44 results about "Differential game" patented technology

The invention relates to the technical field of control, and provides a process disturbance-oriented bidirectional anti-interference control system and method, and the system comprises the steps: receiving an external disturbance signal and controlled object state monitoring data in parallel through a process sensing module, generating a disturbance feature vector through frequency band decomposition, and processing the state monitoring data through a nonlinear manifold learning algorithm, deep recognition of response features is realized; the disturbance coupling module takes the feedforward control pre-estimated quantity and the feedback control compensation quantity as optimization variables and takes the disturbance suppression ratio and the response recovery time as revenue functions to construct a differential game model, and the revenue functions are optimized through a gradient iterative algorithm; the bidirectional fusion module realizes dynamic weighting of two types of control quantities based on the coupling weight matrix; and the intelligent control module obtains a control deviation integral and response recovery time in real time, and dynamically adjusts parameters of the revenue function and parameters of weighted fusion through the reward function.

The invention discloses a rectangular area unmanned aerial vehicle hunting method based on differential game and reinforcement learning. The our unmanned aerial vehicle cluster obtains environment state information; spatial constraint loss is calculated through a rectangular flight area limitation loss function module, time sequence consistency loss is calculated through a batch time sequence behavior constraint module, and the spatial constraint loss and the time sequence consistency loss are combined to generate a state behavior joint hidden variable fusing space and time sequence constraints. And fusing the hidden variable with the current environment state, inputting the fused hidden variable and the current environment state into a differential game value evaluation module, obtaining a future expected value based on Nash equilibrium by solving a differential game between our part and the black flight unmanned aerial vehicle, and updating decision network parameters based on the future expected value. The cluster performs the optimized hunting action and stores empirical data for continuous training. According to the method, the decision stability, the cooperation efficiency and the task success rate of unmanned aerial vehicle cluster hunting in the rectangular limited flight area are effectively improved.

The present application relates to the technical field of smart water affairs and urban public safety, and particularly relates to a kind of artificial intelligence driven urban pipe network flood resilience evaluation method and system, comprising: constructing Riemann metric tensor by using metric matrix containing potential barrier function, and establishing the Riemann manifold embedding space of pipe network physical state;Then, based on fluid energy, a Hamilton function is constructed, and a dynamics prediction model with physical conservation is obtained by using symplectic neural network and symplectic discrete integral format training;Subsequently, a second-order Hamilton-Jacobi-Aleksandrov partial differential equation describing stochastic differential game is constructed, and a physical perception neural network is used to solve and extract the resilience safety boundary;Finally, real-time data is mapped to the manifold space, the Riemann gradient is calculated, and a quadratic programming problem is solved to generate control instructions. The problem of lack of physical constraints and safety bottom line in pipe network control under extreme random working conditions is solved, and real-time closed-loop control with physical consistency is realized.

The invention discloses a synchronous orbit electric propulsion satellite short-range game control method based on differential game, which comprises the following steps: under an LVLH coordinate system, establishing a near-circular orbit relative motion dynamic model of spacecrafts of both pursuit and escape parties, and further determining a state space form in a spacecraft pursuit and escape game process; based on a state space form in a spacecraft pursuit game process, introducing a Hamiltonian function, and constructing a short-distance pursuit game model with optimal time; introducing a co-state equation, and describing a short-range pursuit game process as a two-point boundary value problem; and converting the problem into a parameter optimization problem, and carrying out optimization solution on the problem by utilizing an improved difference method so as to realize short-range game control.

This invention provides a spacecraft escort control method based on zero-sum differential games, belonging to the field of aerospace technology. The control method includes the following steps: obtaining spacecraft escort problem parameters for both the pursuing and defending satellites; establishing a spacecraft escort model based on zero-sum differential games and inputting initial parameters into the model; transforming the spacecraft escort model into a two-point boundary value problem model; using a genetic algorithm to solve the two-point boundary value problem model to obtain the state trajectory within the game time, thus completing the spacecraft escort control. This spacecraft escort control method can effectively solve the spacecraft escort problem under continuous thrust, and is an effective extension of existing continuous thrust escort problem models and methods. The model employs a differential game method described by a zero-sum cost function, thus simplifying the solution of the difficult-to-solve weighted performance index escort problem and improving the solution efficiency of the spacecraft escort game problem.

