A safety-first mobile robot control method and system
By using scLTL and FSARS algorithms in the robot control system, we simulate and predict the risk of staff movement and plan the safety path, and solve the safety and efficiency problems of robot multi-task control in the dynamic environment in the existing technology, achieving safety-first multi-task control Task job path planning.
Patent Information
- Application Number
- CN202410364872.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-28
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-03-28
AI Technical Summary
The existing robot multi-task control system fails to effectively consider the uncertainty of random movement of workers, making it difficult to plan safe paths in a dynamic environment, with high computing costs and poor dynamic adaptability in the environment, making it impossible to ensure the safety and efficiency of robot operations.
Using a finite state automatic mechanism based on scLTL to build a logical standardized task model, predict the risk of staff movement through simulation, design cost functions, and use the FSARS search algorithm to plan security paths to ensure the safety and efficiency of robots performing multi-task operations in a dynamic environment. .
It realizes the optimal path for robots that are safety-first in a dynamic environment, avoids conflicts with staff in advance, improves the safety and efficiency of robot multi-task operations, reduces computing costs, and enhances environmental adaptability.
Smart Images

Figure CN118567345B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of robot control, and in particular to a control method for a mobile robot under risk conditions. Background Art
[0002] With the advancement of artificial intelligence technology, the demand for robots to perform multiple tasks is also increasing. However, the control technology of robots in performing multiple tasks still needs to be improved, especially in the dynamic environment of reality. The uncertain factors in the environment bring many risks to the robot's multi-task performance, and put forward higher requirements for the robot's multi-task control and safe path planning. At present, with the continuous application of robots in logistics warehousing and factories, the demand for dynamic operation of robots is constantly increasing, especially when robots and workers work in the same environment. However, due to the dynamic change characteristics of the movement of workers, the robot poses a threat to the safety of workers in the process of performing tasks, and also affects the robot's own operation safety and efficiency. The existing robot multi-task control system does not consider the uncertainty of random movement of workers to deal with dynamic risks in the environment, and provide safe control strategies and path planning for robot multi-task operations.
[0003] Existing research focuses on real-time obstacle avoidance and automatic navigation of robots. However, for robots in logistics warehouses and factories, simply realizing real-time obstacle avoidance and automatic navigation cannot meet the needs. First, the real-time obstacle avoidance algorithm will cause the robot to constantly avoid obstacles and fail to complete the task within the specified time, resulting in task failure and unable to guarantee the efficiency of logistics warehouses and factories. Second, it is necessary to constantly plan the path in real time according to the environment, which has a high computational cost and is not practical in the same environment as logistics warehouses and factory workers. At present, this issue is still in the research stage, and there is no theoretical algorithm at home and abroad to provide safe control strategies and path planning for multi-task operations of robots.
[0004] The Chinese patent with the patent number 2023112051910 and the name of it is a mobile robot control method and system in a dynamic environment. It still has shortcomings such as high computing cost, poor large-environment learning effect, and the inability to guarantee the safety and optimality of the planned path. In addition, it has poor dynamic adaptability to the environment. When the environment mutates and the pre-planned path becomes invalid, it takes a long time to re-plan, making it difficult to apply in actual scenarios.
[0005] In view of the above problems, the present invention proposes a safety-first mobile robot control method and develops a mobile control system for a mobile robot to perform multi-task operations in an uncertain environment taking environmental risks into consideration. Summary of the invention
[0006] The present invention provides a safety-first mobile robot control method and system, which aims to consider the risks brought by uncertain factors in the environment, provide an optimal control strategy for the robot to perform multi-task operations, and plan a safe operation path.
[0007] The technical solution to achieve the purpose of the present invention is:
[0008] A safety-first mobile robot control method comprises the following steps:
[0009] Build a logical specification task model based on scLTL, write all tasks into the logical specification task model, and build a finite state automaton of the logical specification task model;
[0010] Observe the status of the staff at time t0, simulate and predict the movement of the staff according to the status of the staff at time t0, and calculate the simulation prediction data to predict the risks caused by the movement of the staff;
[0011] The cost function is designed considering the predicted risk. In the process of searching the environment, a path planning algorithm for single-task operation is obtained based on the principle of cost-optimal priority.
