89 results about "Dynamic window approach" patented technology

The invention discloses an intelligent navigation system of a sweeping robot, which relates to the technical field of household intelligent equipment and comprises a multi-source sensor array, a dynamic environment modeling module, a path planning engine, a multi-mode positioning module, an autonomous decision controller, a cloud collaborative optimization unit and a fault degradation processing module. Omnibearing environment perception is provided through a multi-source sensor array, data are fused in real time through a dynamic environment modeling module, a grid map is updated in an incremental mode, obstacle semantic information is marked, an improved A * algorithm and a dynamic window method are fused through a path planning engine, and an optimized path giving consideration to global efficiency and local dynamic obstacle avoidance is generated. The multi-modal positioning module fuses SLAM resolving and IMU data through tight coupling to achieve centimeter-level high-precision and high-robustness positioning, and the autonomous decision controller intelligently and dynamically adjusts the cleaning priority, the advancing speed and the exception handling strategy according to the path, environment state and pose information, so that the cleaning behavior better meets the requirement.

The invention discloses an AUV (Autonomous Underwater Vehicle) path planning and control integrated method for a complex marine environment, and provides an intelligent path control scheme integrating reinforcement learning and local obstacle avoidance mainly aiming at underwater navigation problems such as submarine topography fluctuation, frequent ocean current disturbance and uncertain dynamic obstacle distribution. According to the method, a realistic three-dimensional ocean grid environment is constructed, an improved Q-learning algorithm is utilized to complete global path planning, a group of optimal path point sequences are generated, and it is ensured that the AUV has the macroscopic navigation capability. Local obstacle avoidance is performed in combination with a three-dimensional dynamic window method (3D DWA), and sudden obstacles are dynamically avoided through speed sampling and trajectory prediction. A DDPG control strategy based on deep reinforcement learning is further introduced, end-to-end training of path planning and control execution is realized, and the system can directly output a control instruction according to a real-time sensing state. A cosine attenuation learning rate, a self-adaptive exploration mechanism, an energy consumption penalty function and a dynamic obstacle prediction and control action filtering technology are fused in the method, so that the intelligence of path planning, the stability of control output and the robustness of obstacle avoidance response are effectively improved. According to the invention, the autonomous navigation capability, the environmental adaptability and the safety task completion efficiency of the AUV in a complex marine environment can be obviously enhanced, and reliable intelligent control technical support is provided for underwater operation.

The invention proposes a robot path planning algorithm based on a particle swarm optimization algorithm and a dynamic window method, and relates to the field of control theories and electronic information, and the method comprises the steps: introducing an inertia weight updating strategy based on a state factor, setting adaptive parameters and crossover and mutation operators to improve the global search capability, increase the population diversity, and improve the robot path planning precision. On-line self-adaptive updating of inertia weight and learning factors is realized on line in combination with Q-learning, gene combination modes are enriched through crossover operators, convergence and exploratory performance of the algorithm are improved, an obstacle avoidance strategy of a traditional DWA algorithm is improved, weight parameters of a dynamic window evaluation function are dynamically adjusted according to real-time information of a target and an obstacle, and an obstacle avoidance algorithm is established. According to the method, the global planning is adopted, the path points generated through global planning are adopted as temporary targets, fusion of MOQLCOPSO and the improved DWA algorithm is achieved, local optimum is effectively avoided, the planning efficiency and path safety are improved, and the method is suitable for mobile robot navigation under the complex three-dimensional terrain.

The invention provides an autonomous navigation and intelligent obstacle avoidance method of a biomimetic robot, and relates to the technical field of biomimetic robots, which realizes real-time estimation of six-degree-of-freedom poses by collecting environmental data; visual features are extracted by using a convolutional network, an enhanced memory unit is constructed in combination with pose information, a triple index path memory library is maintained through a dynamic memory updating mechanism, and feature similarity retrieval and navigation instruction generation in a backtracking mode are supported; defining a state vector based on a resultant force of a robot pose, a target direction and an artificial potential field, training a strategy network through deep reinforcement learning, and outputting a motion instruction according to Q value maximization; fusing an artificial potential field and a dynamic window method, and optimizing an objective function to obtain optimal linear velocity and angular velocity; and solving a joint angle through inverse kinematics, and combining attitude error feedback to realize body attitude maintenance and foot end trajectory tracking.

