergency data processing and decision-making mode is activated, enabling the soft robot to make autonomous judgments and decisions based on local information, thereby avoiding dangerous situations due to loss of remote control and ensuring the continuity of tasks and the safety of the robot under extreme communication conditions.

[0142] Through the aforementioned technical solutions, this application significantly improves the reliability and real-time performance of soft robots in acquiring potential risk information, particularly communication signal strength, in complex pipeline environments. The deployment of wireless signal relay units and the mechanism for dynamically adjusting communication parameters effectively solve the problems of communication susceptibility to environmental interference and severe signal attenuation in traditional solutions, ensuring stable information flow between the robot and the remote processing system. Data priority encoding and fragmentation mechanisms guarantee the priority transmission of critical information when communication resources are limited, avoiding decision-making delays or errors due to communication instability. Especially in the event of a complete communication outage, the introduction of a local emergency data processing and decision-making mode endows the robot with the ability to autonomously respond to emergencies, greatly enhancing the robot's task robustness and safety, effectively reducing the risk of task failure or robot damage caused by communication failures. This enables soft robots to complete exploration and operation tasks in complex pipeline areas more safely and efficiently.

[0143] In some preferred embodiments, it is assumed that a soft robot is performing an inspection task in a long, winding underground pipe. The pipe's interior, with its metallic structure and humid environment, severely interferes with wireless signal transmission. To address this challenge, three wireless signal relay units are deployed along the route before the robot enters the pipe. When the robot reaches a sharp bend, the communication signal strength with the remote processing system suddenly drops, accompanied by an increase in data transmission latency. At this point, the robot's internal communication monitoring module detects this change and immediately triggers a communication optimization strategy. The robot body slightly adjusts its local deformation, for example, by fine-tuning the bending angle of its flexible segments to change its antenna orientation. Simultaneously, the two wireless signal relay units closest to the robot dynamically adjust their transmission power and direction based on the received signal strength feedback, forming a more stable signal bridge.

[0144] Meanwhile, the image data of pipe wall cracks being collected by the robot (non-urgent) will be temporarily downgraded in transmission priority, while sensor data on a sudden increase in the concentration of flammable gas in the pipe (urgent) will be immediately prioritized, encoded, and fragmented, and sent first through the optimized communication link. If the communication quality deteriorates further, or even completely fails, the robot will immediately activate its local emergency data processing and decision-making mode. In this mode, the robot will suspend the transmission of non-essential image data and instead use the locally stored pipe map and preset safety protocols to autonomously assess the risks posed by the increased flammable gas concentration. It may autonomously execute strategies such as partial retreat or searching for a safe area, while continuously attempting to restore communication with the relay unit until communication is restored or the local safety protocol instructs it to take other actions.

[0145] In some implementations, the steps described above—adjusting the local deformation and motion rhythm of the robot body in response to atypical interaction patterns to optimize the physical interaction between the robot body and the inner wall of the pipe, and driving the robot body to deform and move according to the adjusted local deformation and motion rhythm—require predicting the energy consumption rate and heat accumulation rate that each alternative motion strategy may result in, based on the currently identified atypical interaction patterns and alternative motion strategies. However, such predictions may be based on static or general models and fail to fully consider the dynamic changes in the internal environment of the pipe, such as the rheological behavior of the medium on the inner wall of the pipe or the dynamic response of obstacles. If the above problems are not addressed, the prediction results may deviate from the actual situation, resulting in the robot body selecting a non-optimal motion strategy, thereby affecting the efficiency of the task, energy consumption, and the safety of the robot body.

[0146] In response, this application further proposes the above-mentioned steps of predicting the energy consumption rate and heat accumulation rate that each alternative motion strategy may result from based on the currently identified atypical interaction patterns and alternative motion strategies. By dynamically identifying the rheological behavior of the medium inside the pipe wall or the dynamic response characteristics of obstacles, and correcting the prediction parameters accordingly, the accuracy and adaptability of the prediction are improved.

[0147] Based on the currently identified atypical interaction patterns and alternative exercise strategies, the above predicts the potential energy expenditure rate and calorie accumulation rate for each alternative exercise strategy, including:

[0148] Based on the dynamic changes in the physical interaction information, the rheological behavior of the medium inside the pipe wall or the dynamic response characteristics of obstacles can be identified.

[0149] Based on the identified rheological behavior or the dynamic response characteristics, the prediction parameters for energy consumption rate and heat accumulation rate are corrected;

[0150] Based on the corrected prediction parameters, the energy consumption rate and heat accumulation rate of each alternative exercise strategy in the current dynamic environment are estimated.