The application relates to the field of energy management, and discloses an energy consumption management optimization method and system of a comprehensive energy station, which are used for improving the economic optimization ability and the autonomous collaborative operation level of the comprehensive energy station when facing complex uncertainty. The energy consumption management optimization method of the comprehensive energy station comprises the following steps: constructing a device characteristic model representing long memory dynamics of a device based on a fractional calculus theory; establishing a random scene based on an alpha-stable distribution theory, which describes non-Gaussian fluctuations of wind and light loads; constructing an economic cost density field representing energy quality space-time dissipation based on a local entropy production minimization principle; establishing a differential game framework taking the economic cost field as a target and the above model as a constraint; and finally realizing closed-loop optimization through rolling execution and model online correction. The application breaks away from the dependence on a machine learning model, and realizes the economic efficiency optimization and autonomous collaborative operation of the energy station under complex uncertainty from the physical mechanism.

The invention discloses a differential game collaborative guidance method under data packet loss and time delay, which comprises the following steps of: after a data retransmission mechanism is introduced to establish a mapping relation between a packet loss probability and a time delay size, designing a proportional time delay accumulation error based on a neighbor node signal received by a missile node and a reference signal, and ensuring that the probability of packet loss is reduced in the error conversion mode; when the error converges to zero, the states of all the missiles can achieve cooperation. And respectively designing feed-forward and feedback control components by using the smoothed error in combination with instruction filtering inversion control and a differential game method so as to improve the robustness and the optimal performance of the designed guidance law. Meanwhile, for the problem that the optimal performance index function is difficult to solve in the feedback component design process, an evaluation network fitting performance index function is introduced in combination with an adaptive dynamic programming technology. According to the scheme, the error bounded property can be proved through the designed Lyapunov function, and the stability performance of a missile cluster using the guidance scheme under the condition of communication data packet probability loss and communication time delay is guaranteed.

This application discloses a dynamic emergency decision-making method, device, equipment, and storage medium for urban flood disasters, relating to the field of emergency management and response scheduling technology. The method includes: first, establishing a differential equation for the safety state transition of an urban area, dividing the area into four safety states: normal, affected, recovering, and paralyzed, and defining their transition relationships; based on this, constructing a stochastic differential game model for flood disaster emergency decision-making; solving for the saddle point strategy using a forward-backward scanning algorithm to obtain the optimal emergency prevention and control strategy; then, conducting multi-dimensional simulation verification using typical flood disaster events as samples; finally, adjusting the strategy based on the simulation results to form a target emergency prevention and control strategy for actual emergency decision-making. This application can take into account both spatiotemporal dynamic characteristics and stochastic uncertainties in urban flood disaster emergency decision-making, achieving dynamic adaptation and continuous, efficient decision-making between the disaster system and the emergency response system.

The application discloses a heterogeneous multi-agent multi-target defense decision control method based on differential game. The method can comprise: determining an agent team in a three-dimensional half space, including a defender and an attacker; constructing a multi-target defense game environment model and a heterogeneous multi-agent dynamics model; determining a termination condition and a benefit function of the multi-target attack and defense game; constructing a single-to-single attack and defense confrontation optimal control strategy and a corresponding value function based on the differential game according to the multi-target defense game environment model and the heterogeneous multi-agent dynamics model; and determining multi-agent game scene task allocation according to the benefit function and the value function. The application utilizes the advantages of air-ground cooperation to improve the success rate of target area defense, provides strict theoretical derivation and mathematical proof for the target defense game of air units and ground units, solves the multi-to-multi target defense problem of a large number of agents, and has practicality in solving the actual application of air-ground cooperative combat tasks.

The invention belongs to the technical field of collaborative interception guidance, and particularly relates to a bias differential game collaborative interception method based on region coverage. The method comprises the steps of establishing a probability distribution model of a target in an interception plane, performing Monte Carlo sampling, and converting a coverage strategy design into an optimal coverage problem of a finite center point in the interception plane; segmenting, iteratively solving the standard maneuver of each interceptor through a Voronoi diagram method to obtain the standard maneuver aTsi of each interceptor; and based on the standard maneuvering aTsi of each interceptor, deploying to a corresponding optimal array position and carrying out offset item design of a differential game guidance law to jointly realize the offset of the coverage range of each interceptor, and finally completing the optimal coverage and interception of the target. According to the method provided by the invention, the defects that an existing disc coverage model has limitation on maneuvering distribution simplification and only a fixed solution can be applied mechanically are overcome; the miss distance and the overload demand of the interceptor are obviously reduced, and the interception efficiency is improved.