The invention discloses an industrial transfer robot path planning method based on artificial intelligence, relates to the technical field of industrial transfer robot path planning and multi-robot cooperative control, and provides a systematic breakthrough for narrow roadway passage of an industrial transfer robot through deep cooperation of the game theory and federal learning. A robot meeting scene is modeled as a limited action game, and a revenue function is designed in combination with priority weight and security constraint, so that the robot can autonomously generate a forward, yielding or keeping strategy, and bidirectional deadlock is avoided from the source; through a synchronous output mechanism of action probability distribution and a time window, the strategy performability is remarkably improved; the track layer directly generates a synchronization path on the cost graph according to the time window, and motion interruption caused by repeated sampling in a dynamic window method is avoided; even if track deviation or communication abnormity occurs, the rollback strategy can guarantee that the system recovers to be stable immediately, and continuous operation of the production line is maintained.

The invention discloses an anti-impact drilling robot navigation system and method based on multi-mode 3D target detection. The anti-impact drilling robot navigation system comprises a multi-sensor sensing module, a 3D target detection module, a path planning module and a control module. The multi-sensor sensing module is used for acquiring surrounding environment data and the real-time position and posture of the anti-impact drilling robot; the 3D target detection module adopts an improved MVXNet model to identify environmental data; the path planning module is of a double-layer architecture, wherein one layer adopts an improved dung beetle optimization algorithm to carry out global path planning; on the other layer, an improved dynamic window method is adopted for dynamic obstacle avoidance and local planning adjustment of a global path, and finally a corresponding control instruction is generated through a control module to enable the anti-impact drilling robot to advance according to a plan. Therefore, high-robustness, high-adaptability and high-precision navigation driving of the anti-impact drilling robot in a complex environment is realized, and the driving efficiency of the anti-impact drilling robot is effectively improved on the premise of ensuring the advancing safety of the anti-impact drilling robot.

The invention discloses an intelligent luggage case obstacle avoidance method based on an ultra wide band, and relates to the technical field of intelligent mobile equipment control, and the method comprises the steps: collecting and preprocessing multi-source environment perception data, and generating an environment perception fusion data set; correcting projection in the ranging direction and constructing a local environment grid map based on grid division and grid occupation probability; comparing the bottom direction TOF height value with a ground reference height by adopting a threshold value judgment method, identifying a height difference risk area, and generating a path guide point through Cartesian coordinate conversion; based on the path guide points and the local environment grid map, generating a trajectory candidate set by adopting a dynamic window method, and removing impassable trajectories through a collision detection algorithm in combination with obstacle grid information; through a multi-target scoring function and a double-DIP controller, predicting a collision risk score of the feasible trajectory and generating an obstacle avoidance control instruction; the obstacle avoidance control instruction is executed, environment sudden change data are collected in real time, and the obstacle avoidance control instruction is adjusted in real time through a path replanning algorithm.

The invention belongs to the technical field of industrial robot control and detection, and particularly relates to an inspection robot detection system based on image recognition. After the system is started, an image acquisition and preprocessing module synchronously acquires visible light and thermal imaging images of an industrial scene through a multispectral sub-module; the image quality is optimized through a multi-color space adaptive threshold segmentation algorithm; the industrial visual identification module extracts defect features based on a YOLOv-SCSA network, matches a device template and locates defect positions; the motion control module realizes accurate positioning through a multi-sensor dynamic fusion algorithm, plans an obstacle avoidance path in combination with a dynamic window method, and drives the robot to autonomously inspect; the abnormality diagnosis module judges defect types by using a hybrid classification algorithm, evaluates risks according to parameters and triggers early warning; the data storage and transmission module processes data locally and uploads the data to the cloud, the man-machine interaction module displays the data in real time and supports remote intervention, and a closed-loop inspection process is formed.

