Mountain complex road condition robot dog adaptive cooperative operation control system

Abstract

This invention discloses an adaptive collaborative operation control system for robotic dogs operating in complex mountainous terrain. Relating to the field of collaborative control, the system includes a vehicle-mounted mobile base station, a cluster of robotic dogs, drone collaboration, adaptive collaborative control, and communication and access control modules. The vehicle-mounted base station is equipped with an all-terrain vehicle, constructing a local communication network to provide support even without satellite signals. The robotic dogs are equipped with multiple sensors and achieve terrain adaptation and anti-occlusion tracking through multi-task reinforcement learning. The drones have RTK positioning capabilities and generate 3D terrain models. The adaptive control module implements path planning, dynamic collision avoidance, and task scheduling. The communication module ensures low-latency encrypted transmission, and the access control module implements hierarchical management and operation traceability. The system enables precise collaboration between robotic dogs and drones, adapting to complex mountainous terrain and improving operational efficiency and safety.

Description

Technical Field

[0001] This invention relates to the field of adaptive cooperative operation control systems for robot dogs, specifically an adaptive cooperative operation control system for robot dogs operating in complex mountainous terrain. Background Technology

[0002] In complex scenarios such as emergency rescue, field exploration, mountain inspection, and construction sites, robotic operation systems must cope with multiple challenges, including the lack of satellite signals, rugged terrain, and variable environments. Existing technologies have the following key limitations: communication relies on satellite signals, making it difficult to achieve coordinated scheduling of drones and robotic dogs in signal-free environments; data transmission between devices is interrupted, preventing the formation of closed-loop operations; vehicle-mounted deployment lacks integrated design, resulting in poor compatibility among multiple devices; drone take-off and landing, robotic dog storage, and control module integration are not smooth, leading to low deployment efficiency and difficulty in adapting to the mobility requirements of complex mountainous terrain; multi-robot collaboration mechanisms are imperfect, lacking dynamic priority allocation and terrain-adaptive adjustment strategies, making it difficult for robotic dog swarm motion control and drone inspection tasks to coordinate efficiently, limiting operational efficiency; robotic dogs have insufficient terrain adaptability, and traditional motion control methods are unable to cope with gait instability issues in mixed mountainous terrain, resulting in low efficiency in cross-terrain strategy transfer. Summary of the Invention

[0003] The purpose of this invention is to provide an adaptive collaborative operation control system for a robotic dog in complex mountainous terrain to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an adaptive collaborative operation control system for robotic dogs in complex mountainous terrain, characterized in that it includes a vehicle-mounted mobile base station module, a robotic dog cluster module, a drone collaboration module, an adaptive collaborative control module, a communication module, and an access control module.

[0005] The vehicle-mounted mobile base station module is mounted on an all-terrain vehicle platform. The vehicle platform is equipped with a four-wheel drive system and a drone take-off and landing device. The vehicle-mounted mobile base station module has a built-in privately deployed monitoring and dispatch center, which provides WebSocket, HTTPS and RTMP protocol interfaces to realize device access, task distribution, encrypted data transmission and log management functions, and provides communication support for robot dog clusters and drones in environments without satellite signals.

[0006] The robot dog cluster module includes at least two quadruped robot dogs with terrain adaptation capabilities. Each robot dog is equipped with a depth camera, LiDAR, and a six-axis IMU sensor. It has a built-in motion control unit based on multi-task reinforcement learning and an anti-occlusion perception and tracking unit. The motion control unit generates cross-terrain motion strategies through a Bayesian-optimized self-optimizing terrain course learning strategy. The perception and tracking unit is based on a visible joint weighted fusion framework and a hierarchical memory replay architecture to achieve stable tracking and rapid re-identification of targets in occluded scenarios.

[0007] The drone collaboration module includes at least two drones with RTK positioning capabilities. The drones are equipped with high-definition cameras and lidar, support one-click take-off and landing and autonomous inspection, and can generate a three-dimensional mountain terrain model and transmit terrain data and inspection images to the vehicle-mounted mobile base station module in real time.

[0008] An adaptive cooperative control module, deployed in the monitoring and dispatch center of the vehicle-mounted mobile base station module, includes:

[0009] The collaborative decision-making unit generates global paths for the robot dog cluster and drones based on the generalized Voronoi diagram, adjusts the operation order through a dynamic priority allocation strategy, and adopts a pilot-follower hybrid formation algorithm to achieve the collaborative movement of the robot dog cluster.