The invention provides a distributed energy storage and power scheduling method and device based on a master-slave game, and relates to the technical field of distributed energy storage management, and the method comprises the steps: obtaining the energy storage capacity of an energy storage unit in each time period of the whole day, and constructing a power distribution matrix of the energy storage unit according to the divided time periods; a differential game is adopted to set an energy storage dynamic equation, a target revenue function is set according to the energy storage dynamic equation, and a dynamic weight distribution mechanism is adopted to calculate weight parameters of the target revenue function. Double-layer nested time division is adopted to realize dynamic decoupling of power distribution and real-time power; the dimensionality of the optimization problem is reduced through matrix modeling, and the solving speed is increased in combination with dynamic weight distribution; the differential game drives the dynamic equation, depicts the real-time game behavior of the energy storage and the external environment, and achieves the improvement of earnings.

The application discloses a spacecraft escort problem solving method based on LQR and differential game, constructs a game model of the escort problem based on a CW equation, and comprises a cost function, optimization variables, dynamic constraints, acceleration capacity constraints and the like; the cost function adopts a linear quadratic index, the optimization variables are continuous control force accelerations of both attack and defense sides, the dynamic constraints are described by using the CW equation, and the acceleration capacity constraints are modeled by using maximum acceleration limitation; the method has the advantages of simple model, clear physical meaning and convenient solving; in addition, the application converts a complex non-zero-sum game problem into a phased zero-sum game problem and an optimal control problem, can effectively solve the spacecraft escort problem under the action of continuous thrust, is an effective expansion of a continuous thrust escort problem model and method in the prior art, and significantly improves the solving efficiency of the problem. Through the method, an effective cooperative defense threat solution can be provided for on-orbit spacecraft.

The invention discloses a wheel-legged all-terrain chassis longitudinal and transverse movement game optimization method, which comprises the following steps: constructing a wheel-legged all-terrain chassis longitudinal and transverse game optimization framework, the wheel-legged all-terrain chassis longitudinal and transverse game optimization framework comprises an information layer, a distribution optimization layer and an execution layer; in a distribution optimization layer of the game optimization framework, constructing a differential game optimization model, abstracting AFS and DYC into two rational game parties based on a game theory, and finally outputting performance game optimization distribution optimal values of the two parties; and inputting vehicle and road data collected from an information layer into the differential game optimization model, calculating an optimal distribution optimal value of AFS and DYC, sending a control instruction at a distribution optimization layer, realizing the control instruction at an execution layer, and achieving a coupling optimization target of longitudinal and transverse movement of an agricultural machinery chassis. According to the method, through differential game modeling, control quantity distribution is optimized, so that the longitudinal performance of the chassis of the hill agricultural machine is guaranteed, the yawing moment can be generated, and stability control and energy consumption optimization are achieved.

The application discloses a robot path planning method based on deep reinforcement learning, and relates to the fields of intelligent agriculture, path planning, robots and the like. Firstly, a farm environment is perceived, a robot state space and an action space are defined, and a multi-modal weighted combination reward mechanism and an experience replay buffer are set. A learnable weight coefficient is introduced into a Critic network loss function in a traditional DDPG algorithm, and an entropy regularization term is added to a target function of an Actor network. Then, a differential game is generated through an adaptive decay greedy strategy to select a control strategy or a DDPG algorithm to generate an action. Finally, the action or the control strategy is executed, network parameters and target network parameters are updated, and the experience replay buffer is dynamically updated. Compared with other path planning methods, the application improves the adaptability of path planning to a dynamic environment, and also has good efficiency and safety in a complex agricultural environment.

The application discloses a multi-agent differential game control method based on a signal preserving matrix and belongs to the field of multi-agent cluster motion control. The application discloses a multi-agent differential game control method based on a signal preserving matrix, solves the problem that a traditional differential game system dynamic approximate solution construction method loses a feedback signal by introducing a signal preserving matrix, avoids the problem that the traditional differential game system dynamic approximate solution construction method is only applicable to a differential game system composed of 1-relative order subsystems, and obtains an approximate optimal feedback control law of a differential game system composed of 2-relative order subsystems through the signal preserving matrix. The application expands the applicable objects of the differential game system dynamic approximate solution construction method to a large range and has important application value. The application can dynamically obtain an approximate optimal solution of the differential game system and is applicable to cluster motion control of a second-order system such as an unmanned aerial vehicle or a spacecraft.