The invention provides an intelligent inspection robot system, and relates to the technical field of robot inspection. Environment data are collected through a multi-mode sensor; preprocessing the environmental data, including image denoising and enhancement, point cloud filtering and attitude data smoothing, to generate multi-modal fusion input data; inputting the preprocessed data into a Vision Transform-based multi-modal sensing network, extracting global features through a self-attention mechanism, and performing joint embedding with laser radar point cloud and IMU data to generate an environment semantic feature map; based on the environment semantic feature map, utilizing a neural radiation field technology to reconstruct a dynamic 3D scene model in real time, and generating a self-adaptive inspection path through a meta-reinforcement learning algorithm; and the smoothness and safety of the inspection path are optimized by combining a dynamic window method, and a robot motion module is controlled to execute obstacle avoidance and navigation. The technical effect of improving the accuracy and rationality of automatic inspection in the industrial inspection process is achieved.

The invention discloses a plant protection unmanned aerial vehicle operation method and system based on Beidou high-precision positioning and hyperspectral nitrogen nutrition estimation, and the method comprises the steps: determining a plant protection unmanned aerial vehicle operation region, and calculating the flight height of an unmanned aerial vehicle through hyperspectral camera parameters and navigation positioning data; acquiring a hyperspectral image of crops in the operation area of the plant protection unmanned aerial vehicle in real time through a hyperspectral camera, and determining a nitrogen nutrition deficiency area by using a spectral feature selection algorithm based on spectral data in the hyperspectral image; converting the pixel coordinates of the nitrogen nutrition deficiency region into real geographic coordinates; based on the real geographic coordinates of the nitrogen nutrition deficient area, the performance parameters and task requirements of the unmanned aerial vehicle are combined, a dynamic window method is adopted for sowing path planning, and finally accurate sowing irrigation is achieved. According to the method, a dynamic threshold value and maximum entropy threshold value method is introduced, Canny edge detection and multi-source correction are combined, region boundary jitter and area calculation errors are reduced, and a space reference is provided for precise sowing.

The invention discloses an improved unmanned aerial vehicle dynamic window obstacle avoidance method based on an artificial potential field method. The method comprises the following steps: selecting local path points from a global path at certain intervals as sub-target path points for guiding local path planning; carrying out sampling combination on the speed space of the unmanned aerial vehicle in the current state, and generating a plurality of feasible candidate trajectories in a time window; introducing an improved artificial potential field function into an evaluation function of a dynamic window method, and comprehensively evaluating each candidate trajectory; and selecting the trajectory control instruction with the optimal comprehensive evaluation score as the motion input of the next control period, and realizing real-time obstacle avoidance and path correction of the unmanned aerial vehicle in a complex dynamic environment. Obstacle information and the positions of the sub-target path points are updated in real time in the flight process, if the distance between the current position and the positions of the sub-target path points is smaller than a set sub-target path point arrival value, it is regarded that the unmanned aerial vehicle arrives at the sub-target path points, and updating is carried out according to the obtained sub-target path points; and updating the next sub-target path point in the sub-target path point set to realize dynamic path adjustment. The device comprises a processor and a memory.

The invention relates to a hybrid reinforcement learning navigation method and system based on hierarchical course learning and an attention mechanism, and belongs to the technical field of intelligent robots and autonomous navigation. The method comprises the steps of obtaining robot state and environment information through a sensor; extracting context features of the laser radar sequence by using a self-attention network and generating a sub-target candidate set; verifying the track by adopting a dynamic window method; constructing a multi-target reward function and a hierarchical course learning framework; and finally updating the strategy through a mixed loss function. The system correspondingly comprises a sensor module, a feature extraction module, a sub-target generation module, a track verification module, a reward function construction module, a course learning control module and a strategy updating module. The problems that a traditional planner lacks long-term planning and reinforcement learning is low in training efficiency and poor in safety are solved, and the navigation intelligence, safety and learning efficiency of the robot in an unknown complex environment are remarkably improved.

