TWI821001B - System and method for generating global optimal exercise strategy - Google Patents

Abstract

A method for generating a global optimal exercise strategy includes the following steps: obtaining environment information from an environment server; obtaining a physiological state by a physiological sensor and obtaining an equipment state by an equipment sensor when an user performs the first segment of the exercise with the exercise equipment; determining a plurality of exercise strategies; establishing an exercise strategy topology including a plurality of nodes, wherein each node represents one of the plurality of exercise strategies; selecting one of a plurality of sub-nodes in a first layer of the exercise strategy topology according to a plurality of exercise performance prediction values, and outputting the exercise strategy represented by the selected sub-node as an optimal exercise strategy of the next segment, wherein each exercise performance prediction value is obtained by inputting the exercise strategy represented by one of the plurality of nodes, the environmental information, the physiological state and the equipment state represented into an exercise performance model.

Description

System and method for generating global optimal motion strategy

The present invention relates to motion and strategy analysis, particularly a system and method for generating a global optimal motion strategy.

With the rise of health awareness, cultivating exercise habits to stay healthy has become a life goal pursued by modern people. Take the common sport "cycling" as an example: in recent years, the bicycle market has developed vigorously and expanded year by year. The current training method for moving from beginner to professional driver not only strengthens the driver's leg strength and turning speed, but also includes: measuring the driver's power output and becoming familiar with the power output of various gears. The former includes measuring critical power (CP), functional threshold power (FTP) and upper limit of rotational speed. The latter requires the driver to be familiar with the power output under each gear ratio and how to choose the appropriate gear ratio at each gradient.

However, both CP and FTP require one to several hours of measurement and require professional equipment and analysts. Shifting strategies, on the other hand, require a lot of actual riding to establish. Even professional drivers mostly choose shifting strategies based on personal experience. Currently, there is no judgment mechanism that can confirm that the shifting strategy selected by the driver is the best for the entire route.

In short, for beginners, the difficulty in challenging long-distance riding is: not knowing how to choose the gear ratio (gear) and cadence that suits their current riding conditions. So far, there is still no conclusion on the optimal gear shifting strategy. Professional cyclists can only match their own heart rate and power performance based on experience. Adjust gear ratio and cadence. From being familiar with gear ratios to shifting strategies, it takes a long time to accumulate experience and is not something that can be accomplished overnight.

In view of this, the present invention proposes a system and method for generating a global optimal motion strategy.

A method for generating a global optimal exercise strategy according to an embodiment of the present invention is suitable for users who perform exercise with exercise equipment. The motion has a plurality of segments that are continuous in time, and these segments include a first segment and at least a second segment. The method includes the following steps performed by the host: obtaining a plurality of environment information corresponding to a plurality of segments from an environment server. When the user performs the first segment of exercise with the exercise equipment, the physiological state is obtained from the physiological sensor provided on the user, and the equipment status is obtained from the equipment sensor provided on the exercise equipment. A plurality of exercise strategies are determined, which at least correspond to a plurality of setting values of a setting of the exercise equipment. A motion strategy topology including multiple nodes is established, where each node represents a motion strategy, the branching degree of each node is associated with the number of motion strategies, and the height of the motion strategy topology is associated with the number of segments. Select one from a plurality of sub-nodes in the first layer of the sports strategy topology according to the plurality of sports performance prediction values, and output the sports strategy represented by the sub-node as the best sports strategy of at least one second segment, wherein each The sports performance prediction value is obtained by inputting the sports strategy, environmental information, physiological state and equipment status represented by a node into a sports performance model.

A system for generating a global optimal exercise strategy according to an embodiment of the present invention is suitable for users who perform exercise with exercise equipment. The motion has a plurality of segments that are continuous in time, and these segments include a first segment and at least a second segment. The system includes: an environment server, a physiological sensor, an equipment sensor and a host. Environment server storage corresponds to multiple segments multiple environmental information. The physiological sensor is disposed on the user and obtains the user's physiological state when the user performs the first segment of exercise with the exercise equipment. The equipment sensor is disposed on the sports equipment and obtains the equipment status of the sports equipment when the user performs the first segment of exercise with the sports equipment. The host communication is connected to the environment server, physiological sensors and equipment sensors. The host is configured to execute a plurality of instructions to output at least a second segmented optimal exercise strategy. The instructions include: determining a plurality of exercise strategies, which at least correspond to multiple setting values of a setting of the exercise equipment. A motion strategy topology including multiple nodes is established, where each node represents a motion strategy, the branching degree of each node is associated with the number of motion strategies, and the height of the motion strategy topology is associated with the number of segments. Select one from a plurality of sub-nodes in the first layer of the sports strategy topology according to the plurality of sports performance prediction values, and output the sports strategy represented by the sub-node as the best sports strategy of at least one second segment, wherein each The sports performance prediction value is obtained by inputting the sports strategy, environmental information, physiological state and equipment status represented by a node into a sports performance model.