[0010] The terrain adaptive unit dynamically adjusts the robot dog's motion parameters and gait strategy based on the 3D terrain model transmitted by the UAV, and combines the reciprocal speed obstacle prediction algorithm to achieve dynamic collision avoidance between robot dogs and with obstacles.

[0011] The task scheduling unit supports parallel processing of multiple tasks and can allocate the robot dog's inspection and data acquisition tasks according to the job requirements. Combined with the three-dimensional digital acceptance function, it can complete the verification of job results.

[0012] The communication module adopts a layered communication protocol and a real-time data pipeline. It constructs a local communication network through the vehicle-mounted mobile base station module to realize low-latency data interaction between the robot dog cluster, drones and the monitoring and dispatch center, and supports data transmission encryption and device authentication.

[0013] The access control module supports hierarchical management of personnel permissions, and implements operation authorization through a dual verification mechanism of password and verification code. It also has an operation history traceability function to ensure the traceability of the operation process.

[0014] Furthermore, the all-terrain vehicle platform of the vehicle-mounted mobile base station module is equipped with reversing radar, tire pressure alarm, traction control system, vehicle stability system and hill descent control function, and adopts AT all-terrain tires and a full-size spare tire; the drone take-off and landing device of the vehicle platform and the drone collaboration module achieve precise docking through mechanical positioning structure and Ethernet interface; the command and control platform in the cabin and the remote control handle of the robot dog cluster module and the drone collaboration module achieve low-latency control signal transmission through wired communication link; the external display screen synchronizes the equipment status and task progress data output by the monitoring and dispatch center of the vehicle-mounted mobile base station module in real time.

[0015] Furthermore, the multi-task reinforcement learning framework of the robot dog's motion control unit includes a policy distillation module. The policy distillation module integrates cross-entropy loss, mean squared error loss, and Dice coefficient to construct a composite objective function, and dynamically adjusts the weight allocation through reinforcement learning. The motion control unit receives terrain adaptation parameters from the adaptive cooperative control module through the communication network of the vehicle-mounted mobile base station module, and adjusts the robot dog's gait strategy in real time to match changes in mountain road conditions.

[0016] Furthermore, the visible joint weighted fusion framework of the robot dog's perception and tracking unit includes a symmetrical joint encoding mechanism for the hip, knee, and ankle joints. The hierarchical memory replay architecture integrates short-term trajectory caching and a long-term online learning feature library. The perception and tracking unit uploads target tracking data to the vehicle-mounted mobile base station module in real time and fuses it with the terrain 3D model transmitted by the UAV collaborative module to provide environmental perception support for the motion control unit.

[0017] Furthermore, the RTK base station of the UAV collaborative module supports simultaneous reception of five satellite frequencies: GPS, GLONASS, BeiDou, Galileo, and QZSS. The mountain 3D terrain model generated by the UAV is transmitted in real time to the monitoring and dispatch center of the vehicle-mounted mobile base station module via the RTMP protocol. After being processed by the adaptive collaborative control module, it is synchronized to the motion control unit and terrain adaptive unit of the robot dog cluster module.

[0018] Furthermore, the collaborative decision-making unit of the adaptive collaborative control module smooths the global path using Bézier curves, and the pilot-follower hybrid formation algorithm realizes formation formation, maintenance, and switching; the collaborative decision-making unit receives terrain data from the UAV collaborative module and real-time pose data from the robot dog cluster module, adjusts the operation order through a dynamic priority allocation strategy, and sends the generated global path instructions to each robot dog and UAV through the WebSocket protocol.

[0019] Furthermore, the reciprocal speed obstacle prediction algorithm of the terrain adaptive unit is based on a BP neural network to construct a speed prediction model. By dynamically adjusting the expansion radius, it generates a set of achievable speeds, thereby realizing distributed cooperative collision avoidance of the robot dog cluster. The terrain adaptive unit calls the three-dimensional terrain model features transmitted by the UAV collaborative module in real time, combines them with the sensor data of the robot dog cluster module, dynamically optimizes the motion parameters of the robot dog, and feeds back the adjustment instructions to the motion control unit of the robot dog.