The invention provides an inspection robot obstacle avoidance method based on fusion of an artificial potential field method and a dynamic window method, which belongs to the technical field of robot path planning and comprises the following steps: firstly, respectively constructing a gravitational potential field function and a repulsive potential field function according to the distance from a robot to a target point and the distance from the robot to an obstacle; a traditional repulsive force potential field function is modified into a sampling interval, negative gradients of the gravitation potential field function and the repulsive force potential field function are solved to obtain a resultant force set borne by the robot, an optimal path is selected from multiple paths according to an evaluation function, and finally the cyclic sampling prediction process is repeated until the robot reaches the end point. According to the method, the local minimum value points can be effectively avoided, the path with the least redundant nodes is selected, and the obstacle avoidance efficiency is improved on the premise of ensuring safety.

The invention discloses a two-way A star algorithm and dynamic window method fused path planning method and system. The method comprises the following steps: calculating a global path from a starting point to an end point by using a two-way A star algorithm; traversing the global path to remove redundant nodes in the global path; finding an initial current node in the global path from which the redundant nodes are removed; judging whether the current node is an end point, if the current node is the end point, fusing local paths among the nodes to obtain a final planned path from the starting point to the end point, otherwise, determining a next current node according to the current node and the global path; and calculating a local path from the current node to the next current node by using a dynamic window method, and returning the next current node as a new current node to continue iteration. The invention aims to solve the problems of calculation efficiency and path smoothness of the existing path planning method, improve the navigation capability in a dynamic environment, and improve the safety, movement efficiency and real-time obstacle avoidance capability of the robot.

The invention provides an inland river navigation collision avoidance method and system fusing improved A * and a dynamic window method, and the method comprises the following steps: 1, constructing a two-dimensional grid map of an inland river navigation environment, setting a starting point and an ending point of ship navigation, and determining the distribution information of obstacles in the map; 2, performing global path planning based on an improved A algorithm; step 3, optimizing the global path; 4, performing local planning and fusion based on a dynamic window method DWA; and step 5, repeating the step 4 until the local target point is an end point, and outputting a final navigation path to realize autonomous collision avoidance of the inland ship. According to the technical scheme, navigation safety, efficiency and environmental adaptability are considered.

The invention relates to a self-adaptive safe path planning method for a gas inspection robot, and belongs to the technical field of robot path planning. According to the method, firstly, a global path is generated by improving an A star algorithm, smoothness is optimized, then a multi-sensor data sensing environment is fused, local real-time obstacle avoidance planning is carried out by adopting a dynamic window method, and a deep reinforcement learning model is introduced to realize dynamic decision making. According to the method, the problems that in the prior art, in a dynamic complex environment, path planning is rigid, obstacle avoidance is not timely, and high-risk areas cannot be adaptively avoided are effectively solved, the efficiency and safety of path planning can be remarkably improved, and the autonomous operation capacity and the intelligent decision-making level of the robot in the underground complex environment are enhanced.

The application relates to an unmanned vehicle adaptive path planning method based on a dynamic window method and a proximal strategy. First, an agent-environment interaction model for an unmanned vehicle is constructed, a proximal strategy optimization learning (PPO) model based on an actor-critic framework is established, a reward function is defined according to the principle of a dynamic window method algorithm (DWA) and main evaluation elements, model parameters in the input layer, the output layer size, the number of hidden layers and the number of neurons are determined, and a DWA-PPO deep reinforcement learning model is constructed. Then, the established DWA-PPO deep reinforcement learning model is used for iterative training, and finally, a network model capable of representing the potential relationship between surrounding environment information and evaluation function weight parameters is converged, and the construction of the adaptive PPO-ADWA algorithm is completed. Finally, the feasibility and effectiveness of the unmanned vehicle adaptive path planning strategy based on the PPO-ADWA are verified through simulation comparison experiments.