[0012] Based on the finite state automaton, the FSARS search algorithm is constructed. By calling the path planning algorithm, the safe path of the robot's multi-task operation is solved based on the safety priority search strategy.
[0013] Furthermore, the logical specification task model is:
[0014]
[0015] Where T represents the Boolean operator true; prp is the set of atomic propositions; represents an atomic proposition; X represents the next step; express The next state is true; F stands for future; express The future is real; U: express True until satisfy.
[0016] Furthermore, the finite state automaton is:
[0017]
[0018] Where Q is a finite state set, 2 PrP is a finite set of atomic propositions, δ is Q×2 PrP →Q state transition function, q0 is the initial state, q F is a finite set of accepting states.
[0019] Further, the risks posed by the movement of staff are:
[0020]
[0021] Where R(v, t) represents the risk value caused by the movement of K workers at position v. is the simulated prediction value, which represents the conditional probability that worker k occupies v at time t, t0<t≤t0+T, where T is the prediction time period.
[0022] Furthermore, the simulation prediction value Specifically:
[0023]
[0024] Among them, N v is the number of visits to position v at time t, N s is the number of simulations.
[0025] Furthermore, the cost function is:
[0026]
[0027] Among them, α, β are balance parameters, γ is the scale parameter, g(v) is the cost function from the starting position to the current position v, h(v) is the cost estimation function from position v to the target position, and z(v) is the risk cost function of the current state, that is:
[0028] z(v)=V r (v)+E r (v)+F r (v)
[0029] Among them, V r (v) is the collision risk of the robot colliding with the person at position v, E r (v) is the risk of the robot colliding with a person when it moves from the current position v to the next position v′, F r (v) is the future risk corresponding to position v.
[0030] Furthermore, the V r (v) E r (v) F r (v) specifically:
[0031] V r (v) = R(v, t)
[0032] E r (v) = R (v, t + 1) * R (v', t)
[0033]
[0034] Among them, R(v, t) is the collision risk corresponding to v at time t, R(v, t+1) is the collision risk corresponding to time t+1, R(v′, t) is the collision risk corresponding to v′ at time t, v′ is the position at time t+1, and neighbors represents the set of neighbors of position v.
[0035] Furthermore, in the FSARS search algorithm, each node n is defined as nq, ns, n.path, n.value, and np, where nq represents a finite state automaton. The status value of is used to control the progress of the task, ns represents a target point of the robot, n.path represents the path from the parent node to the current node, which is determined by the path planning algorithm, n.value represents the value of the current node, which is used to evaluate the safety of the node, and np represents the parent node of n.
[0036] Furthermore, the FSARS search algorithm selects the node with the smallest n.value to expand based on the safety-first search principle. If a node n is expanded for the first time and its successor nodes n′ have not been visited, a random strategy is adopted. A successor node is randomly selected in the process, and the update of n.value is:
[0037] n′.value=n′.p.value+∑ v∈n′.path (V r (v)+E r (v)).
[0038] A system for implementing the safety-first mobile robot control method comprises a finite state automaton construction unit, a risk calculation unit, a path planning unit and an FSARS search unit. The finite state automaton construction unit constructs a logical specification task model based on scLTL, writes all tasks into the logical specification task model, and constructs a finite state automaton of the logical specification task model; the risk calculation unit observes the state of the staff at time t0, simulates and predicts the movement of the staff according to the state of the staff at time t0, and statistically simulates and predicts the risk caused by the movement of the staff; the path planning unit designs a cost function considering the predicted risk, and in the process of searching the environment, obtains a path planning algorithm for single-task operations based on the principle of optimal cost priority; the FSARS search unit constructs an FSARS search algorithm based on a finite state automaton, and solves the safe path of the robot's multi-task operation based on a safety-first search strategy by calling the path planning algorithm.
[0039] Compared with the existing mobile robot control system, the double-layer finite state automaton risk search method considering environmental risks proposed in the present invention has the following beneficial effects:
[0040] (1) The present invention proposes a safe-A* method that considers the dynamic factors in the environment and plans a safe path for the robot in dynamic and complex scenes, so that the robot can avoid conflicts with workers in the environment in advance, replacing the existing real-time obstacle avoidance algorithm, thereby ensuring the safety of workers and the safety of robot operations;
[0041] (2) The present invention proposes a safety-first search method for multi-tasking operations of robots. The method aims to determine the optimal safe path when multi-tasking, provide a safe path from a global perspective of multi-tasking execution, and improve the efficiency of multi-tasking operations of robots. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 This is a control system architecture diagram of the present invention.