[0020] Furthermore, the task scheduling unit supports precise planning of inspection routes, real-time transmission of operation data, and three-dimensional digital acceptance functions. It can dynamically adjust the task execution priority according to the complexity of mountainous terrain. After verifying the operator's task execution permissions through the permission management module, the task scheduling unit assigns inspection and data collection tasks to the robot dog cluster module. After receiving the operation data transmitted back by the robot dog and drone, it compares and verifies the data with preset standards through the three-dimensional digital acceptance function and synchronizes the results to the monitoring and scheduling center of the vehicle-mounted mobile base station module.

[0021] Furthermore, the encryption mechanism of the communication module includes symmetric encryption processing of transmitted data and hash verification of stored data, and supports UPS power supply for vehicle charging, mains charging and solar panel charging; the communication module realizes differentiated data transmission between the vehicle mobile base station module, the robot dog cluster module and the drone collaboration module through a layered communication protocol, wherein the control signal adopts a low-latency transmission channel, and the terrain data and image data adopt a high-bandwidth transmission channel, while the device identity authentication mechanism ensures that only authorized devices can access the local area communication network.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] A local communication network is constructed using vehicle-mounted mobile base stations, combined with a private monitoring and dispatch center, enabling stable communication, task distribution, and data transmission between the robot dog swarm and drones in environments without satellite signals, overcoming the signal dependence bottleneck. The all-terrain vehicle platform, equipped with four-wheel drive, all-terrain tires, and a dedicated take-off and landing device, allows for integrated storage and rapid deployment of drones and robot dogs, adapting to the mobility requirements of complex mountainous terrain and significantly improving adaptability to various operational scenarios. The adaptive collaborative control module, through global path planning, dynamic priority allocation, and a hybrid leader-follower formation algorithm, achieves precise collaboration between the robot dog swarm and drones, resulting in more efficient inspection route planning and orderly, conflict-free task execution. The robot dogs employ multiple... The task reinforcement learning framework combines a Bayesian-optimized self-optimizing terrain course learning strategy with a policy distillation technique based on a composite loss function to generate a universal motion strategy across terrains. This results in stable gait in complex terrains such as steep slopes and gullies, significantly improving terrain adaptability. The perception and tracking unit, based on a visible joint weighted fusion framework and a hierarchical memory replay architecture, coupled with an information entropy-spatiotemporal joint evaluation model, can quickly re-identify targets even after partial or complete occlusion, solving the failure problem of traditional tracking systems. It features symmetric encryption for data transmission, storage hash verification, hierarchical management of personnel permissions, and operation history tracing. Combined with a device authentication mechanism, it ensures the security of operational data and operational compliance, meeting industry-level security control requirements. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the control system flow of the present invention;

[0025] Figure 2 This is a schematic diagram of trajectory planning classification according to the present invention;

[0026] Figure 3 This is a framework diagram of the mechanical dog control system of the present invention. Detailed Implementation

[0027] 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.

[0028] Please see Figure 1 —3. The present invention provides an adaptive collaborative operation control system for robotic dogs in complex mountainous road conditions, characterized in that it includes a vehicle-mounted mobile base station module, a robotic dog cluster module, a drone collaboration module, an adaptive collaborative control module, a communication module, and an access control module:

[0029] The vehicle-mounted mobile base station module is mounted on an all-terrain vehicle platform. The vehicle platform is equipped with a four-wheel drive system and a drone take-off and landing device. The vehicle-mounted mobile base station module has a built-in privately deployed monitoring and dispatch center, which provides WebSocket, HTTPS and RTMP protocol interfaces to realize device access, task distribution, encrypted data transmission and log management functions. It provides communication support for robot dog clusters and drones in environments without satellite signals.

[0030] As the core supporting unit of the entire system, the vehicle-mounted mobile base station module's all-terrain vehicle platform is designed with extreme environmental adaptability in complex mountainous terrain in mind. The vehicle platform's four-wheel drive system employs intelligent torque distribution technology, dynamically adjusting the power output to each wheel based on real-time terrain feedback. Combined with the deep tread pattern of the AT all-terrain tires, this effectively enhances its ability to traverse complex terrains such as mud, rocks, and steep slopes. The vehicle's IP54 protection rating withstands dust and rain in mountainous environments, ensuring the stable operation of onboard equipment under harsh conditions.