The invention discloses an indoor dynamic environment modeling and self-adaptive way-finding obstacle avoidance algorithm based on multi-sensor fusion, belongs to the technical field of intelligent navigation, is suitable for indoor autonomous moving scenes such as service robots and intelligent inspection equipment, and solves the problems of weak dynamic environment modeling capability, poor path planning adaptability and low multi-sensor fusion efficiency in the traditional technology. Fusing ultrasonic wave (20Hz), visual sense (30fps) and IMU (100Hz) data, and realizing state estimation through extended Kalman filtering (Q, R matrix definition); constructing a dynamic map by using improved RANSAC (Random Sample Consensus) denoising and an octree (a time decay factor lambda is equal to e-t / 5); path planning and obstacle avoidance are achieved through an improved A * algorithm (heuristic function fusing path length and other three elements) and a dynamic window method (double constraints), the method is used for indoor autonomous navigation and dynamic obstacle avoidance, the actually measured navigation success rate is 92%, the average curvature of the path is reduced by 30%, the positioning error is reduced to 10 cm, and the operation of embedded equipment reaches 28 FPS.

The invention relates to the technical field of tour guide robots, and provides a temple tour guide robot path planning method and system.The temple tour guide robot path planning method comprises the steps that a robot scans environment information to generate three-dimensional point cloud data; global path planning is optimized through a dynamic weight improved A algorithm, a path is predicted in combination with a machine learning model, and an optimal path is obtained; and moving according to the optimal path, and avoiding obstacles in combination with a motion model and a dynamic window method. The machine learning module can enable the robot to avoid a tourist dense area and shorten the moving time, so that the navigation time is shortened; and the path can be adjusted according to weather conditions. By integrating the technologies of path planning, motion prediction, speed optimization, intelligent decision making and the like, autonomous navigation of the robot in a complex dynamic environment is realized. According to the innovative method combining reactive planning and strategic waiting, the robot can flexibly adapt to various dynamic scenes, the service quality is guaranteed, and the practicability is better.

This invention relates to the field of obstacle avoidance technology for mobile robots, specifically a reinforcement learning-based obstacle avoidance method for mobile robots guided by a large language model (LLM). The method includes: using a large language model (LLM) to guide the generation of scene vectors reflecting scene features; and constructing a low-dimensional matrix group conforming to robot dynamics constraints using the Dynamic Window (DWA) method, which together with the scene vectors constitutes the state space for reinforcement learning. This structured, low-dimensional state representation significantly reduces the learning burden of the policy network, enabling the policy to quickly distinguish key states and reduce ineffective exploration in the early stages, thereby reducing training costs. Compared to pure reinforcement learning, which relies on a simulation environment, generating different scene vectors through LLM allows for transfer between different scenes, improving the generalization ability of the method.

The application discloses a real-time path planning method fusing improved locust-dynamic window method, and comprises the following steps: S1, generating an initial path according to an improved A* algorithm, performing global path planning on the path in a radiation environment through an improved locust optimization algorithm, and determining the path from the starting point to the ending point; S2, performing local path planning according to the path through an improved DWA algorithm, and generating a planning path. Through the introduction of a position updating strategy considering radiation distribution information and an improved adaptive comfort zone coefficient, the influence of the radiation risk on path selection is more truly reflected, and the position updating process of the GOA is optimized. This enables the algorithm to effectively guide the locust population to avoid high radiation areas in a nuclear environment, and significantly reduces the radiation exposure risk. Through the combination of the global optimization capability of the improved A* algorithm, the initial population distribution of the locust optimization algorithm is improved, and then the convergence speed of the locust optimization algorithm is improved.