[0043] Figure 2 Randomly move the model graph for the workers.
[0044] Figure 3 It is a risk probability distribution diagram.
[0045] Figure 4 This is a risk distribution example diagram.
[0046] Figure 5 This is the upper layer search process diagram of FSARS.
[0047] Figure 6 The figure is a flow chart of the method of the present invention.
[0048] Figure 7 Schematic diagram of the robot's starting position, task settings, and human observation position in the embodiment.
[0049] Figure 8 It is a state transition diagram in the embodiment. DETAILED DESCRIPTION
[0050] The present invention considers the uncertain risks in the environment and proposes a two-layer finite state automaton (FSA) risk search method considering environmental risks, the FSARS algorithm. The lower layer proposes a safe-A* safety path planning method, which allows the robot to generate an optimal path with safety priority in a dynamic risk environment; the upper layer is a multi-task safe path search based on the principle of safety priority, which solves the conflict between the robot and the staff at the planning level and provides safe decision-making and path planning for the robot's multi-task operation. Considering the scenario where the robot and the staff work in the same environment, s is the state set of the robot, s tis the state of the robot at time t, s0 is the initial state of the robot, s g ={g1, ..., g I} is the target point set corresponding to task I; H is the state set of the staff, is the state at time t; A = {forward, backward, left, right, wait} is the set of actions of the robot and the staff. The present invention plans a safe path for the robot at the planning level to avoid conflicts with the staff and ensure the safety of the staff and the robot.
[0051] The proposed control system architecture is shown in the attached Figure 1 As shown in the figure, the control system consists of the following 4 parts:
[0052] S1: Multi-task encoder (finite state automaton building block);
[0053] S2: Risk calculation unit;
[0054] S3: Lower layer safety path planning (path planning unit);
[0055] S4: Upper level FSARS search algorithm (FSARS search unit).
[0056] Among them, S3 and S4 are the core technologies of the present invention.
[0057] The technical principles of each unit of the present invention are as follows:
[0058] S1: Multi-task encoding controller
[0059] This project mainly uses scLTL logic specifications to control the robot's multi-tasks, writes multiple tasks into a formula, and performs unified control. The technical principles are as follows:
[0060] First, we transform the multi-task into atomic propositions and use formula (1) to organize and express them.
[0061]
[0062] in,
[0063] T represents the Boolean operator "true";
[0064] prp is an atomic proposition;
[0065] Represents atomic propositions;
[0066] X(next) indicates the next step;
[0067] Xφ means that the next state of φ is true;
[0068] F (future) means future;
[0069] Fφ=TUφ means that φ is true in the future;
[0070] U(until): This means that φ1 is true until φ2 is satisfied.
[0071] Through the above logical expression, the atomic propositions of all tasks can be written into formula (1) for unified control and management.
[0072] The scLTL formula (1) Convert to finite state automaton FSA
[0073]
[0074] Where, Q: finite state set, 2 PrP : a finite set of atomic propositions, δ: Q×2 PrP →Q state transition function, q0: initial state, q F : A finite set of accepting states. Through a finite state automaton FSA state transfer.
[0075] S2: Risk Calculation
[0076] The present invention mainly considers that humans and robots work in the same environment, and the movement of humans depends on their own will and is not affected by the control system. The present invention considers a situation where risk calculation is the most difficult, that is, the movement of humans adopts a completely random strategy. Figure 2 As shown, for each action a (a∈A={forward,backward,left,right,wait}) of a person, the probability P a is unknown, where P a ∈{P f , P b , P r , P l , P w}, P f Represents the probability of a person moving forward, P b represents the probability of a person moving backwards, P r represents the probability of a person moving to the right, P l represents the probability of a person moving to the left, P w It represents the probability that a person waits in place without moving.
[0077] In the present invention, since P a Since it is unknown, a random strategy is adopted, that is, the actions taken by a person each time he moves are randomly selected.