To sum up, the system and method for generating a global optimal motion strategy proposed by the present invention can save users who are not yet proficient in sports the time of professional measurements, and users can take measures without accumulating a large amount of sports experience. Optimal motion strategies for global motion.

The above description of the present disclosure and the following description of the embodiments are used to demonstrate and explain the spirit and principles of the present invention, and to provide further explanation of the patent application scope of the present invention.

100,200: A system that generates a globally optimal motion strategy

10:Host

12:Notification interface

14: Setting adjustment device

20:Equipment sensor

30:Physiological sensor

40:Environment server

S1~S5, S51~S56, S511~S513: steps

R: root node

N1~N3: child nodes

L1~L3: leaf nodes

P: Visit path

G1, G2, G3: movement strategy

FIG. 1A is a block diagram of a system for generating a globally optimal motion strategy according to an embodiment of the present invention; Figure 1B is a block diagram of a system for generating a globally optimal motion strategy according to another embodiment of the present invention; Figure 2 is a flow chart of a method for generating a globally optimal motion strategy according to an embodiment of the present invention; Figure 3 is a motion strategy A schematic diagram of the topology; Figure 4 is a detailed flow chart of the steps in Figure 2; Figure 5 is a detailed flow chart of the update procedure; and Figures 6A to 6D are examples of finding the best movement strategy in the movement strategy topology.

The detailed features and characteristics of the present invention are described in detail below in the implementation mode. The content is sufficient to enable anyone familiar with the relevant art to understand the technical content of the present invention and implement it accordingly. Based on the content disclosed in this specification, the patent scope and the drawings, , anyone familiar with the relevant arts can easily understand the relevant concepts and features of the present invention. The following examples further illustrate the aspects of the present invention in detail, but do not limit the scope of the present invention in any way.

The system and method for generating a global optimal exercise strategy proposed by the present invention are suitable for users who perform exercise with exercise equipment. The movement has a plurality of segments that are continuous in time. For example: riding a bicycle on a route, the route is divided into multiple road segments (for example, based on slope or altitude), and each road segment corresponds to a segment. For example: use fitness equipment to perform multiple sets of training, each set of training corresponds to a segment. Please note that the "best" mentioned in the following refers to the best in the whole world (multiple motion segments).

FIG. 1A is a block diagram of a system for generating a globally optimal motion strategy according to an embodiment of the present invention. The system 100 for generating a global optimal exercise strategy includes a host 10 , an equipment sensor 20 , a physiological sensor 30 and an environment server 40 .

The host 10 is used to execute multiple instructions. These instructions can be comprehensively evaluated based on the user's physiological state, the equipment state of the sports equipment, and the environmental information of the sports environment, and provide the user with an optimal exercise strategy to achieve a sports goal. For example: when the exercise is riding a bicycle, the exercise goal is the shortest riding time. The multiple instructions are described in detail later.

In one embodiment, the host 10 may be implemented using one or more of the following examples: a central processor unit (CPU), a microcontroller (MCU), an application processor (Application Processor), AP), field programmable gate array (FPGA), Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), system-on -a-chip (SOC), deep learning accelerator (deep learning accelerator), or any electronic device used to execute the plurality of instructions.

The equipment sensor 20 is provided on the sports equipment and is connected to the host 10 through communication. When the user performs the first segment of the exercise with the exercise equipment, the equipment sensor 20 obtains the equipment status of the exercise equipment and transmits it to the host 10 .

The physiological sensor 30 is installed on the user and is communicatively connected to the host 10 . When the user performs the first segment of exercise with the exercise equipment, the physiological sensor obtains the user's physiological state.

The environment server 40 stores a plurality of environment information corresponding to a plurality of segments, and is communicatively connected to the host 10 .

In one embodiment, the sports equipment is a bicycle, the sport is cycling, and the multiple segments included in the sport correspond to multiple riding sections. The multiple environmental information includes: corresponding to At least one of distance, slope, temperature, average elevation, and elevation difference for each ride segment. The physiological state includes: at least one of the user's pre-riding heart rate and the user's post-riding heart rate. The equipment status includes: at least one of vehicle speed, rotational speed and pedal power.

FIG. 1B is a block diagram of a system for generating a globally optimal motion strategy according to an embodiment of the present invention. Compared with the embodiment of FIG. 1A , the system 200 shown in FIG. 1B further includes a notification interface 12 and a setting adjustment device 14 .