[0151] Specifically, "identifying the rheological behavior of the medium on the inner wall of the pipe or the dynamic response characteristics of obstacles based on the dynamic changes of the physical interaction information" refers to analyzing the changing patterns of the continuously acquired physical interaction information (e.g., pressure, friction, deformation feedback, etc.) between the soft robot body and the inner wall of the pipe over time to determine the dynamic response of the flow characteristics of the medium on the inner wall of the pipe (e.g., non-Newtonian fluid behavior such as viscosity, shear thinning, or shear thickening) or obstacles (e.g., movable deposits, elastically deformable pipe structures) to the robot body's motion. For example, when the robot body moves in the pipe, if the tactile sensor detects a sudden increase in friction or resistance and exhibits periodic fluctuations, this may indicate the presence of viscous fluid or elastic obstacles on the inner wall of the pipe.

[0152] The phrase "adjusting the prediction parameters for energy consumption rate and heat accumulation rate based on the identified rheological behavior or dynamic response characteristics" can be understood as follows: once a specific rheological behavior or dynamic response characteristic is identified, the system adjusts the key parameters used to predict energy consumption and heat accumulation based on a preset physical model or machine learning model. For example, if a high-viscosity fluid is identified, the energy consumption parameter related to frictional resistance will be increased; if an elastic obstacle is identified, the energy consumption parameter related to work done by deformation will be adjusted. These parameter adjustments aim to make the prediction model more closely reflect the current actual environmental conditions.

[0153] In practical applications, "estimating the energy consumption rate and heat accumulation rate of each alternative motion strategy under the current dynamic environment based on the corrected prediction parameters" specifically involves using the corrected prediction parameters to simulate or calculate each preset alternative motion strategy (e.g., different creep frequencies, deformation amplitudes, propulsion speeds, etc.), thereby determining the energy consumption rate and heat accumulation rate that might result from executing the strategy under the current dynamic environment. For example, for an alternative strategy, under the corrected parameters, the driving torque and deformation work required when passing through a specific rheological medium or dynamic obstacle can be calculated, thereby estimating the corresponding energy consumption and heat generation.

[0154] This application's solution, by introducing the identification of dynamic changes in physical interaction information, can capture in real time the rheological behavior of the medium inside the pipe or the dynamic response characteristics of obstacles. It is precisely because these dynamic characteristics are accurately identified that the predicted parameters for energy consumption rate and heat accumulation rate can be corrected promptly and accurately. This correction mechanism ensures that the prediction model can adapt to the constantly changing pipe environment, thereby avoiding errors caused by predictions based on static or general models. By estimating the energy consumption and heat accumulation of each alternative motion strategy in a dynamic environment, the system can obtain more realistic and reliable strategy evaluation data, providing a solid foundation for subsequent action strategy selection.

[0155] Through the above technical solution, soft robots can more accurately predict the energy required and heat generated to execute different motion strategies in complex dynamic pipe environments. This dynamic and adaptive predictive capability significantly improves the accuracy of strategy selection and avoids energy waste or overheating risks caused by inaccurate predictions. Compared to methods that rely solely on static models for prediction, this application enables the robot to select safer and more efficient motion strategies when facing rheological media or dynamic obstacles, thereby effectively extending task execution time, increasing the likelihood of task completion, and reducing the risk of robot wear and failure.

[0156] In some preferred embodiments, a specific example is given below. Suppose a soft robot, in a task to detect leaks inside a pipe, enters a section of the pipe filled with high-viscosity slurry.

[0157] First, the robot's tactile and pressure sensors continuously acquire information about its physical interactions with the mud. The system analyzes the dynamic changes in these interactions; for example, it detects that during the robot's peristaltic movement, the shear force and resistance on its surface increase significantly, exhibiting a nonlinear response related to the mud's rheological properties. Based on these dynamic changes, the system identifies the medium inside the pipe as exhibiting the rheological behavior of a high-viscosity non-Newtonian fluid.

[0158] Secondly, based on the identified high-viscosity non-Newtonian fluid behavior, the system corrects relevant parameters in its internally stored energy consumption and heat accumulation prediction models. For example, the coefficient related to frictional resistance is significantly increased, and the parameter related to deformation efficiency is adjusted according to the viscoelastic properties of the mud. Finally, using these corrected prediction parameters, the system estimates various alternative motion strategies (e.g., low-speed large-deformation creep, medium-speed small-deformation creep, high-speed vibration propulsion, etc.). The results show that the low-speed large-deformation creep strategy, although with a short single propulsion distance, has the lowest energy consumption and heat accumulation rates, and causes the least wear on the robot body. While the high-speed vibration propulsion strategy is theoretically fast, it leads to extremely high energy consumption and heat accumulation in high-viscosity mud, potentially rapidly depleting the battery or causing overheating. Therefore, the system can select the most suitable low-speed large-deformation creep strategy for the current high-viscosity mud environment based on more accurate prediction results, thereby ensuring that the robot body can safely and effectively pass through this complex area and complete the detection task.