[0078] In logistics warehouses, factories and other scenarios, the environment is usually defined as a 2D grid world, where each grid v = (x, y) represents a state (position). There are K workers in the environment. For any worker k, is the observed state at time t0, where Indicates that grid v is occupied by worker k at time t0. For any worker k, according to the observed state Perform a random simulation to predict the trajectory of worker k and calculate the conditional probability that worker k occupies v at time t (t0<t≤t0+T). The calculation formula is as follows,
[0079]
[0080] Among them, N v is the number of visits to position v at time t, N s is the number of simulations.
[0081] For robots, the risk distribution caused by the movement of workers is complex, and the probability distribution is as follows: Figure 3 As shown, it contains multiple peaks, which is a problem that traditional Gaussian distribution cannot solve. In the present invention, we calculate the risk value of v caused by the movement of K workers, R(v, t), and the calculation formula is as follows:
[0082]
[0083] Through the above calculations, we can get the risk set from time t0 to t0+T, Thus, it is easy to t Or the risk of getting v at any time t.
[0084] Figure 4 An example is shown: a scene of a person in a 7×7 grid world. The person starts at position (3, 6). We show the risk of each v from time t=0 to t=6. ξ1, ξ2, ..., ξ6 are Figure 4 (a), 4(b), ... and 4(f).
[0085] S3: Lower layer safety path planning
[0086] Input: the robot's starting point s0, the target point g, and the observed positions of the K staff members
[0087] Output: path χ = {s0, ..., g}
[0088] To plan safe paths in risky environments, we propose a safe-A* method that uses the principle of best-first search and considers the risks caused by workers in the environment. This method evaluates each state v through formula (5) to perform a best-first search.
[0089] f(v)=g(v)+h(v)+z(v)……Formula (5)
[0090] Among them, f(v) is the cost function of position v, each state v = (x, y) represents the position coordinates, g(v) is the cost function from the starting position to the current position v, h(v) is the cost estimation function from v to the target position, and z(v) is the risk cost function of the current state. The risk cost z(v) of position v depends not only on the current position v, but also on future risks. The detailed calculation formula of z(v) is
[0091] z(v)=V r (v)+E r (v)+F r (v) ……Formula (6)
[0092] Where V r (v) is the collision risk of the robot colliding with the person at position v, E r (v) is the risk of collision between the robot and the human when the robot moves from the current position v to the next position v′ (at time t+1), F r (v) is the future risk corresponding to state v. r (v) and E r (v) can be calculated by formula (7)(8).
[0093] V r (v) = R (v, t) ... Formula (7)
[0094] E r (v)=R(v,t+1)*R(v′,t)...Formula (8)
[0095] Among them, R(v, t) is the collision risk corresponding to v at time t, R(v, t+1) is the collision risk corresponding to time t+1, and R(v′, t) is the collision risk corresponding to v′ at time t.
[0096] Usually, given a certain path for the robot, the future risk F corresponding to v at time t is r (v), can be calculated theoretically by formula (9).
[0097] F r (v) = V r (v′)+E r(v′)+V r (v″)+E r (v″)+…Formula (9)
[0098] Where v′ is the position at time t+1, and v″ is the position at time t+2.
[0099] However, the key to the problem is that the robot's path is unknown and needs to be planned. In other words, formula (9) cannot be used to calculate future risks in reality. Therefore, we adopt an approximate method to calculate future risks. The specific calculation method is defined by formula (10):
[0100]
[0101] Through formula (10), we approximate the difficult problem. Although we cannot obtain the definite path of the robot, it is easy to obtain the neighbors of the current position v. Therefore, we use the sum of the risks of all neighbors to approximate the future risk.
[0102] So far, although we can evaluate the value of the current position v by formula (5), as shown in formula (5), g(v) and h(v) are distances, and z(v) is the risk value, which is usually much smaller than g(v) and h(v). Therefore, we need to set a balance parameter to balance g(v), h(v) and z(v).
[0103]
[0104] Among them, α, β are balance parameters, and γ is a scale parameter.
[0105] Finally, the value of each v is calculated by formula (11). In the process of searching the environment, the principle of best priority is used to generate a safe path for the robot. The safe-A* method proposed in this patent inherits the best priority principle of classic A*, fully considers the risks brought by the movement of workers in the environment, and can definitely be solved by safe-A*, and ensures that the first path solved is the optimal solution.