The host 10, the notification interface 12, the setting adjustment device 14 and the equipment sensor 20 are all provided on the sports equipment.

The notification interface 12 is communicatively connected to the host 10 . The notification interface 12 is used to notify the user of the optimal exercise strategy output by the host 10 . In one embodiment, the notification interface 12 is a screen, a speaker, or any human-machine interface that can be used to notify the user.

The setting adjustment device 14 is communicatively connected to the host computer 10 . The setting adjustment device 14 is used to adjust the settings of the sports equipment. In one embodiment, the setting adjustment device 14 is a shift lever of an electronic transmission, which can adjust settings of the bicycle, such as gears or gear ratios. In one embodiment, the user manually adjusts the settings of the exercise equipment through the setting adjustment device 14 according to the optimal exercise strategy. In another embodiment, the setting adjustment device 14 automatically adjusts the settings of the sports equipment according to the optimal exercise strategy output by the host 10 .

FIG. 2 is a flowchart of a method for generating a globally optimal motion strategy according to an embodiment of the present invention. In one embodiment, the aforementioned system 100 or 200 may be used to perform the method shown in FIG. 2 .

In step S1, the host obtains a plurality of environment information corresponding to a plurality of segments from the environment server. Taking cycling as an example, the environmental information that the host can obtain includes a route package The number of road sections included, the length, slope, temperature, average altitude and altitude difference of each road section.

In step S2, when the user performs the first segment of the exercise with the exercise equipment, the user's physiological state is obtained from the physiological sensor, and the equipment state of the exercise equipment is obtained from the equipment sensor. In one embodiment, after the user rides the first road section in a random fixed gear, the system will obtain: physiological status such as heart rate before riding, heart rate after riding; and equipment status such as vehicle speed, rotational speed, and pedal power. . These riding data from the first segment will be used in subsequent steps to assess the user's individual abilities.

In one embodiment of the present invention, the system can automatically detect which segment of the exercise the user performs." In one embodiment, the host obtains the current location information and altitude information through the GPS locator and altitude detector installed on the bicycle, and then compares the above information with the environmental information to determine where the user is currently located. A road section. In other embodiments, the user can specify the segment to which the current motion belongs. The present invention does not limit the way the system learns motion segments.

In step S3, the host determines a plurality of exercise strategies, which at least correspond to a plurality of setting values of the exercise equipment. In one embodiment, the motion strategy is a combination of rotational speed and gear position. For example: the two rotational speeds that the user can achieve include R1 and R2, and the three gears that the bicycle can adjust include G1, G2 and G3. Therefore, the host can decide 6 movement strategies, including (R1,G1), (R1,G2), (R1,G3), (R2,G1), (R2,G2) and (R2,G3). In another embodiment, the exercise strategy is a combination of heart rate zones and gears. The heart rate zones include: the lower end of the aerobic endurance zone (62.5% MHR), the upper end of the aerobic endurance zone (72.5% MHR), the aerobic power zone (80% MHR), the maximum efficiency zone (90% MHR), and the speed burst zone (97.5 % MHR) and other five intervals, where MHR represents the maximum heart rate.

In step S4, the host establishes a motion strategy topology including multiple nodes. In one embodiment, the motion strategy topology may adopt a tree structure. The motion strategy trees described below are all embodiments that adopt a tree structure corresponding to the motion strategy topology. In one embodiment of the motion strategy tree, each node represents a motion strategy, the branching degree of each node is associated with the number of motion strategies, and the height of the motion strategy tree is associated with the number of segments.

Taking bicycle riding as an example, Figure 3 is a schematic diagram of the motion strategy tree. Assume that the riding route includes four sections, the user has completed the first section, and the remaining three sections need to decide their respective movement strategies. Therefore, the height of the motion policy tree established in step S4 is 3. It is assumed that three movement strategies have been determined in step S3, corresponding to the three gears G1, G2 and G3 respectively. Therefore, in Figure 3, each node represents one of the three movement strategies G1, G2, and G3. Except for leaf nodes, each node has three branches, representing three movement strategies that can be selected for the next road section. Starting from the root node R of the motion policy tree, along the edges connecting the nodes, you can visit the leaf nodes of the motion policy tree. Taking the visiting path P in Figure 3 as an example, it represents that the movement strategy G2 is selected for the second section, the movement strategy G1 is selected for the third section, and the movement strategy G3 is selected for the fourth section.