[0031] The built-in private monitoring and scheduling center adopts a distributed architecture design, supporting concurrent access from multiple devices and parallel task processing. The task distribution logic is dynamically adjusted based on real-time device status and task priority. For example, if a robot dog experiences a task execution delay due to terrain obstacles, the scheduling center will automatically allocate redundant resources or adjust the task order to ensure overall operational efficiency. The provided WebSocket, HTTPS, and RTMP protocol interfaces adopt standardized adaptation schemes, achieving seamless integration with different devices through protocol conversion middleware. It also supports interface expansion and protocol upgrades to meet the needs of subsequent system iterations.

[0032] The data encryption transmission mechanism is implemented throughout the entire data acquisition, transmission, and storage process. During transmission, a symmetric encryption algorithm is used to encrypt sensitive data in real time. During storage, hash verification ensures data integrity, effectively preventing security risks such as communication interception and data tampering that may occur in mountainous environments. The log management module not only records routine information such as equipment operating status and task execution progress, but also retains special logs for critical operations. Log data supports offline storage and online traceability, providing a complete basis for system maintenance and troubleshooting.

[0033] The robot dog cluster module contains at least two quadruped robot dogs with terrain adaptation capabilities. Each robot dog is equipped with a depth camera, LiDAR, and a six-axis IMU sensor. It has a built-in motion control unit and an anti-occlusion perception and tracking unit based on multi-task reinforcement learning. The motion control unit generates cross-terrain motion strategies through a Bayesian-optimized self-optimizing terrain course learning strategy. The perception and tracking unit is based on a visible joint weighted fusion framework and a hierarchical memory replay architecture to achieve stable tracking and rapid re-identification of targets in occluded scenarios.

[0034] Each quadruped robot dog in the swarm employs a multi-sensor fusion perception architecture. Data from depth cameras, LiDAR, and six-axis IMU sensors are processed complementaryly through real-time data fusion algorithms. LiDAR is responsible for long-range terrain obstacle detection and 3D spatial modeling, depth cameras accurately capture close-range target details and environmental texture information, and the six-axis IMU collects robot dog posture data and motion status in real time. The data from these three sources are fused through time synchronization and spatial calibration techniques to form a comprehensive and reliable environmental perception result.

[0035] The multi-task reinforcement learning framework of the motion control unit adopts a two-stage training logic of "professional skill learning - general skill extraction". In the professional skill learning stage, the self-optimizing terrain course learning strategy continuously iterates training samples through Bayesian optimization, dynamically generating training sequences adapted to different terrain features in mountainous areas, enabling the robot dog to gradually master targeted motion skills. In the general skill learning stage, the strategy distillation module integrates and extracts various professional skills through a composite objective function. The reinforcement learning dynamic weight allocation mechanism adjusts the weight ratio according to the real-time feedback of the priority of different terrain tasks, effectively solving the gradient conflict problem of multi-task, and forming a general motion strategy adapted to complex mixed terrain.

[0036] The visible joint weighted fusion framework of the perception and tracking unit addresses the partial occlusion scenarios common in mountainous environments. It extracts and weights the spatial location features of visible joints through a symmetrical joint encoding mechanism for the hip, knee, and ankle joints, enabling accurate target location estimation even when limbs are partially obscured by vegetation or rocks. The hierarchical memory replay architecture includes a short-term trajectory caching module that stores recent target motion trajectory data in real time, while a long-term online learning feature library continuously accumulates target appearance features and movement patterns. Combined with an information entropy-spatiotemporal joint evaluation model, key feature information is dynamically filtered to ensure rapid re-identification based on historical features and environmental context even after complete target occlusion.