[0106] S4: upper layer FSARS (FSA-Risk-Search) search algorithm, S3 is the lower layer of S4, embedded in S4, S4 updates the node information in S4 by continuously calling S3 to generate paths
[0107] Input: robot's starting point s0, target points g1, g2, ...g I
[0108] Output: Path x = {s0, ..., g1, ..., g i}
[0109] The upper layer FSARS is a graph search based on FSA, which plans the optimal safe path for the robot to perform multiple tasks. According to the robot's task, S1 can easily get based on We construct the FSARS search architecture, where each node n contains:
[0110] nq: a The status value is used to control the progress of the task.
[0111] ns: a target point of the robot,
[0112] n.path: the path from the parent node to the current node,
[0113] n.value: the value of the current node, used to evaluate the security of the node.
[0114] np: the parent node of n.
[0115] Each edge represents a path x from a node to its successor node.
[0116] FSARS search process Figure 5 As shown. The search graph starts from the root node n0, as Figure 5 (Left) As shown, for the root node n0,
[0117] n0.q=q0(q0 is initial state),
[0118] n0.s=s0 is the initial state of the robot,
[0119] Indicates that the root node is empty.
[0120] n0.value=0,
[0121] Indicates that it has no parent node.
[0122] The entire search process of FSARS can be divided into three stages: selection, expansion, and path construction, as follows:
[0123] Selection: This stage determines which node to select for expansion. In the present invention, FSARS is based on the principle of security-first search, so FSARS always selects the node with the minimum value; the smaller the n.value value, the safer the node. If a node is expanded for the first time, that is, none of its successor nodes have been visited, this algorithm adopts a random strategy and randomly selects a successor node. If all the successor nodes of the node have been visited, the node is called a fully expanded node.
[0124] Extension: Each selected node n is based on The state transfer is extended, that is, the successor node of node n can be expanded from We select a successor node with the smallest value as the successor node n′ to expand node n, which is the parent node of n′.
[0125] For the successor node n′,
[0126] n′.q=q′, where q′ comes from
[0127] ns=g i ,
[0128] n′.path=χ is planned by the lower layer safe-A*,
[0129] n′.value consists of two parts, one inherited from the parent node and the other determined by n′.path. Formula (12) shows how n′.value is calculated.
[0130] np=n.
[0131]
[0132] Figure 5 (Left) shows an example of expanding the root node n0 to the node n1, which is an expansion process with no successor nodes visited.
[0133] Note that a node in FSARS can be visited multiple times. Figure 5 (middle) shows that from node n m Before starting the expansion, the red node has been visited. The existing expansion starts from node n2, where
[0134] nq=q,
[0135] ns=g i ,
[0136] n.path=χ,
[0137] n.value←Formula (12),
[0138] np=n2.
[0139] The red edges and nodes represent the nodes from node n m The new expansion started,
[0140]
[0141]
[0142]
[0143]
[0144]
[0145] like Figure 5 As shown in (middle), one expansion starts from node n2, and the other expansion starts from node n m Starting from, except that nq = n′.q, the two extensions are very different, as follows:
[0146] The target point is different, Although they share the same q-state,
[0147] Different paths, Because the target points of the two expansions are different,
[0148] The most important thing is that the values of the nodes brought by the two expansions may be different.
[0149] In FSARS, the smaller the n.value is, the safer the node is. That is, the new extension is safer than the previous extension, so we replace the previous information with the new extension.
[0150] nq=q,
[0151]
[0152]
[0153]
[0154]
[0155] if We keep the previous node information.
[0156] Path generation: The FSARS proposed in the present invention performs a security-first search. The search process ends at the target node (if a node nq = q F , then this node is called the target node. The target node represents that all tasks have been completed). After the search process is completed, FSARS starts from the target node and traverses backwards until it reaches the root node, and generates the robot operation path. Figure 5 (Right) As shown, the constructed robot safe operation path is shown as blue nodes and edges.