In step S5, the host selects one of multiple sub-nodes in the first level of the sports strategy tree based on multiple sports performance prediction values, and outputs the sports strategy represented by this sub-node as the best sports strategy for the next segment. Each sports performance prediction value is obtained by inputting the sports strategy represented by a node, multiple environmental information, physiological status and equipment status into the sports performance model. The details of the sports performance model are first explained below, and then the details of selecting the optimal sports strategy are explained. In addition, in an embodiment of step S5, the algorithm used by the host when selecting child nodes is a Monte Carlo tree search algorithm.

In one embodiment, the sports performance model runs on the host. In another embodiment, the sports performance model runs on an external server, and the host communicates with the external server to transmit a plurality of sports performance evaluation parameters to the sports performance model, and receives the sports output of the sports performance model from the external server. Performance predictions.

Taking riding a bicycle as an example, the plurality of sports performance evaluation parameters include: the riding record of the first road section (environmental information, physiological state and equipment status of the first road section), environmental information of at least a second road section (which may include More sections after the second section), the expected movement strategy (corresponding to a node in the movement strategy tree), and the accumulated physical energy consumption from the starting point to the present (can be calculated from the aforementioned information). The sports performance prediction value is the estimated riding time.

In one embodiment, the sports performance model is a feed forward neural network trained based on a sports history. The exercise history includes multiple pieces of data, and each piece of data includes a timestamp, reference environmental information, reference physiological status, and reference equipment status. Therefore, based on the timestamps of the two pieces of data, the interval between the two pieces of data can be calculated. Before training the sports performance model, one or more pieces of data with high similarity in the exercise process can be combined into a single piece of data for training. The sports performance prediction value output by the sports performance model is the estimated time to complete a segment of the movement.

As can be seen from the above, using the sports performance model can estimate the estimated riding time of the specified sports strategy on the specified road section. On the other hand, according to the definition of motion policy tree, if the branch degree is B and the number of segments is K, then the number of leaf nodes is B ^K . The number of leaf nodes represents the difficulty of searching for the best strategy in the motion strategy tree. Taking cycling as an example, a route may need to be divided into dozens of segments in order to reflect the environmental changes of each segment. As a result, expanding all movement strategies for all road segments will result in a huge movement strategy tree, and it becomes impractical to calculate the estimated riding time of all nodes. Therefore, one embodiment of the present invention proposes an optimal solution search method in step S5, whereby the optimal motion strategy is selected in the motion strategy tree.

Fig. 4 is a detailed flow chart of step S5 in Fig. 2, including steps S51 to S56. FIG. 5 is a detailed flow chart of the update procedure mentioned in step S51 and step S52. Figures 6A to 6D are an example of finding the best movement strategy in the movement strategy tree. Each step in FIG. 4 and FIG. 5 is explained below with reference to the examples of FIG. 6A to FIG. 6D .

In step S51, the host performs a single random visit operation on each child node and performs an update procedure. The child nodes are nodes on the first level of the motion strategy tree, such as child nodes N1, N2 and N3 in Figure 6A. Numbers in nodes represent performance estimates. The random visit operation starts from the child node, along the edge between nodes, randomly selects a node in the next layer to visit until reaching the leaf node. As shown in Figure 6A, starting from each child node, a single random visit operation is performed until the leaf node L1/L2/L3.

The update program includes the process shown in Figure 5. Specifically, in step S511, the host accumulates the number of visits to each child node. As shown in Figure 6A, the number of visits to sub-node N1 to sub-node N3 is all once, and is recorded as N=1.

In step S512, the host accumulates the total number of visits to the motion policy tree. The total number of visits is the sum of the number of visits to all child nodes. As shown in Figure 6A, the total number of visits t=3 is marked on the root node R.

In step S513, the host calculates the evaluation parameters of each child node based on the multiple sports performance prediction values, the number of visits, and the total number of visits of multiple visited nodes passed by a single random visit operation. One embodiment of the present invention proposes a dedicated Upper Confidence Bound (UCB1) formula as an evaluation parameter, as shown below:

。 Among them, UCB 1 is the evaluation parameter, W is the weighted sum of the time consumption of all known subtree paths, n is the number of visits to the child node, t is the total number of visits, ln() is the natural logarithm, α and β are weight constants. In one embodiment, α =20,

.

In Figure 6A, since step S513 is executed for the first time, each child node has only one visit path. Therefore, W of child node N1 =58+29+53=140; W of child node N2 =50+25+38=113; W of child node N3 =37+11+16=64. The evaluation parameters of each child node are calculated according to the above UCB1 formula, as shown in Figure 6A.