[0159] In some embodiments described above, this application proposes to assess the remaining safe operating time and task completion probability by combining the robot's operational status information, potential risk information in the environment, and alternative motion strategies. However, in practical applications, the local characteristics of the pipe's inner wall can significantly affect the robot's operational status, environmental risks, and the actual effectiveness of different motion strategies. Failure to fully consider these local characteristics may lead to discrepancies between the assessment results and the actual situation, thereby affecting the accuracy and adaptability of the robot's decision-making. Therefore, this application further proposes a more refined assessment method that dynamically adjusts the weights of various assessment parameters by identifying the local characteristics of the pipe's inner wall, thereby achieving a more accurate and environmentally adaptable strategy assessment.

[0160] The above-mentioned assessment, combining operational status information, potential environmental risk information, and alternative movement strategies, evaluates the remaining safe operating time and task completion probability of alternative movement strategies, specifically including:

[0161] Based on physical interaction information, identify local features of the inner wall of the pipe;

[0162] Based on the characteristics of the local area, the weights of the robot's operating status information, potential environmental risk information, and strategy prediction results are adjusted, and the remaining safe operating time and task completion probability of each alternative motion strategy are evaluated.

[0163] Specifically, identifying local features of the pipe's inner wall refers to using physical interaction information acquired by various sensors (such as tactile sensors, vision sensors, and sonar sensors) mounted on the robot to perceive and analyze the physical properties and geometry of the pipe's inner wall. These local features may include, but are not limited to, the roughness, material, diameter variation, degree of curvature, presence of obstacles, viscosity, or corrosiveness of the fluid medium. For example, tactile sensors can detect the friction or texture of the inner wall, while vision sensors can identify the pipe's geometry or obstacles. Adjusting the weights based on local features can be understood as dynamically changing the proportions of operational status information (such as battery level, internal temperature, and sensor health), potential environmental risk information (such as flammable gas concentration and communication signal strength), and strategy prediction results (such as energy consumption rate and heat accumulation rate) in the overall assessment based on the specific conditions of the identified pipe inner wall. The aim is to make the assessment process more closely aligned with the current specific environment, thereby improving the accuracy of the assessment and the effectiveness of the decision-making. For example, when the robot body is in a narrow and rough pipe bend, the weight of internal temperature and the health status of tactile sensors in the operating status information may be increased. At the same time, the weight of the predicted energy consumption rate and heat accumulation rate will also be increased accordingly to prioritize the wear and overheating risk of the robot body.

[0164] This application's solution introduces the identification of local features within the pipe's inner wall, enabling the robot to dynamically adjust the weights of evaluation parameters based on its specific environment. This mechanism addresses the potential problems of fixed or generalized evaluation parameter weights in traditional methods, avoiding evaluation biases caused by inappropriate weights in complex areas. Specifically, when the robot enters an area with unique local features, such as a high-friction or highly corrosive pipe section, the system, based on these identified features, correspondingly increases the weights of operational status information related to friction and corrosion (such as internal temperature and material wear) and potential environmental risk information (such as corrosive gas concentration). Therefore, when evaluating alternative motion strategies, these key factors receive higher priority, prompting the robot to choose more conservative and safer strategies, rather than simply pursuing task completion speed or minimum energy consumption. This dynamic adjustment mechanism ensures that the evaluation results more accurately reflect the real risks and opportunities in the current environment, thus providing a more solid foundation for the robot's adaptive decision-making.

[0165] Through the above technical solution, this application can significantly improve the accuracy and robustness of the soft robot's adaptive capabilities in complex pipeline environments. By identifying local features of the pipeline's inner wall and dynamically adjusting the weights of evaluation parameters accordingly, the robot's decision-making process becomes more intelligent and context-aware. This enables the robot to more accurately assess the safety and effectiveness of different motion strategies when facing various complex deformation areas, thereby avoiding energy waste, accelerated component wear, or even task failure caused by inappropriate strategies. Furthermore, this refined evaluation mechanism helps optimize resource allocation; for example, efficiency can be prioritized in low-risk areas, while safety can be prioritized in high-risk areas, thus maximizing the robot's lifespan and reducing operational risks while ensuring the likelihood of task completion.