[0157] The pseudo code of the FSARS algorithm is as follows:
[0158]
[0159]
[0160] Based on the system, the method proposed by the present invention is as follows: Figure 6 As shown, the steps include:
[0161] Step 1: Initialize various parameters;
[0162] Step 2: Multi-task settings;
[0163] Step 3: Use scLTL to make multi-task atomic propositions and build the corresponding finite state automaton FSA;
[0164] Step 4: Observe environmental information;
[0165] Step 5: Calculate the risk based on the information observed in step 4 (staff location, etc.);
[0166] Step 6: FSARS performs safe path planning by calling the safe-A* path planning algorithm;
[0167] Step 7: The robot performs multi-task operations. If the operation is interrupted during the operation, it returns to step 4 to recalculate the risk and plan the path;
[0168] Step 8: The work is completed.
[0169] The present invention adopts syntactically co-safe linear temporal logic (scLTL) logic specification for multi-task control, considers the risks brought by random movement of workers in dynamic environment, and plans safe paths for robots. The focus of the present invention is to avoid risks in the environment at the planning level, generate the optimal safe path for mobile robots to perform multi-task operations in the same environment with people, and avoid conflicts with workers in advance. The core of the present invention is to propose a two-layer finite state automaton (FSA) risk search method, FSARS, that considers environmental risks. The lower layer proposes a safe-A* safe path planning method that allows robots to generate optimal paths with safety priority in dynamic risk environments, rather than the shortest path. At the same time, the upper layer is the search architecture of FSARS based on the principle of safety priority. The optimal safe path is planned for robots to perform multi-tasks to avoid collisions with randomly moving workers. The proposed algorithm considers the uncertain risks caused by the movement of workers in the environment, thereby avoiding risks in advance at the planning level, and providing the optimal safe path for mobile robots to achieve multi-task operations in uncertain environments.
[0170] Example
[0171] by Figure 7 For example, the robot's starting position is s0 = (1, 0), and the task target point set is s g ={g1,g2,g3}={(5,5),(1,5),(3,1)}, assuming that there are K=1 workers who can move randomly in the environment, at t=0, the position of the person is observed Based on the observation, risk simulation and calculation can be performed to obtain the risk set ξ, such as Figure 4 As shown. Write the robot's I = 3 tasks into the scLTL formula and construct like Figure 8 shown.
[0172] The FSARS algorithm is used for path planning. After 100 simulations, the average value is calculated to evaluate and verify the proposed algorithm. The simulation results are shown in Table 1.
[0173] Table 1 Evaluation and verification simulation results
[0174]
[0175] The average number of conflicts of the proposed algorithm is significantly lower than that of the RL, FSA-MDP and MCTS algorithms, and the job success rate of the proposed algorithm is improved to 0.9, which is better than that of the RL, FSA-MDP and MCTS algorithms.
[0176] The present invention is a mobile robot control method with safety priority, and proposes a mobile control system for a mobile robot to perform multi-task operations in an uncertain environment taking into account environmental risks. The system adopts syntactically co-safe linear temporal logic (scLTL) logic specifications for multi-task control, considers the risks of uncertain factors in a dynamic environment, performs safe path planning, avoids risks in the environment at the planning level, generates an optimal safe path for the mobile robot to perform multi-task operations in the same environment as humans, and avoids conflicts with workers in advance. The core of the present invention is to propose a two-layer finite state automaton (FSA) risk search method considering environmental risks, FSA-Risk-search, FSARS. The lower layer proposes a safe-A* safe path planning method, which allows the robot to generate an optimal path with safety priority in a dynamic risk environment, rather than the shortest path. At the same time, the upper layer is a search architecture of FSARS with safety priority as the principle. The optimal safe path is planned for the robot to perform multi-tasks to avoid collisions with randomly moving workers. Note that the present invention is not aimed at real-time collision avoidance. Compared with other MDP and MCTS algorithms, the FSARS method proposed in the present invention has significantly improved safety. The present invention can be applied to robot multi-tasking operations, mainly including factory automation, logistics warehousing robots and humans working in the same environment.
Claims
1. A safety-first mobile robot control method, It is characterized in that Includes steps: Build a logical specification task model based on scLTL, write all tasks into the logical specification task model, and build a finite state automaton of the logical specification task model; Observation t 0 The status of the staff at the moment, according to t 0 The staff's status at all times, simulate and predict the staff's movement, compile statistical simulation and prediction data, and predict the risks caused by the staff's movement; The cost function is designed considering the predicted risk. In the process of searching the environment, a path planning algorithm for single-task operation is obtained based on the principle of cost-optimal priority. Based on the finite state automaton, the FSARS search algorithm is constructed. By calling the path planning algorithm, the safe path of the robot's multi-task operation is solved based on the safety priority search strategy.