Please note that although the optimal solution search method proposed in step S5 by one embodiment of the present invention applies the steps of Monte Carlo Tree Search (MCTS) traditionally used in chess games, there are still some implementation details. The differences are as follows. Difference 1: There is no factor of changing hands. In a chess game, both parties alternate moves, and the best solution search method proposed in one embodiment of the present invention has only one user in the entire process. Therefore, the optimal solution search method proposed by an embodiment of the present invention does not need to consider the position of the opponent's moves like the traditional method. During the playout process, there is no need to consider the reverse situation value caused by alternating moves. Instead, it directly uses random The selection is executed to the end, as shown in step S51 and step S52. Difference 2: The Upper Confidence Bound (UCB1) formula is different. In MCTS, which is traditionally used for chess games, the value of a move is usually calculated based on the number of wins and losses in the route. An embodiment of the present invention proposes The optimal solution search method does not have the concept of victory or defeat, so it cannot use any UCB1 formula with the concept of victory or defeat. Instead, it uses the UCB1 formula dedicated to an embodiment of the present invention proposed in step S513.

In step S52, among multiple child nodes, the host selects the child node with the largest evaluation parameter to perform a single random visit operation and perform an update procedure. As shown in Figure 6B, the evaluation parameter 1.5135 of the child node N3 is greater than the evaluation parameter 1.5 of the child node N2 and the evaluation parameter 1.4965 of the child node N1. Therefore, the host selects sub-node N3 for a single random visit operation, generates another visit path (37-44-59), and updates the total number of visits ( t =4) and the number of visits of sub-node N3 ( n =2) . Please note that after the update program is executed, the evaluation parameter of the selected child node N3 drops to 1.2264, which is lower than the evaluation parameters of the other two child nodes N1 and N2. Therefore, the next time step S52 is executed, the host will select a child node other than child node N3, that is, explore other possible best solutions.

In step S53, the host determines whether there is a candidate child node among all child nodes whose visit count is not less than the first threshold. If it is determined to be "yes", continue to step S54. If the determination is "No", then return to step S52. In one embodiment, the first threshold is 100, and the maximum number of visits to a child node in Figure 6B is only 2. Therefore, step S52 will be executed repeatedly until the determination of step S53 is "yes".

，且因子節點N3走訪次數為100次，達到第一閾值。因此可執行步驟S54。 In Figure 6C, all subtree path nodes of child node N3 have been explored. Calculation is based on the 37-18-43 path of child node N3. The total time consumption of this path is 98 (=37+18+43). The total time consumption of the paths are: 115, 91, 136, 96, 140, 64, 79, 96, and it is assumed that the number of visits to each path is: 9, 12, 13, 11, 11, 11, 11, 11, 11, Then W of child node N3 =98×9+115×12+91×13+136×11+96×11+140×11+64×11+79×11+96×11=10166, n =9+12+ 13+11+11+11+ 11+11+11=100, and it is known that t =124, so the child node N3

, and the number of visits to factor node N3 is 100 times, reaching the first threshold. Therefore step S54 can be executed.

In step S54, the host determines whether the number of visits to the original root node is not less than a second threshold (the second threshold must be greater than the first threshold). If the determination is "No", continue to step S55. If it is determined to be "yes", continue to step S56. In one embodiment, the second threshold is 1000, and the number of visits to the original root node in Figure 6C is only 124. Therefore, step S55 is continued.

In step S55, the host uses the candidate child node as the new root node and returns to step S51. Taking FIG. 6C as an example, the host regards the candidate child node (child node N3) as the new root node, and then repeats the process from step S51 to step S54 until the determination in step S54 is "no".

，且原始根節點的走訪次數為1000，已達第二閾值。因此，在步驟S56中，主機從運動策略樹選擇走訪次數具有最大值(假設357就是N1、N2、N3的最大n值)的子節點N3，並輸出此子節點N3代表的運動策略作為第二路段的最佳運動策略。 In Figure 6D, each sub-tree path node of sub-node N3 has been explored. Taking the 37-18-43 path of sub-node N3 for calculation, the total time consumption of this path is 98 (=37+18+43). The total time consumption of the paths are: 115, 91, 136, 96, 140, 64, 79, 96, and it is assumed that the number of visits to each path is: 13, 32, 26, 29, 29, 51, 120, 35, 22, Then W of child node N3 =98×13+115×32+91×26+136×29+96×29+140×51+64×120+79×35+96×22=33745, n =13+32+ 26+29+29+51+120+35+22=357, and it is known that t =1000, so the UCB of child node N3 1=

, and the number of visits to the original root node is 1000, which has reached the second threshold. Therefore, in step S56, the host selects the child node N3 with the maximum number of visits from the movement strategy tree (assuming 357 is the maximum n value of N1, N2, and N3), and outputs the movement strategy represented by this child node N3 as the second The optimal movement strategy for the road segment.