[0166] In some preferred embodiments, assuming a soft robot, while exploring an industrial pipeline, identifies severe corrosion and localized narrowing on the inner wall of the pipeline using physical interaction information acquired by its tactile and visual sensors, the system identifies this "corrosion and localized narrowing" as a localized feature of the pipeline's inner wall. In response to this feature, the system dynamically adjusts the weights of evaluation parameters: for example, increasing the weight of the robot's "internal temperature" and "health of the tactile sensors" in the operational status information, as corrosion and narrowing areas may lead to higher friction and wear; simultaneously, increasing the weight of "flammable gas concentration" in the environmental potential risk information, as corrosion may be accompanied by a risk of gas leakage; and correspondingly increasing the weights of "heat accumulation rate" and "energy consumption rate" in the strategy prediction results. Based on these adjusted weights, the system re-evaluates all alternative motion strategies. For example, a fast-moving strategy initially considered efficient might be downgraded after weight adjustments due to its high heat accumulation and wear risk, while a slower, gentler creeping strategy, even with slightly higher energy consumption, would be preferred because of its lower wear on the robot and slower heat accumulation. This weight adjustment based on local features allows the robot to make decisions that better suit the actual needs of the current environment, thus enabling it to safely and efficiently traverse complex areas.

[0167] In some embodiments described above, while adjusting the weights of the robot's operational status information, potential environmental risks, and strategy prediction results based on local features of the pipe's inner wall is proposed, in real-world complex pipe environments, local features may encompass multiple rheological behaviors or dynamic response characteristics. These characteristics may interact in complex ways. Adjusting weights based solely on single or simple local features may fail to accurately reflect the true risks of the current environment and the robot's actual state, leading to biased evaluation results and impacting the optimization of action strategies. Failure to address this issue could result in the robot making suboptimal decisions in complex environments, potentially even posing safety risks. Therefore, this application further proposes a more refined weight adjustment method. By classifying rheological behaviors or dynamic response characteristics and analyzing their coupling effects, more precise and dynamic weight adjustment can be achieved.

[0168] In some embodiments of this application, adjusting the weights of the robot's operational state information, potential environmental risk information, and strategy prediction results based on local region characteristics includes:

[0169] The identified rheological behaviors or dynamic response features are classified to identify the corresponding rheological modes or dynamic response modes;

[0170] Based on the classified rheological modes or dynamic response modes, analyze the interactions between different modes and determine the coupling effect;

[0171] Based on the classified rheological modes or dynamic response modes and the coupling effect, the weights of the robot's operating state information, the potential environmental risk information, and the strategy prediction results are dynamically adjusted.

[0172] Specifically, classifying identified rheological behaviors or dynamic response characteristics and identifying corresponding rheological modes or dynamic response modes involves using pattern recognition and cluster analysis on features extracted from physical interaction information to summarize complex fluid dynamic behaviors (e.g., turbulence, laminar flow, shear thinning, thixotropy, etc.) or the mechanical responses of obstacles (e.g., elastic deformation, plastic deformation, frictional damping, etc.) into several discrete, identifiable modes. For example, machine learning algorithms (e.g., support vector machines, neural networks, or decision trees) can be used to train sensor data to automatically identify different rheological modes (e.g., high-viscosity fluids, low-viscosity fluids, non-Newtonian fluids, etc.) or dynamic response modes (e.g., rigid obstacles, flexible obstacles, viscous deposits, etc.). The aim is to simplify complex environmental information into actionable categories, laying the foundation for subsequent refined analysis.

[0173] This process involves analyzing the interactions between different modes based on their classified rheological or dynamic response modes, and determining the coupling effects. This can be understood as assessing the combined impact of multiple modes—either simultaneously or sequentially—on the robot's operational status and environmental risks. For example, when a robot faces both a high-viscosity fluid (a rheological mode) and a flexible obstacle (a dynamic response mode), these two modes may mutually reinforce or weaken their effects on the robot's motion drag, energy consumption, and local deformation. By establishing physical or data-driven models, the coupling effects between these modes can be quantified. For instance, a high-viscosity fluid may increase the deformation drag of a flexible obstacle, thereby further increasing energy consumption. The aim is to gain a more comprehensive understanding of the impact of complex environments on the robot and avoid errors arising from isolated analyses.

[0174] In practical applications, based on the classified rheological modes or dynamic response modes and the coupling effects, the weights of the robot's operational status information, the potential environmental risks, and the strategy prediction results are dynamically adjusted. Specifically, the importance of each evaluation parameter in the overall evaluation is adjusted in real time according to the identified specific mode combinations and their coupling effects. For example, when a strong coupling effect between high-viscosity fluid and flexible obstacles is identified, the weights of energy consumption rate and internal temperature in the evaluation can be increased, while the weight of task completion probability can be decreased, prioritizing the safe operation of the robot. This dynamic adjustment mechanism ensures that the evaluation process can adapt to rapidly changing environments, making the final selected action strategy more robust and safer.