2. A safety-first mobile robot control method according to claim 1, It is characterized in that The logical specification task model is: Where T represents the Boolean operator true; prp is the set of atomic propositions; represents an atomic proposition; X represents the next step; express The next state is true; F means future; express The future is real; U: express True until satisfy.
3. A safety-first mobile robot control method according to claim 2, It is characterized in that The finite state automaton is: Where Q is a finite state set, 2 PrP is a finite set of atomic propositions, δ is Q×2 PrP →Q state transition function, q 0 is the initial state, q F is a finite set of accepting states.
4. A safety-first mobile robot control method according to claim 1, It is characterized in that The risks posed by the movement of workers are: Among them, R(v, t) represents the risk value caused by the movement of K workers at position v at time t. is the simulated prediction value, which represents the conditional probability that worker k occupies v at time t, t 0 <t≤t 0 +T, T is the prediction time period.
5. A safety-first mobile robot control method according to claim 4, It is characterized in that The simulation predicted value Specifically: Among them, N v is the number of visits to position v at time t, N s is the number of simulations.
6. A safety-first mobile robot control method according to claim 4, It is characterized in that The cost function is: Among them, α, β are balance parameters, γ is the scale parameter, g(v) is the cost function from the starting position to the current position v, h(v) is the cost estimation function from position v to the target position, and z(v) is the risk cost function of the current state, that is: with(v)=V r (in)+E r (v)+F r (in) Among them, V r (v) is the collision risk of the robot colliding with the person at position v, E r (v) is the risk of the robot colliding with a person when it moves from the current position v to the next position v′, F r (v) is the future risk corresponding to position v.
7. A safety-first mobile robot control method according to claim 6, It is characterized in that The V r (v) E r (v) F r (v) specifically: V r (v)=R(v,t) E r (v)=R(v,t+1)*R(v′,t) Among them, R(v, t) is the risk value caused by the movement of K staff members at position v at time t, R(v, t+1) is the risk value caused by the movement of K staff members at time t+1, R(v′, t) is the risk value caused by the movement of K staff members at position v′ at time t, v′ is the position at time t+1, and neighbors represents the set of neighbors of position v.
8. A safety-first mobile robot control method according to claim 7, It is characterized in that In the FSARS search algorithm, each node n is defined as nq, ns, n.path, n.value, and np. nq represents a finite state automaton. The status value of is used to control the progress of the task, ns represents a target point of the robot, n.path represents the path from the parent node to the current node, which is determined by the path planning algorithm, n.value represents the value of the current node, which is used to evaluate the safety of the node, and np represents the parent node of n.
9. A safety-first mobile robot control method according to claim 8, It is characterized in that The FSARS search algorithm selects the node with the smallest n.value to expand based on the safety-first search principle. If a node n is expanded for the first time and its successor nodes n′ have not been visited, a random strategy is adopted. A successor node is randomly selected from the set, and the successor node n′.value is: n′.value=n′.p.value+∑ v∈n′.path (V r (v)+E r (v)) The successor node with the smallest value is used as the successor node n′ to expand node n, and n is the parent node of n′.
10. A system for implementing the safety-first mobile robot control method according to any one of claims 1 to 9, It is characterized in that It includes a finite state automaton construction unit, a risk calculation unit, a path planning unit and an FSARS search unit. The finite state automaton construction unit constructs a logical specification task model based on scLTL, writes all tasks into the logical specification task model, and constructs a finite state automaton of the logical specification task model; the risk calculation unit observes t 0 The status of the staff at the moment, according to t 0 The state of the staff at all times is used to simulate and predict the movement of the staff, and the simulation prediction data is statistically simulated to predict the risks brought by the movement of the staff; the path planning unit designs a cost function considering the predicted risks, and in the process of searching the environment, based on the principle of optimal cost priority, obtains a path planning algorithm for single-task operations; the FSARS search unit constructs an FSARS search algorithm based on a finite state automaton, and by calling the path planning algorithm, solves the safe path for the robot's multi-task operation based on a safety-first search strategy.