The application discloses a kind of low-dimensional control methods for radial underactuated spacecraft pursuit-evasion game, comprising the following steps: step S1, based on differential game theory, construct the 24-dimensional two-point boundary value problem of radial underactuated spacecraft optimal time game and high-dimensional game control strategy;Step S2, deduce the state condition of reducing underactuated high-dimensional two-point boundary value to 12-dimensional two-point boundary value problem, and design low-dimensional game optimal control strategy.The application demonstrates the feasibility of underactuated pursuit-evasion game when radial thrust is missing, deduces the low-dimensional optimal time game control strategy for realizing three-dimensional pursuit and evasion only through trace and normal thrust, expands the theoretical analysis and engineering application of underactuated spacecraft, and the method can be used for underactuated spacecraft interception, on-orbit service, approach of out-of-control spacecraft and other applications.

The application discloses a rectangular area unmanned aerial vehicle encircling method based on differential game and reinforcement learning. The environment state information is obtained by the unmanned aerial vehicle cluster; the spatial constraint loss is calculated through the rectangular flight area restriction loss function module, and the time sequence consistency loss is calculated through the batch time sequence behavior constraint module, so that the state behavior joint hidden variable of the fusion space and time sequence constraint is generated. The hidden variable and the current environment state are fused and input into the differential game value evaluation module, the differential game between the unmanned aerial vehicle cluster and the black unmanned aerial vehicle is solved, the future expected value based on Nash equilibrium is obtained, and the decision network parameter is updated. The cluster executes the optimized encircling action, and the experience data is stored for continuous training. The application effectively improves the decision stability, cooperative efficiency and task success rate of the unmanned aerial vehicle cluster in the rectangular limited flight area.

The application discloses an aircraft intelligent evasion guidance method and system based on a generative adversarial imitation learning. Firstly, a three-dimensional aircraft evasion simulation environment is constructed, and evasion demonstration data is generated based on program maneuver or differential game expert strategy. Secondly, a GAIL framework including a strategy network, a value network and a double-branch discriminator is constructed, and the strategy network is pre-trained by using behavior cloning. Then, fusion GAIL training is performed through mechanisms such as progressive GAIL reward introduction, PPO strategy optimization, adaptive supervision decay and EMA strategy smoothing. Finally, the trained strategy model is deployed in an online guidance system, and evasion control instructions are output according to the real-time state of a red aircraft. The GAIL imitation learning framework is adopted, and strategy learning is guided relying on expert demonstration data, so that a complex multi-objective reward function does not need to be artificially designed, and the ability of strategy autonomous optimization and surpassing the performance of experts is reserved.

The application belongs to the technical field of aerospace, and relates to a strategy parameter estimation method for an incomplete information differential countermeasure game. A neural network model is constructed by using a time sequence feature extraction module, a self-attention mechanism and a multi-task learning module in a gated recurrent unit neural network. Input features and output features of the neural network model are determined by using a relative dynamics model and an objective function. Training data is constructed by using original data generated by Monte Carlo simulation. Optimal hyperparameter combinations are determined by using a training loss function and analyzing the influence of hyperparameters on the performance of the neural network model, and the neural network model is trained. In actual game, relative state derivatives and extended state quantities are calculated based on relative state information and self-manipulation control information, the optimal neural network model is input, and the output result is subjected to inverse normalization processing to obtain actual estimation values of target strategy parameters.

The invention discloses a novel power grid dispatching control method and system based on artificial intelligence, and relates to the technical field of power grid dispatching control, and the method comprises the steps: carrying out the topological fusion modeling of power grid multi-source real-time measurement data through employing a space-time diagram convolution network, and obtaining a dynamic digital twinborn body; performing risk field prediction on the dynamic digital twin by adopting a space-time attention mechanism to obtain a time-varying risk distribution result; performing ultra-short-term scheduling analysis on the time-varying risk distribution result by adopting a gating space-time unit to obtain a dynamic weight of a scheduling instruction; performing plan optimization on the dynamic weight of the scheduling instruction by adopting a Transform-OPF hybrid architecture to obtain a power generation plan, and verifying the power generation plan to obtain a scheduling action; updating the scheduling action by adopting a differential game mechanism to obtain an updated strategy parameter; and carrying out dynamic calibration on the updated strategy parameters by adopting Kalman filtering to obtain a digital twinborn parameter correction, and updating the dynamic digital twinborn based on the digital twinborn parameter correction.