[0175] This application addresses the problem of evaluation bias that may arise from adjusting weights based solely on single or simple local region features in complex pipeline environments by introducing classification and coupling effect analysis of rheological behavior or dynamic response characteristics. Through this technical solution, the accuracy and robustness of the soft robot's adaptive capabilities in complex pipeline environments are significantly improved. Compared to solutions that adjust weights based solely on local region features, this solution, through refined classification of rheological behavior or dynamic response characteristics and in-depth analysis of the coupling effects between different modes, makes the weight adjustments for robot operating status information, potential environmental risks, and strategy prediction results more closely aligned with actual working conditions. This refined and dynamic weight adjustment mechanism can more accurately capture the true impact of multi-factor interactions in complex environments, thereby avoiding decision-making biases caused by simple evaluations. Consequently, when facing variable and complex pipeline deformation areas, the robot can make more informed and safer action strategy choices, effectively reducing the possibility of equipment damage or task failure due to misjudgment risks, and significantly improving the safety and efficiency of task completion.

[0176] In some embodiments described above, this application proposes adjusting the weights of the robot's operational status information, potential environmental risk information, and strategy prediction results based on local characteristics of the pipe's inner wall to assess the remaining safe operating time and task completion probability of alternative motion strategies. However, in real-world complex pipe environments, the physical interaction information of the pipe's inner wall may exhibit highly dynamic and variable characteristics. For example, rheological behavior or obstacle responses may switch rapidly, or multiple modes may coexist and overlap. If weight adjustments are made solely based on static or slowly changing local characteristics, the instantaneous complexity of the current environment may not be reflected in a timely and accurate manner, leading to discrepancies between the assessment results and the actual situation, thus affecting the effectiveness of the soft robot's adaptive capabilities. To address this, this application further proposes a more refined weight adjustment mechanism. By analyzing the dynamic changes in physical interaction information in real time, it identifies and responds to rapid switching points or overlapping areas of rheological modes or dynamic response modes, thereby achieving dynamic and precise adjustment of the weights of operational status information, potential environmental risk information, and strategy prediction results.

[0177] According to the above method, adjusting the weights of the robot's operating state information, the potential environmental risk information, and the strategy prediction result based on the local region characteristics includes:

[0178] Based on the continuously acquired physical interaction information, the instantaneous rate of change and trend of change of the physical interaction information are analyzed in real time;

[0179] Based on the instantaneous rate of change and the trend of change, identify the rapid switching points or superposition regions of rheological modes or dynamic response modes;

[0180] In response to the rapid switching or superposition of the rheological mode or dynamic response mode, the weights of the robot's operating status information, the potential environmental risk information, and the strategy prediction results are dynamically adjusted according to the type and intensity of the switching or superposition.

[0181] If multiple modalities are identified as superimposed, the weights are adjusted according to the superposition effect of the multiple modalities.

[0182] Specifically, the continuously acquired physical interaction information refers to the physical quantity data such as contact, friction, pressure, and deformation that occur between the soft robot body and the inner wall of the pipe. This data can be acquired through tactile sensors, pressure sensors, and vision sensors integrated on the robot body. Real-time analysis of the instantaneous rate of change and trend of this physical interaction information aims to capture subtle and rapid dynamic characteristics of the environment or interaction state, such as sudden changes in the viscosity of the medium inside the pipe, rapid movement of obstacles, or instantaneous changes in the cross-sectional shape of the pipe. The instantaneous rate of change can be understood as the magnitude of change of a physical quantity within a very short time, while the trend refers to the direction and persistence of this change.

[0183] Identifying rapid switching points or overlapping regions of rheological modes or dynamic response modes based on the instantaneous rate of change and trend refers to processing the analysis results using pattern recognition or machine learning algorithms to determine which rheological mode (e.g., Newtonian fluid, non-Newtonian fluid, plastic fluid, etc.) or dynamic response mode (e.g., elastic collision, viscous damping, frictional sliding, etc.) the current interaction state belongs to, and paying particular attention to the moments when rapid transitions occur between these modes (switching points) or the regions where multiple modes coexist and influence each other (overlapping regions). For example, a robot may move from a low-viscosity region to a high-viscosity region, or be subjected to both compression and friction while passing through a narrow bend.