In the above embodiment, the condition for step S56 to be executed is that the determination of step S54 is "yes", that is, the number of visits to the original root node of the motion strategy tree is greater than or equal to the second threshold. In another embodiment, in order to speed up the execution of step S5, an early termination condition may be added to step S5. For example: exceeding the total time limit, exceeding the safe heart rate for a long time, using fixed muscle groups overtime or compensating, etc. Once it is determined that the above conditions are true, the motion strategy represented by the child node is output in advance.

The traditional Monte Carlo tree search algorithm is suitable for chess games, so its early termination condition can only be the end of the chess game. On the other hand, sports with multiple segments that are continuous in time, such as cycling, jogging, fitness, etc., are intended to complete the entire movement process, and therefore do not have the winning or losing conditions of a chess game. In other words, the optimal solution search method proposed in step S5 according to an embodiment of the present invention does not have a fixed early termination condition. If the exercise goal is to pursue the shortest riding time, there is no early termination condition because the user will definitely ride the entire route. Therefore, the optimal solution search method proposed by the present invention has the following advantages: the early termination condition can be replaced, the early termination condition does not need to exist, and the early termination condition can be a mixture of multiple conditions, such as heart rate and total time consumption.

To sum up, when the present invention is applied to bicycle riding, it can save the measurement time of CP and FTP, and can adopt the best shifting strategy globally (in the entire riding route) without accumulating a lot of experience. Specifically, the present invention establishes a riding performance prediction model under various factors by collecting various riding data in advance, and uses this model to output sports performance prediction values, such as the time required to ride a certain road section, thereby Avoid the need to calculate CP and FTP. After building the model, users can set the system target to the shortest riding time. The method proposed by the present invention to generate the global optimal motion strategy will use the route information to continuously call the model to expand all motion strategies and automatically find the global approximate optimal motion strategy. The solution allows users to achieve the best performance possible with their current physical abilities without having extensive riding experience.

Although the present invention is disclosed in the foregoing embodiments, they are not intended to limit the present invention. All changes and modifications made without departing from the spirit and scope of the present invention shall fall within the scope of patent protection of the present invention. Regarding the protection scope defined by the present invention, please refer to the attached patent application scope.

S1~S5: steps

Claims (8)