The application provides a hoisting path intelligent planning method, comprising the following steps: S1, crane motion state and environment modeling; S2, construction of a space-time sequence of a historical trajectory of a work equipment; S3, position recommendation based on a space-time attention network (STAN); S4, neural network prediction dynamic window method weight; S5, local path planning based on a weighted DWA algorithm; S6, real-time obstacle avoidance and closed-loop updating. The application mines the time sequence and spatial correlation characteristics in the historical trajectory of the crane through a space-time attention mechanism, and realizes real-time path planning and obstacle avoidance in combination with a dynamic window method, and simultaneously uses a neural network to adaptively adjust the weight of a dynamic window evaluation function, thereby improving the path planning safety, stability and operating efficiency of the crane in a complex outdoor work environment.

The invention provides a robot automatic obstacle avoidance system and method based on a multi-source sensor, and relates to the technical field of mobile robotics.The method comprises the steps that environmental data are collected through the multi-source sensor, obstacle information is obtained, and the environmental data are converted into a local coordinate system of a robot; an improved dynamic window method is adopted to construct a space-time potential field between the robot and an obstacle, a safety channel around the robot is calculated, an adaptive weight factor is introduced to dynamically adjust the safety channel, an obstacle avoidance track of the robot is generated, a gradient descent method is adopted to optimize the speed and the steering angle of the robot, and the obstacle avoidance track of the robot is obtained. And enabling the robot to move along the obstacle avoidance track. Therefore, the problems of insufficient environmental perception precision, weak dynamic prediction capability and low obstacle avoidance decision-making efficiency can be solved.

The invention relates to the field of intelligent driving, in particular to a predictive path planning method suitable for a narrow garage, and the method comprises the steps: S100, collecting the three-dimensional point cloud data and image data of a garage environment through an environment sensing module, marking a static obstacle boundary, a dynamic obstacle boundary and a drivable region, and building an environment model; step S200, pre-generating a collision-free optimal reference path by adopting an A * algorithm optimized by a dynamic window method; when the planning path has an emergent obstacle, an alternative path is quickly generated; and S300, executing the collision-free optimal reference path through the control execution module, and controlling the vehicle to steer and run according to the collision-free optimal reference path and a path generated by environment pre-judgment. According to the invention, the problems of path adjustment lag, lack of pre-judgment, low parking efficiency and insufficient safety caused by dependence on high-precision sensors, environmental factors and the like during path planning of the vehicle in a narrow garage environment are solved.

The invention provides an indoor mobile robot layered dynamic obstacle avoidance method based on a humanoid decision mechanism, which comprises the following steps: step 1, an indoor mobile robot acquires observation and positioning information of an obstacle through a sensor, and constructs a multi-model predictor for prediction; 2, carrying out hierarchical decision-making based on distance driving, and executing a step 3 or a step 4; 3, global path planning: generating a global path according to an improved APF method, fusing obstacle motion state prediction information, controlling the robot to move, and executing the step 1 again; and step 4, local obstacle avoidance control: based on the latest global path, determining the optimal speed by adopting an improved dynamic window method DWA, controlling the robot to move, and executing the step 1 again. The method can effectively improve the path planning efficiency of the robot in a complex dynamic indoor environment and the response ability of the robot to deal with sudden dynamic obstacles, and has the advantages of low calculation amount, high planning efficiency, automatic response and the like.

The invention discloses a real-time path planning method fusing an improved locust-dynamic window method, and the method comprises the following steps: S1, generating an initial path according to an improved A * algorithm, carrying out the global path planning of a path in a radiation environment through introducing an improved locust optimization algorithm, and determining a path from a starting point to an end point; and S2, performing local path planning according to the path through an improved DWA algorithm, and generating a planned path. By introducing a position updating strategy considering radiation distribution information and an improved adaptive comfort zone coefficient, the influence of radiation risk on path selection is reflected more truly, and the GOA position updating process is optimized. The algorithm can effectively guide the locust population to avoid a high-radiation area in a nuclear environment, and the radiation exposure risk is remarkably reduced. The initial population distribution of the locust optimization algorithm is improved by combining and improving the global optimization capability of the A * algorithm, so that the convergence speed of the locust optimization algorithm is improved.