[0184] In practical applications, responding to rapid switching or superposition of rheological or dynamic response modes, and dynamically adjusting the weights of the robot's operational status information, potential environmental risks, and strategy prediction results based on the type and intensity of the switching or superposition, means that once a rapid switching or superposition of modes is detected, the system will assess the impact of this switching or superposition on the robot's operational status, environmental risks, and strategy effectiveness based on a preset rule base or learning model. For example, if the sudden appearance of a high-intensity viscous fluid mode is detected, it may be necessary to increase the weight of energy consumption rate prediction while decreasing the weight of task completion probability, because the robot may require more energy to overcome resistance in this environment. The type of switching or superposition can refer to a specific combination of modes, while the intensity refers to the degree of hindrance or influence on the robot's movement.

[0185] Furthermore, if multiple modal superpositions are identified, the weights are adjusted based on the superposition effect of these modalities. This means that when multiple rheological modes or dynamic response modes act on the robot body simultaneously, the system not only considers the influence of a single mode but also analyzes the potential synergistic or antagonistic effects between these modes. For example, in regions with both high friction and high viscosity, the superposition effect may lead to greater energy consumption and heat accumulation than considering friction or viscosity alone. Therefore, a larger adjustment to the corresponding weights is needed to more accurately assess the risks and task feasibility. The weighting adjustment can employ linear weighting, nonlinear weighting, or weighting methods based on fuzzy logic.

[0186] This application's solution, through real-time and refined dynamic analysis of continuously acquired physical interaction information, can capture the instantaneous changes and multimodal superposition phenomena of rheological behavior or obstacle response in complex deformation regions of pipelines. Through this technical solution, this application can significantly improve the adaptive capability and decision-making accuracy of soft robots when traversing complex deformation regions of pipelines. Compared to methods that only adjust weights based on local region features, this solution analyzes the instantaneous rate and trend of change in physical interaction information in real time, and identifies rapid switching points or superposition regions of rheological modes or dynamic response modes, enabling the system to respond more quickly and precisely to dynamic changes in the environment. Especially when facing complex situations with multiple superimposed modes, by considering their superposition effects and applying weighted adjustments, it can more comprehensively and accurately assess potential risks and task feasibility, thereby avoiding energy waste, robot damage, or task failure caused by insufficient single-modal analysis or sluggish dynamic response. This dynamic and refined weight adjustment mechanism ensures that the soft robot can always choose the optimal action strategy in highly uncertain and rapidly changing environments, effectively improving its safety and task completion efficiency.

[0187] refer to Figure 4 , Figure 4 This is a schematic diagram of the structure of a system for collaboration between a small edge model and a large cloud model of a soft robot, provided in an embodiment of the present invention, including:

[0188] The input terminal is used to acquire physical interaction information between the robot body and the inner wall of the pipe; based on the physical interaction information, it identifies atypical interaction patterns between the robot body and the inner wall of the pipe.

[0189] The adjustment end is used to adjust the local deformation and movement rhythm of the robot body in response to the atypical interaction mode, so as to optimize the physical interaction state between the robot body and the inner wall of the pipe, and drive the robot body to deform and move according to the adjusted local deformation and movement rhythm.

[0190] The feedback end is used to continuously acquire the physical interaction information and adjust the local deformation and motion rhythm based on the continuously acquired physical interaction information.

[0191] This system uses an input terminal to perceive the physical interaction information between the robot body and the inner wall of the pipe in real time, and intelligently identifies atypical interaction patterns, thus providing a basis for subsequent adaptive adjustments. Upon receiving the identification results of atypical interaction patterns, the adjustment terminal can quickly respond and dynamically adjust the local deformation and movement rhythm of the robot body to optimize its physical interaction with the pipe wall, ensuring the robot body can smoothly pass through complex areas. Simultaneously, the feedback terminal continuously monitors physical interaction information and provides real-time feedback corrections to the adjustment terminal's strategy based on this information, forming a closed-loop adaptive control mechanism. This system architecture enables the soft robot to autonomously and efficiently cope with the small but complex structural changes inside the pipe, significantly improving its ability to pass through harsh environments and its task completion efficiency, effectively solving the problems of poor adaptability and easy jamming in complex environments associated with traditional methods.

[0192] The specific methods and principles for acquiring physical interaction information between the robot body and the inner wall of the pipe, identifying atypical interaction patterns, adjusting local deformation and motion rhythm, and providing feedback adjustments have already been described in the above embodiments, and will not be repeated here. It should be emphasized that the system proposed in this application achieves a clearer system architecture and more efficient functional coordination by modularizing these functions into input, adjustment, and feedback ends.