[0037] The drone collaboration module includes at least two drones equipped with RTK positioning capabilities. These drones are equipped with high-definition cameras and LiDAR, supporting one-click takeoff and landing and autonomous inspection. They can generate 3D mountain terrain models and transmit terrain data and inspection images to the vehicle-mounted mobile base station module in real time. The RTK base station of the drone collaboration module supports simultaneous reception of five satellite frequencies: GPS, GLONASS, BeiDou, Galileo, and QZSS. The 3D mountain terrain model generated by the drones is transmitted in real time to the monitoring and dispatch center of the vehicle-mounted mobile base station module via the RTMP protocol. After processing by the adaptive collaborative control module, it is synchronized to the motion control unit and terrain adaptation unit of the robot dog cluster module. Through multi-satellite signal fusion calculation, centimeter-level positioning accuracy is achieved in the complex electromagnetic environment of mountains, providing a reliable position benchmark for 3D terrain modeling and precise inspection. The high-definition camera and LiDAR of the drones work in a collaborative mode. The high-definition camera is responsible for collecting terrain texture and target appearance information, while the LiDAR quickly acquires terrain elevation data. The two data are synchronously fused to generate a high-precision 3D mountain terrain model. The autonomous inspection function of the drones supports two operating modes: preset path inspection and dynamic adjustment inspection. Preset-path inspection automatically generates the optimal inspection route based on the user-input work area and task requirements. The route planning fully considers factors such as mountainous terrain undulations and obstacle distribution to ensure comprehensive inspection coverage. Dynamically adjusted inspection relies on real-time reception of commands from the vehicle-mounted mobile base station module and environmental feedback from the robot dog cluster to dynamically adjust the flight trajectory and inspection focus. For example, when the robot dog encounters complex obstacles in a certain area, the drone can specifically enhance terrain detection and data transmission in that area. Key data such as the mountainous 3D terrain model and inspection images are transmitted in real-time to the vehicle-mounted mobile base station module via the RTMP protocol to meet real-time decision-making needs. Simultaneously, the drone has a built-in local storage module to cache and back up transmitted data. In the event of a brief communication signal interruption, data can be automatically retransmitted after the signal is restored to ensure data integrity. The drone's one-click take-off and landing function is equipped with multiple safety mechanisms. Before take-off, it automatically detects the equipment status and flight environment. During landing, it uses dual calibration through visual positioning and RTK positioning to achieve precise landing and adapt to complex take-off and landing conditions in mountainous areas.

[0038] An adaptive cooperative control module, deployed in the monitoring and dispatch center of the vehicle-mounted mobile base station module, includes:

[0039] The collaborative decision-making unit generates global paths for the robot dog cluster and drones based on the generalized Voronoi diagram, adjusts the operation order through a dynamic priority allocation strategy, and adopts a pilot-follower hybrid formation algorithm to achieve the collaborative movement of the robot dog cluster.

[0040] By combining elevation data and obstacle distribution characteristics of mountainous terrain, a global path is generated that maximizes the safety margin between obstacles and the path. After path generation, it is smoothed using Bézier curves to eliminate sharp inflections, ensuring continuous and stable movement trajectories for the robot dog swarm and drones, reducing the difficulty of motion control in complex terrain. A dynamic priority allocation strategy establishes a priority evaluation model based on multiple dimensions such as task urgency, equipment load status, and terrain complexity. For example, urgent inspection tasks have higher priority than regular data collection tasks, robot dogs with lower equipment loads have higher priority than those with saturated loads, and tasks in complex terrain areas have higher priority than those in flat terrain areas. Through this strategy, the collaborative decision-making unit can adjust the task order in real time, ensuring that critical tasks are executed first and improving overall operational efficiency. The navigation-following hybrid formation algorithm is designed for the collaborative movement of the robot dog swarm, supporting three core functions: formation formation, maintenance, and switching. During the formation formation phase, the optimal formation is automatically selected based on task requirements and the number of devices. During the formation maintenance phase, the robot dynamically adjusts its own motion parameters by sensing the position and movement status of adjacent robot dogs in real time to maintain the integrity of the formation. During the formation switching phase, when encountering special terrains such as narrow passages or dense obstacles, the robot automatically switches to a formation that adapts to the terrain and quickly restores the original formation after passing through, ensuring the flexibility and stability of coordinated movement.

[0041] The terrain-adaptive unit dynamically adjusts the robot dog's motion parameters and gait strategy based on the 3D terrain model transmitted by the UAV. Combined with a reciprocal speed obstacle prediction algorithm, it enables dynamic collision avoidance between robot dogs and with obstacles. Through terrain feature extraction algorithms, it identifies key parameters such as terrain type, slope, and obstacle size. Combined with real-time motion status data fed back by the robot dog swarm, it dynamically adjusts the robot dog's motion parameters and gait strategy. For example, on steep slopes, it automatically increases stride length, reduces movement speed, and adjusts the center of gravity distribution to improve stability; on gravelly terrain, it adopts a small stride and high-frequency gait to reduce the impact of uneven ground on movement.