The application discloses a kind of based on patrol trolley robot police training intelligent tracking evaluation method and system, it is related to intelligent monitoring technical field.The method includes: demarcate training area and robot activity area in training site, and patrol trolley robot is based on laser radar and dynamic window method (DWA) in activity area autonomous movement and real-time obstacle avoidance;Through the AI vision module and laser radar carried, the preset training team is accurately identified and locked, and the pan-tilt and the moving chassis are controlled to realize stable follow-up shooting, and training video is collected in real time.Further, based on the multi-angle video shot, using the action recognition algorithm based on skeleton key point detection, the team action specification of training team and the action specification of individual student are intelligently evaluated, and the evaluation report containing quantitative score, deviation positioning and rectification suggestion is automatically output.The application solves the problem that traditional police training evaluation relies on manual, strong subjectivity and low efficiency, realizes the automation tracking of training process, multi-angle recording and intelligent quantitative evaluation.

The application discloses a TD3 map-free navigation method based on a dynamic window method, and particularly relates to the technical field of path planning, and acquires current state information and a target position of a robot; after the current state information and the target position information are preprocessed, the information is input into an improved TD3 network to obtain TD3 output actions, and improved DWA output actions are obtained by combining an improved evaluation function and an escape mechanism; the two obtained actions are input into an action selector to output an optimal action; the robot executes the optimal action to complete a map-free navigation task. The application improves the TD3 network structure, and applies DWA (dynamic window method) and LSTM (long short-term memory network) to map-free navigation. The network structure can directly output robot actions, namely linear velocity and angular velocity, according to the input target point relative position and the robot state, and realizes end-to-end map-free navigation.

The application belongs to the technical field of ocean robot path planning, and particularly relates to an improved dynamic window collision avoidance method for fast approaching of an ocean robot to a target point, a program, equipment and a storage medium. In the process of performing the task of fast approaching of the ocean robot to the target point, real-time state information and obstacle information obtained are used, a dynamic action space window of the ocean robot is obtained through calculation of a dynamic prediction time based on a dynamic window algorithm, and threshold processing is performed on a distance evaluation function in consideration of the problem of excessively moving away from the obstacle in the avoidance process. The optimization processing method of the dynamic prediction time and the threshold distance function is used to solve the problem of speed reduction in approaching the temporary target point and the problem of excessively moving away from the obstacle in the collision avoidance process of the ocean robot, a more effective dangerous avoidance path is solved, the ocean robot is ensured to safely and quickly complete the obstacle avoidance process, and the dangerous avoidance efficiency and the navigation safety of the ocean robot are improved.

The invention discloses a robot navigation path generation system and method, and relates to the field of machine learning, and the method comprises the steps: collecting multi-source data in real time, carrying out the preprocessing, generating a multi-source feature set, carrying out the processing of the multi-source feature set through an attention mechanism, obtaining an environment feature vector, inputting the environment feature vector into an LSTM network, and outputting a dynamic obstacle trajectory prediction result. The method comprises the following steps: constructing a three-dimensional environment model, projecting the three-dimensional environment model into a two-dimensional grid map, processing the two-dimensional grid map by using an A * algorithm to obtain a global path point sequence, generating a local cost map through a clustering algorithm based on the three-dimensional environment model and the current state of the robot, and obtaining a global path point sequence according to the local cost map, the current state of the robot and the global path point sequence. Performing local obstacle avoidance by using a dynamic window method to generate a final executable path; according to the method, the limitation of traditional fixed weight fusion and a simple prediction model is overcome by utilizing the cooperation of the attention mechanism and the LSTM network.