The present application relates to the field of control, in particular to a kind of target information missing under the differential game guidance method for maneuvering target.The present application discloses a kind of differential game guidance method for near space high dynamic target scene under considering target incomplete information.The method comprises the following steps: the terminal interception problem is expressed, and the vertical plane is taken as the main research plane, and a two-dimensional missile-target relative motion engagement model is established;A linear extended observer is designed to estimate the target acceleration information;Then, the expression of zero-control miss distance is given, and the linear quadratic differential game (LQDG) guidance law under incomplete target information is obtained.The linear quadratic differential game (LQDG) guidance law under incomplete target information of the present application solves the problem of low control accuracy of interceptor under incomplete target information in the prior art.

This invention proposes a multi-hunter-single escape robustness reconstruction and control method and system based on Set Transformer-HAPPO. The method and system construct a three-dimensional multi-agent arrival-avoidance differential game model with dynamic attrition and replenishment mechanisms. A two-layer network model consisting of a centralized evaluation network and multiple distributed decision networks is designed. The centralized evaluation network introduces Set Transformer to handle the variable number of agent inputs, successfully reducing the computational complexity of traditional attention mechanisms, which increases quadratically with the number of agents, to linear complexity. This makes centralized training of large-scale agent clusters possible with limited computing resources. The distributed decision networks employ a fixed k-neighbor observation mechanism to maintain a constant input dimension, significantly reducing the deployment cost of traditional methods, which increases linearly with the number of application scenarios, to a constant cost requiring only one training iteration. Simultaneously, a two-layer robustness mechanism is formed through adversarial failure training and active mask filtering and dynamic topology reconstruction during the deployment phase, effectively improving the system's failure robustness. The method and system proposed in this invention effectively solve the shortcomings of existing technologies in terms of dynamic scale adaptation, resistance to disabling attacks, cross-scale migration and computational efficiency. It has the advantages of adaptive training scale, strong failure robustness and efficient training deployment, and can achieve stable collaborative capture of a single escapee by multiple pursuers in adversarial environments.

A front wheel-differential cooperative steering control method and system of a multi-axle distributed drive unmanned vehicle, the main steps of which include: obtaining the vehicle state and the desired path; based on the quadratic differential game algorithm of Nash equilibrium theory, calculating the optimal front wheel steering angle and differential steering yaw moment in the path tracking process of the unmanned vehicle to reduce the turning radius and increase flexibility; identifying the stability of the vehicle based on the phase plane of the center of mass side slip angle; using a weight control strategy based on fuzzy rules, the differential steering yaw moment and stability control yaw moment are coordinated and distributed according to the current stability of the vehicle. This method can effectively solve the balanced control of front wheel steering-differential steering and the coordinated control of path tracking performance-stability, and is suitable for multi-axle distributed drive unmanned vehicles that need to consider path tracking accuracy and driving stability.

The invention discloses a dynamic emergency decision-making method, device and equipment for urban flood disasters and a storage medium, and relates to the technical field of emergency management and response scheduling. Dividing the area into four safety states of normal, disaster, recovery and paralysis, and defining a transfer relation thereof; a random differential game model for flood disaster emergency decision making is constructed based on the random differential game model; solving a game saddle point strategy through a forward-backward scanning algorithm, and obtaining an optimal emergency prevention and control strategy; carrying out multi-dimensional simulation verification by taking a typical flood disaster event as a sample; and finally, according to a simulation result adjustment strategy, a target emergency prevention and control strategy is formed and used for actual emergency decision making. According to the method, the spatial-temporal dynamic characteristics and the random uncertainty can be considered in urban flood disaster emergency decision making, and dynamic adaptation and continuous efficient decision making of a disaster system and an emergency response system are achieved.

The invention discloses a fuzzy load frequency control method based on a game theory, and the method is suitable for a modern power system containing an ESS, so as to deal with a frequency stability problem caused by FDI attack and external disturbance. According to the method, firstly, valve position saturation characteristics and energy storage system dynamics are introduced into a load frequency control model, T-S fuzzy modeling is adopted to express a nonlinear system as a convex combination of local linear subsystems, and therefore accurate approximation of the nonlinear characteristics of the system is achieved. And then constructing a time-varying weight FDI attack model from an attacker perspective to describe scaling type and disturbance type measurement signal tampering behaviors, modeling a controller design problem as a double-person zero sum differential game between a power system and an attacker, and solving an optimal control strategy under a minimum-maximum performance criterion. Aiming at a non-convexity problem introduced by game performance constraint, a GA is adopted to optimize fuzzy PI controller parameters, and an optimal gain meeting a robustness condition is obtained through global search.