[0042] The reciprocal speed obstacle prediction algorithm is based on a backpropagation (BP) neural network to build a speed prediction model. By learning the historical motion data of the robot dog cluster and the interaction patterns with the terrain, it accurately predicts the movement trajectory and potential collision risks of adjacent robot dogs. The algorithm dynamically adjusts the expansion radius to generate a set of reachable speeds adapted to different terrain conditions. When a robot dog encounters dynamic obstacles or other robot dogs in a mountainous environment, it can quickly select the optimal obstacle avoidance speed from the set of reachable speeds, achieving distributed cooperative collision avoidance and preventing motion stalemates or collision accidents.

[0043] The task scheduling unit supports multi-task parallel processing and can allocate inspection and data acquisition tasks for the robot dogs according to operational needs. Combined with the 3D digital acceptance function, it verifies the operational results. For complex mountainous operations, the overall task is broken down into sub-tasks such as inspection, data acquisition, and 3D modeling. Tasks are assigned according to the performance characteristics and real-time status of each robot dog; for example, inspection tasks in complex terrain areas are assigned to robot dogs with stronger terrain adaptability. The precise inspection route planning function combines the 3D terrain model provided by the UAV with the preset operational range to generate the optimal inspection route. Route planning fully considers factors such as inspection coverage density and movement costs, ensuring that the robot dog's movement distance and energy consumption are minimized while meeting operational requirements. The real-time data transmission function allows the robot dogs to transmit collected terrain data and target information to the vehicle-mounted mobile base station module in real time. The scheduling unit performs real-time analysis and quality assessment of the transmitted data. When data is missing or abnormal, it automatically instructs the robot dog to re-collect data. The 3D digital acceptance function verifies the operational results by comparing them with preset standards. For example, it compares inspection data with historical terrain data to identify terrain changes; it compares collected target information with preset features to verify the accuracy of target recognition. The acceptance results are synchronized in real time to the monitoring and dispatch center of the vehicle-mounted mobile base station module, providing a basis for evaluating the operation effect and adjusting subsequent tasks.

[0044] The communication module employs a layered communication protocol and real-time data pipeline. A local area communication network is constructed via a vehicle-mounted mobile base station module, enabling low-latency data interaction between the robot dog cluster, drones, and the monitoring and dispatch center. It supports encrypted data transmission and device authentication. By optimizing the communication protocol and reducing data redundancy, command transmission latency is controlled within milliseconds to meet real-time control requirements. Large-capacity data such as terrain and imagery data utilizes high-bandwidth transmission channels. Data compression and chunked transmission techniques improve data transmission rates and reduce transmission time. The local area communication network is built upon wireless communication devices based on the vehicle-mounted mobile base station module, employing a multi-channel redundancy design. When a channel is affected by terrain obstruction or electromagnetic interference, it automatically switches to a backup channel to ensure communication continuity. The device authentication mechanism uses a dual authentication method. Each access device is equipped with a unique device identifier, requiring key verification upon network access to prevent unauthorized device access and ensure network security. The power supply supports a combined UPS and vehicle-mounted power supply mode. The UPS uses lithium iron phosphate batteries, featuring high energy density and long cycle life. It supports three charging methods: vehicle-mounted charging, AC charging, and solar panel charging, adapting to diverse power supply needs in mountainous operation scenarios. The UPS has a built-in intelligent battery management system that monitors battery status in real time. When it detects low battery voltage or abnormal temperature, it automatically triggers a protection mechanism and switches to vehicle-mounted power supply to ensure the continuous operation of the communication module and the entire system.