A method for generating a global optimal exercise strategy, suitable for users who perform a sport with a sports equipment, wherein the sport has a plurality of temporally continuous segments, and the segments include a first segment and at least a first segment. Two segments, and the method includes executing with a host: obtaining a plurality of environmental information corresponding to the segments from an environment server; when the user performs the first segment of the exercise with the sports equipment , obtain a physiological state from a physiological sensor provided on the user, and obtain an equipment state from an equipment sensor provided on the sports equipment; determine a plurality of exercise strategies, which at least correspond to the Multiple setting values for a setting of sports equipment; establishing a sports strategy topology including a plurality of nodes, wherein each of the nodes represents one of the sports strategies, and the branching degree of each of the nodes Correlated with the number of the sports strategies, the height of the sports strategy topology is related to the number of the segments; and selecting one from a plurality of child nodes of the first layer of the sports strategy topology based on a plurality of sports performance prediction values. or, and output the exercise strategy represented by the child node as the best exercise strategy of the at least one second segment, wherein each of the exercise performance prediction values is input to the exercise represented by one of the nodes. The strategy, the environmental information, the physiological state and the equipment state are combined into a sports performance model; wherein the sports performance model is a neural network trained based on a sports history; the sports history includes multiple pieces of data, and the multiple data Each piece of data includes a timestamp, reference environmental information, reference physiological state, and reference equipment state; and Each of the athletic performance predictions is an estimated time to complete one of the segments of the exercise. A method for generating a global optimal exercise strategy, suitable for users who perform a sport with a sports equipment, wherein the sport has a plurality of temporally continuous segments, and the segments include a first segment and at least a first segment. Two segments, and the method includes executing with a host: obtaining a plurality of environmental information corresponding to the segments from an environment server; when the user performs the first segment of the exercise with the sports equipment , obtain a physiological state from a physiological sensor provided on the user, and obtain an equipment state from an equipment sensor provided on the sports equipment; determine a plurality of exercise strategies, which at least correspond to the Multiple setting values for a setting of sports equipment; establishing a sports strategy topology including a plurality of nodes, wherein each of the nodes represents one of the sports strategies, and the branching degree of each of the nodes Correlated with the number of the sports strategies, the height of the sports strategy topology is related to the number of the segments; and selecting one from a plurality of child nodes of the first layer of the sports strategy topology based on a plurality of sports performance prediction values. or, and output the exercise strategy represented by the child node as the best exercise strategy of the at least one second segment, wherein each of the exercise performance prediction values is input to the exercise represented by one of the nodes. The strategy, the environmental information, the physiological state and the equipment state are obtained by converting the sports performance model into a sports performance model; wherein one of the sub-nodes of the first layer of the sports strategy topology is selected based on the sports performance prediction values. , and output the motion strategy represented by the child node as the best motion strategy of the at least one second segment, including: Step 1: Perform a single random visit operation on each of the child nodes and perform an update procedure. The update procedure includes: accumulating the number of visits to each of the child nodes; accumulating the total visits of the motion strategy topology. times; and calculate the evaluation parameters of each of the sub-nodes based on the predicted sports performance values of the multiple visited nodes passed by the single random visit operation, the number of visits and the total number of visits; step 2, in the Among these sub-nodes, select the sub-node with the maximum evaluation parameter to perform the single random visit operation and perform the update procedure; step 3, repeat step 2 until there is a candidate sub-node of the sub-nodes. The number of visits is not less than the first threshold; and step 4, use the candidate child node as the root node of the motion strategy topology, and repeat steps 1 to 3; and step 5, when the original root node of the motion strategy topology has When the number of visits is not less than the second threshold, one of the child nodes of the first layer of the motion strategy topology is selected, and the number of visits of the selected child node has the maximum value. The method of generating a global optimal sports strategy as described in claim 1, wherein one is selected from the sub-nodes of the first layer of the sports strategy topology according to the sports performance prediction values, and the representative of the sub-node is output The movement strategy as the optimal movement strategy of the at least one second segment includes: step 1, performing a single random visit operation on each of the child nodes and performing an update procedure, the update procedure includes: The number of visits to each of the child nodes is accumulated; the total number of visits to the motion strategy topology is accumulated; and the motion performance prediction values, the number of visits, and the number of visits are based on the multiple visited nodes passed by the single random visit operation. The total number of visits is used to calculate the evaluation parameters of each of the sub-nodes; step 2, among the sub-nodes, select the sub-node with the maximum evaluation parameter to perform the single random visit operation and perform the update procedure ; Step 3, repeat Step 2 until there is a candidate child node among the child nodes whose number of visits is not less than the first threshold; and Step 4, use the candidate child node as the root node of the motion strategy topology, and Repeat steps 1 to 3; and step 5, when the number of visits of the original root node of the motion strategy topology is not less than the second threshold, select one of the child nodes of the first layer of the motion strategy topology. Or, the number of visits of the selected child node has the maximum value. The method of generating a global optimal sports strategy as described in claim 1 or 2, wherein the sports equipment is a bicycle, the sport is cycling, and the segments correspond to multiple riding sections; the environmental information includes: corresponding At least one of distance, slope, temperature, average altitude, and altitude difference in each of the riding sections; the physiological state includes: the user's pre-riding heart rate and the user's post-riding heart rate At least one of; the equipment status includes: at least one of vehicle speed, rotational speed and pedal power; The setting includes: a gear ratio or gear; and each of the sports performance prediction values is at least one of riding time and calorie consumption. A system for generating a global optimal exercise strategy, suitable for users who perform an exercise with a sports equipment, wherein the exercise has a plurality of time-continuous segments, and the segments include a first segment and at least a first segment. Two segments, and the system includes: an environment server that stores a plurality of environmental information corresponding to the segments; a physiological sensor that is disposed on the