The application relates to a kind of multi-agent system formation strategy based on hierarchical differential game: first, the communication topology between agents is established by using graph theory, and agent model and obstacle environment model are established;Secondly, for unknown obstacle environment, in the strategy layer, the multi-agent formation problem with known obstacle constraints is converted into distributed differential game, and a cost function is designed for game participants to establish a game model;Using Pontryagin minimum principle, the existence and uniqueness of local Nash equilibrium solution are analyzed, and the expression form of local Nash equilibrium solution is solved, and the global convergence condition of local Nash equilibrium solution is given;In the planning layer, the local Nash equilibrium solution from the strategy layer is modified in real time by using the quadratic programming model based on rolling optimization, and finally a hybrid formation strategy is formed;In theory, it is guaranteed that after the agent successfully avoids the unknown obstacle, the hybrid formation strategy can converge to the local Nash equilibrium solution.

The application provides a dual-arm robot cooperative control method and device based on differential game, relates to the technical field of robot control, and comprises the following steps: collecting multi-modal data of a dual-arm robot, inputting the multi-modal data into a graph neural network to extract observation feature vectors; obtaining a state vector of the dual-arm robot, inputting the observation feature vectors and the state vector into an extended Kalman filter for recursive estimation to obtain a state estimation value of the dual-arm robot; constructing a left-arm control benefit function and a right-arm control benefit function of the dual-arm robot; based on the state estimation value, iteratively calculating a Nash equilibrium point of the left-arm control benefit function and the right-arm control benefit function through an argmin function to obtain left-arm optimal impedance parameters and right-arm optimal impedance parameters corresponding to the Nash equilibrium point; and driving the dual-arm robot to complete a cooperative operation task according to the left-arm optimal impedance parameters and the right-arm optimal impedance parameters.

The invention discloses a hybrid Pareto differential game-based modular robot approximate optimal tracking control method under a cooperative competition framework, and belongs to the technical field of robot control algorithms. The method comprises the steps that a modular robot kinetic model is established, a kinetic equation is constructed through a Newton-Euler iteration method, uncertainty is processed, and a state variable and a disturbance term are generated; constructing a global cost function and a Hamiltonian function based on the state variable and the disturbance term; according to the global cost function and the Hamiltonian function, exporting a control law by using an evaluation neural network and a hybrid Pareto game, and designing a weight updating mechanism; and based on the control law, verifying the control performance through a modular robot platform experiment. The problems that in the prior art, a modular robot is low in control precision and poor in robustness under uncertain disturbance are solved, and the system stability and the energy consumption efficiency are improved.

This application discloses a method and apparatus for multi-to-multi attack and defense game in an obstacle environment, relating to the field of unmanned aerial vehicle (UAV) control technology. The method includes: establishing a multi-to-multi attack and defense problem model for a UAV swarm; establishing a one-to-one attack and defense subgame problem; expanding the defense victory region under differential game based on the one-to-one attack and defense subgame problem to obtain an expanded defense victory region under differential game; expanding the defense victory region under reinforcement learning to obtain an expanded defense victory region under reinforcement learning; and performing multi-to-multi defense decision based on the multi-to-multi attack and defense problem model, the expanded defense victory region under differential game, and the expanded defense victory region under reinforcement learning to obtain a multi-to-multi defense decision planning result. In the attack and defense problem of a UAV swarm with polygonal obstacles, this application enables the defending UAV to intercept as many attacking UAVs as possible, maximizing the protection of the target area.

The invention relates to the technical field of unmanned vehicle control, and discloses a steering control method and device for an unmanned vehicle, and the method comprises the following steps: aligning multi-modal data through a dynamic calibration model, optimizing a fusion weight based on a quality index game, and solving an optimal steering instruction in combination with a vehicle state. Calibration parameters are updated through closed-loop feedback, and high-precision robust control is achieved. The system comprises a multi-source sensor calibration module, a dynamic calibration operation module, a game weight distribution module, a differential game solving module and a closed-loop feedback module. The device comprises a laser radar array, a multispectral camera group, a heterogeneous computing unit, a redundant steering actuator and a synchronous control bus. According to the method, the space-time consistency and fusion robustness of sensing data are improved through a dynamic space-time calibration model and multi-modal game optimization, the path tracking precision is optimized in combination with a differential game equation, self-correction is realized by using closed-loop feedback, and the real-time performance of the system is guaranteed by depending on a heterogeneous computing architecture.