[0045] The access control module supports hierarchical management of personnel permissions, using a dual verification mechanism of password and verification code to authorize operations. It also features operation history tracking to ensure the traceability of the work process. Appropriate operation permissions are assigned based on the operational needs of different departments and the division of responsibilities among personnel roles. The dual password and verification code mechanism provides security for system login; passwords are stored encrypted to prevent leakage, and verification codes are time-sensitive to effectively prevent security risks such as brute-force attacks and stolen logins. The operation history tracking function records all user operations in detail, including key information such as operator, operation time, operation content, and operation result. Operation logs support multi-dimensional retrieval and export, providing complete evidence for security audits and accountability. The system authentication mechanism is implemented throughout the entire operation process; the system verifies the user's permissions in real time when performing any operation, and operations without corresponding permissions will be rejected. Simultaneously, the system supports dynamic permission adjustments; administrators can modify user permissions in real time based on changes in personnel positions and task requirements, ensuring the flexibility and timeliness of access control and adapting to the collaborative needs of personnel in mountainous operation scenarios.

Claims

1. A machine dog adaptive cooperative operation control system for complex mountain road conditions, characterized in that, It includes a vehicle-mounted mobile base station module, a robot dog cluster module, a drone collaboration module, an adaptive collaboration control module, a communication module, and a permission management module. The vehicle-mounted mobile base station module is mounted on an all-terrain vehicle platform. The vehicle platform is equipped with a four-wheel drive system and a drone take-off and landing device. The vehicle-mounted mobile base station module has a built-in privately deployed monitoring and dispatch center, which provides WebSocket, HTTPS and RTMP protocol interfaces to realize device access, task distribution, encrypted data transmission and log management functions, and provides communication support for robot dog clusters and drones in environments without satellite signals. The robot dog cluster module includes at least two quadruped robot dogs with terrain adaptation capabilities. Each robot dog is equipped with a depth camera, LiDAR, and a six-axis IMU sensor. It has a built-in motion control unit based on multi-task reinforcement learning and an anti-occlusion perception and tracking unit. The motion control unit generates cross-terrain motion strategies through a Bayesian-optimized self-optimizing terrain course learning strategy. The perception and tracking unit is based on a visible joint weighted fusion framework and a hierarchical memory replay architecture to achieve stable tracking and rapid re-identification of targets in occluded scenarios. The drone collaboration module includes at least two drones with RTK positioning capabilities. The drones are equipped with high-definition cameras and lidar, support one-click take-off and landing and autonomous inspection, and can generate a three-dimensional mountain terrain model and transmit terrain data and inspection images to the vehicle-mounted mobile base station module in real time. An adaptive cooperative control module, deployed in the monitoring and dispatch center of the vehicle-mounted mobile base station module, includes: The collaborative decision-making unit generates global paths for the robot dog cluster and drones based on the generalized Voronoi diagram, adjusts the operation order through a dynamic priority allocation strategy, and adopts a pilot-follower hybrid formation algorithm to achieve the collaborative movement of the robot dog cluster. The terrain adaptive unit dynamically adjusts the robot dog's motion parameters and gait strategy based on the 3D terrain model transmitted by the UAV, and combines the reciprocal speed obstacle prediction algorithm to achieve dynamic collision avoidance between robot dogs and with obstacles. The task scheduling unit supports parallel processing of multiple tasks and can allocate the robot dog's inspection and data acquisition tasks according to the job requirements. Combined with the three-dimensional digital acceptance function, it can complete the verification of job results. The communication module adopts a layered communication protocol and a real-time data pipeline. It constructs a local communication network through the vehicle-mounted mobile base station module to realize low-latency data interaction between the robot dog cluster, drones and the monitoring and dispatch center, and supports data transmission encryption and device authentication. The access control module supports hierarchical management of personnel permissions, and implements operation authorization through a dual verification mechanism of password and verification code. It also has an operation history traceability function to ensure the traceability of the operation process.

2. The adaptive cooperative operation control system for machine dogs in complex mountainous terrain as described in claim 1, characterized in that, The all-terrain vehicle platform of the vehicle-mounted mobile base station module is equipped with reversing radar, tire pressure alarm, traction control system, vehicle stability system and hill descent control function, and adopts AT all-terrain tires and a full-size spare tire; the UAV take-off and landing device of the vehicle platform and the UAV collaboration module achieve precise docking through mechanical positioning structure and Ethernet interface; the command and control platform in the cabin and the remote control handle of the robot dog cluster module and the UAV collaboration module achieve low-latency control signal transmission through wired communication link; the external display screen synchronizes the equipment status and task progress data output by the monitoring and dispatch center of the vehicle-mounted mobile base station module in real time.