user, and when the user performs the exercise with the exercise equipment Obtain a physiological state of the user during the first segment; the equipment sensor is disposed on the sports equipment, and obtains the sports equipment when the user performs the first segment of the exercise with the sports equipment. an equipment state; and a host, communicatively connected to the environment server, the physiological sensor and the equipment sensor, the host is used to execute a plurality of instructions to output the optimal movement strategy of the at least one second segment , the instructions include: determining a plurality of exercise strategies, which at least correspond to a plurality of setting values of a setting of the exercise equipment; establishing an exercise strategy topology including a plurality of nodes, wherein each of the nodes Representing one of the motion strategies, the branching degree of each of the nodes is related to the number of the motion strategies, the height of the motion strategy topology is related to the number of the segments; and based on multiple motions The performance prediction value selects one of a plurality of sub-nodes of the first layer of the sports strategy topology, and outputs the sports strategy represented by the sub-node as the best sports strategy of the at least one second segment, wherein the Each of the athletic performance prediction values is entered into the The movement strategy, the environmental information, the physiological state and the equipment state represented by one of the nodes are obtained by converting the movement strategy, the environmental information, the physiological state and the equipment state to a movement performance model; wherein the movement performance model is a neural network trained based on a movement process; The exercise history includes multiple pieces of data, each of the multiple pieces of data includes a timestamp, reference environmental information, reference physiological state, and reference equipment status; and each of the estimated exercise performance values is the result of completing the exercise. The estimated time for one of these segments. A system for generating a global optimal exercise strategy, suitable for users who perform an exercise with a sports equipment, wherein the exercise has a plurality of time-continuous segments, and the segments include a first segment and at least a first segment. Two segments, and the system includes: an environment server that stores a plurality of environmental information corresponding to the segments; a physiological sensor that is disposed on the user, and when the user performs the exercise with the exercise equipment Obtain a physiological state of the user during the first segment; the equipment sensor is disposed on the sports equipment, and obtains the sports equipment when the user performs the first segment of the exercise with the sports equipment. an equipment state; and a host, communicatively connected to the environment server, the physiological sensor and the equipment sensor, the host is used to execute a plurality of instructions to output the optimal movement strategy of the at least one second segment , the instructions include: determining a plurality of exercise strategies, which at least correspond to a plurality of setting values of a setting of the exercise equipment; Establish a motion strategy topology including a plurality of nodes, wherein each of the nodes represents one of the motion strategies, the branching degree of each of the nodes is associated with the number of the motion strategies, the The height of the sports strategy topology is related to the number of the segments; and selecting one from a plurality of sub-nodes of the first layer of the sports strategy topology according to the plurality of sports performance prediction values, and outputting the sub-node represented by the The exercise strategy serves as the optimal exercise strategy for the at least one second segment, wherein each of the exercise performance prediction values is the exercise strategy, the environmental information, and the physiological input represented by one of the nodes. The state and the equipment state are obtained from a sports performance model; wherein one is selected from the sub-nodes of the first layer of the sports strategy topology according to the sports performance prediction values, and the sub-node represented by the sub-node is output. As the optimal movement strategy for the at least one second segment, the motion strategy includes: step 1, performing a single random visit operation on each of the child nodes and performing an update procedure. The update procedure includes: accumulating the The number of visits to each child node; the total number of visits to the sports strategy topology; and the predicted sports performance values, the number of visits, and the total visits based on the multiple visited nodes passed by the single random visit operation Calculate the evaluation parameters of each of the sub-nodes several times; Step 2, among the sub-nodes, select the sub-node with the maximum evaluation parameter to perform the single random visit operation and perform the update procedure; Step 3 , repeat step 2 until there is a candidate child node among the child nodes of the first layer and the number of visits is not less than the first threshold; and Step 4, use the candidate child node as the root node of the motion strategy topology, and repeat steps 1 to 3; and step 5, when the number of visits of the original root node of the motion strategy topology is not less than the second threshold, One of the child nodes of the first layer of the movement strategy topology is selected, and the number of visits of the selected child node has the maximum value. The system for generating a global optimal sports strategy as described in claim 5, wherein one of the sub-nodes of the first layer of the sports strategy topology is selected according to the sports performance prediction values, and the sub-node representative is output The movement strategy as the optimal movement strategy for the at least one second segment includes: step 1, performing a single random visit operation on each of the child nodes and performing an update procedure. The update procedure includes: accumulation The number of visits to each of the sub-nodes; the total number of visits to the motion strategy topology; and the prediction values of the motion performance of the multiple visited nodes passed by the single random visit operation, the number of visits and the The total number of visits is used to calculate the evaluation parameters of each of the sub-nodes; step 2, among the sub-nodes, select the sub-node with the maximum evaluation parameter to perform the single random visit operation and perform the update procedure; Step 3: Repeat step 2 until there is a candidate child node among the child nodes of the first layer whose visit times are not less than the first threshold; and step 4, use the candidate child node as the motion strategy topology. the root node and repeat steps 1 to 3; and Step 5: When the number of visits to the original root node of the motion strategy topology is not less than the second threshold, select one of the child nodes of the first layer of the motion policy topology, and the selected child node is selected. The number of visits to the node has the maximum value. A system for generating a global optimal sports strategy as described in claim 5 or 6, wherein the sports equipment is a bicycle, the sport is cycling, and the segments correspond to multiple riding sections; the environmental information includes: corresponding At least one of distance, slope, temperature, average altitude, and altitude difference in each of the riding sections; the physiological state includes: the user's pre-riding heart rate and the user's post-riding heart rate At least one of; the equipment status includes: at least one of vehicle speed, rotational speed and pedal power; the setting includes: gear ratio or gear; and each of the sports performance prediction values is the riding time and At least one of caloric expenditure.

Images