[0193] Specifically, the input end can be configured to include various sensors and data preprocessing modules. For example, the input end can integrate a tactile sensor array to directly measure the contact pressure and friction between the robot body and the inner wall of the pipe. Alternatively, the input end can employ a vision-based perception module to indirectly acquire physical interaction information by analyzing image changes on the robot body surface or the inner wall of the pipe. Furthermore, the input end can also include acoustic sensors to capture the sound wave characteristics generated during the robot's movement. The raw data collected by these sensors is then transmitted to a data processing unit, which performs preprocessing operations such as filtering, noise reduction, and feature extraction to facilitate subsequent recognition of atypical interaction patterns.

[0194] The adjustment module can be configured to include a decision-making module and multiple actuator control modules. For example, the decision-making module can select appropriate local deformation and motion rhythm adjustment strategies based on a pre-defined rule base or expert system and atypical interaction patterns identified by the input. Alternatively, the decision-making module can employ a reinforcement learning-based algorithm to learn and optimize the adjustment strategy through continuous interaction with the environment. The actuator control modules are responsible for converting the adjustment commands output by the decision-making module into precise control signals for the robot's internal drive units (such as hydraulic / pneumatic actuators, motors, etc.), thereby achieving changes in the robot's local deformation and motion rhythm.

[0195] The feedback end can be configured to continuously monitor the physical interaction information acquired by the input end and compare it with the current adjustment strategy of the adjustment end. For example, the feedback end can include a performance evaluation module to evaluate the effectiveness of the current adjustment strategy in real time, such as the robot's throughput efficiency, energy consumption, and the pressure distribution of interaction with the inner wall of the pipe. When the evaluation results show that the current strategy has failed to achieve the expected optimization effect, the feedback end will feed this information back to the adjustment end, prompting the adjustment end to re-evaluate and select a better adjustment strategy. As a preferred implementation, the feedback end can also integrate a predictive model to predict possible future interaction patterns based on historical data and current trends, thereby achieving proactive adjustment.

[0196] The core innovation of the system proposed in this application, which integrates a small edge model of a soft robot with a large cloud model, lies in constructing a closed-loop adaptive control architecture that combines perception, decision-making, execution, and feedback. Compared to traditional soft robots that rely on preset motion planning or simple local obstacle avoidance logic, and schemes that overly depend on remote cloud processing systems for global decision-making, this application's system avoids decision-making errors caused by data simplification by acquiring physical interaction information in real time and with precision, and intelligently recognizing atypical interaction patterns at the input end. The adjustment end can quickly and dynamically adjust local deformation and motion rhythm based on the recognition results, directly optimizing the physical interaction state with the inner wall of the pipe at the robot body level, effectively solving problems such as decreased robot motion efficiency and abnormal local pressure distribution. In addition, the feedback end continuously monitors and corrects the adjustment effect, ensuring that the system can maintain optimal adaptive capability when facing complex and ever-changing pipe environments. This systematic design significantly improves the autonomous decision-making and passage capabilities of soft robots in harsh and communication-limited environments, thereby improving the success rate and safety of inspection tasks.

[0197] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for soft robot edge small model and cloud large model collaboration, characterized in that, The method comprises: acquiring physical interaction information between a robot body and an inner wall of a pipeline; identifying an atypical interaction mode between the robot body and the inner wall of the pipeline according to the physical interaction information; adjusting local deformation and motion rhythm of the robot body in response to the atypical interaction mode, so as to optimize the physical interaction state between the robot body and the inner wall of the pipeline, and drive the robot body to deform and move according to the adjusted local deformation and motion rhythm; continuously acquiring the physical interaction information and feedback adjusting the local deformation and motion rhythm according to the continuously acquired physical interaction information.

2. The method of claim 1, wherein, The method of adjusting the local deformation and motion rhythm of the robot body in response to the atypical interaction mode, so as to optimize the physical interaction state between the robot body and the inner wall of the pipeline, and drive the robot body to deform and move according to the adjusted local deformation and motion rhythm, comprises: continuously acquiring running state information of the robot body, the running state information including battery power, internal temperature, and health degree of a tactile sensor; continuously acquiring potential risk information in the current environment, the potential risk information including flammable gas concentration and communication signal strength with a remote processing system; predicting energy consumption rate and heat accumulation rate that each alternative motion strategy may cause according to the currently identified atypical interaction mode and the alternative motion strategies; evaluating safe running remaining time and task completion possibility of the alternative motion strategies in combination with the running state information, the environmental potential risk information, and the alternative motion strategies; adjusting task priority and selecting an action strategy according to the evaluation results and a preset safety protocol.