3. The adaptive cooperative operation control system for machine dogs in complex mountainous terrain as described in claim 1, characterized in that, The multi-task reinforcement learning framework of the robot dog's motion control unit includes a policy distillation module. The policy distillation module integrates cross-entropy loss, mean squared error loss, and Dice coefficient to construct a composite objective function, and dynamically adjusts the weight allocation through reinforcement learning. The motion control unit receives terrain adaptation parameters from the adaptive cooperative control module through the communication network of the vehicle-mounted mobile base station module, and adjusts the robot dog's gait strategy in real time to match changes in mountain road conditions.

4. The adaptive collaborative operation control system for a robot dog in complex mountainous terrain as described in claim 1, characterized in that, The visible joint weighted fusion framework of the robot dog's perception and tracking unit includes a symmetrical joint encoding mechanism for the hip, knee, and ankle joints, and the hierarchical memory replay architecture integrates short-term trajectory caching and long-term online learning feature library. The perception and tracking unit uploads target tracking data to the vehicle-mounted mobile base station module in real time, and merges it with the terrain 3D model transmitted by the UAV collaborative module to provide environmental perception support for the motion control unit.

5. The adaptive collaborative operation control system for a robot dog in complex mountainous terrain as described in claim 1, characterized in that, The RTK base station of the UAV collaboration module supports simultaneous reception of five satellite frequencies: GPS, GLONASS, BeiDou, Galileo, and QZSS. The mountain 3D terrain model generated by the UAV is transmitted in real time to the monitoring and dispatch center of the vehicle-mounted mobile base station module via the RTMP protocol. After being processed by the adaptive collaborative control module, it is synchronized to the motion control unit and terrain adaptive unit of the robot dog cluster module.

6. The adaptive cooperative operation control system for a robot dog in complex mountainous terrain as described in claim 1, characterized in that, The adaptive collaborative control module's collaborative decision-making unit smooths the global path using Bézier curves, and the pilot-follower hybrid formation algorithm enables formation formation, maintenance, and switching. The collaborative decision-making unit receives terrain data from the UAV collaborative module and real-time pose data from the robot dog cluster module, adjusts the operation order through a dynamic priority allocation strategy, and sends the generated global path instructions to each robot dog and UAV via the WebSocket protocol.

7. The adaptive cooperative operation control system for a robot dog in complex mountainous terrain as described in claim 1, characterized in that, The terrain adaptive unit's reciprocal speed obstacle prediction algorithm is based on a BP neural network to build a speed prediction model. It generates a set of achievable speeds by dynamically adjusting the expansion radius, thereby achieving distributed collaborative collision avoidance of the robot dog cluster. The terrain adaptive unit calls the 3D terrain model features transmitted by the UAV collaborative module in real time, combines them with the sensor data of the robot dog cluster module, dynamically optimizes the robot dog's motion parameters, and feeds back the adjustment instructions to the robot dog's motion control unit.

8. The adaptive collaborative operation control system for a robot dog in complex mountainous terrain as described in claim 1, characterized in that, The task scheduling unit supports precise planning of inspection routes, real-time transmission of operation data, and three-dimensional digital acceptance functions. It can dynamically adjust the task execution priority according to the complexity of mountainous terrain. After verifying the operator's task execution permissions through the permission management module, the task scheduling unit assigns inspection and data collection tasks to the robot dog cluster module. After receiving the operation data transmitted back by the robot dog and drone, it compares and verifies the data with preset standards through the three-dimensional digital acceptance function and synchronizes the results to the monitoring and scheduling center of the vehicle-mounted mobile base station module.

9. The adaptive collaborative operation control system for a robot dog in complex mountainous terrain as described in claim 1, characterized in that, The encryption mechanism of the communication module includes symmetric encryption processing of transmitted data and hash verification of stored data, and supports UPS power supply for vehicle charging, mains charging and solar panel charging; the communication module realizes differentiated data transmission between the vehicle mobile base station module, the robot dog cluster module and the drone collaboration module through a layered communication protocol, wherein the control signal adopts a low-latency transmission channel, and the terrain data and image data adopt a high-bandwidth transmission channel, and at the same time, the device identity authentication mechanism ensures that only authorized devices can access the local area communication network.

Images