3. The method of claim 2, wherein, The method of adjusting task priority and selecting an action strategy according to the evaluation results and a preset safety protocol comprises: if the flammable gas risk is high and the internal temperature has reached a high temperature warning, executing a minimum power consumption safety shutdown strategy; if the flammable gas risk is medium and the internal temperature continues to rise, executing a local retreat strategy; if the immediate safety threat is 0 and the task completion possibility is low, selecting a degradation escape strategy with low energy consumption and small wear on the robot body; if the task completion possibility is high and the comprehensive risk is low, selecting a task continuation strategy.

4. The method of claim 3, wherein, The method of evaluating safe running remaining time and task completion possibility of the alternative motion strategies in combination with the running state information, the environmental potential risk information, and the alternative motion strategies comprises: continuously acquiring instantaneous discharge current and voltage of the battery; calculating instantaneous output power of the battery according to the instantaneous discharge current and voltage; correcting actual available capacity of the battery according to the instantaneous output power, temperature, and cycle number of the battery; estimating real remaining battery capacity percentage according to the corrected actual available capacity and the instantaneous discharge current; calculating safe running remaining time according to the real remaining battery capacity percentage and the predicted energy consumption rate.

5. The method of claim 2, wherein, The method of continuously acquiring potential risk information in the current environment, the potential risk information including flammable gas concentration and communication signal strength with a remote processing system, comprises: deploying a plurality of wireless signal relay units that receive communication signals from the robot body; adjusting the posture of the robot body and the transmission power and direction of the wireless signal relay units to optimize the strength of the communication signals and data transmission delay when detecting a decrease in communication quality with the remote processing system; prioritizing and fragmenting data to be sent to the remote processing system according to the urgency of the current task and the importance of data transmission; activating a local emergency data processing and decision-making mode and suspending the sending of non-essential data to the remote processing system when the communication quality is extremely poor or completely interrupted.

6. The method of claim 2, wherein, According to the current identified atypical interaction mode and alternative motion strategy, the energy consumption rate and heat accumulation rate that each alternative motion strategy may cause are predicted, including: According to the dynamic changes of the physical interaction information, the flow behavior of the pipe wall medium or the dynamic response characteristics of the obstacle are identified; According to the identified flow behavior or dynamic response characteristics, the prediction parameters of energy consumption rate and heat accumulation rate are corrected; According to the corrected prediction parameters, the energy consumption rate and heat accumulation rate of each alternative motion strategy in the current dynamic environment are estimated.

7. The method of claim 2, wherein the soft robotic edge small model and the cloud large model are coordinated. The safety operation remaining time and task completion possibility of the alternative motion strategy are evaluated in combination with the running state information, the environmental potential risk information and the alternative motion strategy, including: According to the physical interaction information, the local area characteristics of the pipe wall are identified; According to the local area characteristics, the weights of the running state information of the robot body, the environmental potential risk information and the strategy prediction results are adjusted, and the safety operation remaining time and task completion possibility of each alternative motion strategy are evaluated.

8. The method of claim 7, wherein, According to the local area characteristics, the weights of the running state information of the robot body, the environmental potential risk information and the strategy prediction results are adjusted, including: Classify the identified flow behavior or dynamic response characteristics to identify the corresponding flow mode or dynamic response mode; According to the classified flow mode or dynamic response mode, the interaction between different modes is analyzed to determine the coupling effect; According to the classified flow mode or dynamic response mode and the coupling effect, the weights of the running state information of the robot body, the environmental potential risk information and the strategy prediction results are dynamically adjusted.

9. The method of claim 7, wherein, According to the local area characteristics, the weights of the running state information of the robot body, the environmental potential risk information and the strategy prediction results are adjusted, including: According to the continuously acquired physical interaction information, the instantaneous change rate and change trend of the physical interaction information are analyzed in real time; According to the instantaneous change rate and change trend, the rapid switching point or superposition area of the flow mode or dynamic response mode is identified; In response to the rapid switching or superposition of the flow mode or dynamic response mode, the weights of the running state information of the robot body, the environmental potential risk information and the strategy prediction results are dynamically adjusted according to the type and intensity of the switching or superposition; If multiple modalities are identified to be superimposed, the weights are adjusted according to the superimposition effect of the multiple modalities.

10. A system for soft robot edge small model and cloud large model collaboration, characterized in that, Comprise: an input end for acquiring physical interaction information between a robot body and a pipe inner wall; According to the physical interaction information, identify the atypical interaction mode between the robot body and the pipe inner wall; an adjustment end for adjusting the local deformation and motion rhythm of the robot body in response to the atypical interaction mode, so as to optimize the physical interaction state between the robot body and the pipe inner wall, and drive the robot body to deform and move according to the adjusted local deformation and motion rhythm; a feedback end for continuously acquiring the physical interaction information and adjusting the local deformation and motion rhythm according to the continuously acquired physical interaction information.

Images