JP2009301455A - Input determination means for moving vehicle, and hybrid vehicle and multijoint mobile robot having the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To minimize loss in specific operation by solving an optimization control problem even in a hybrid vehicle and a mobile robot difficult to predict a state in the future thereof. <P>SOLUTION: An input determination means 15 generates internal input 16 with external input 11 and output 12 as the input. A moving vehicle 10 targets an object which has redundancy and a plurality of methods for generating the output 12 corresponding to a certain external input 11. Internal input 16 determines an evaluation function, decided beforehand by the input determination means 15, so as to minimize or the maximize the evaluation function. The input determination means 15 have a future state prediction means 21, an optimization operation means 22 and a constraint condition determination means 23. When a solution obtained by the optimization operation is impossible to be executed, an optimization operation result notice means 27 is arranged in order to notify the outside of the fact, and is materialized as an LED, a liquid crystal display, a memory, and the like. This not only facilitates debugging at the time of development, but also is used as off-line diagnosis. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to various means for determining a certain state, such as a hybrid vehicle or a robot, that is, a method for minimizing the energy used in an object having redundancy. This is suitable when it is uncertain whether to operate in a normal manner.

  In general, when the path of a moving object is determined, the target is formulated as an optimal control problem. An example of this technique is described in Patent Document 1, for example. This known example relates to a four-dimensional navigation object represented by an airplane, and refers to a method of minimizing the amount of shift from a predetermined trajectory and the energy used.

JP 2004-326363 A

  However, since the optimal control problem is generally formulated so that the path integral from the initial state to the terminal state is minimized or maximized, it is necessary to predict in advance the changes in the flight path and the outside world with high accuracy. Don't be. The route of the hybrid vehicle, which is the main subject of the present invention, can change from moment to moment according to the driver's intention. In addition, it is expected that the conventional road map cannot be used particularly in the case of a disaster, and therefore it is necessary to generate an input command signal to the component on the spot.

In order to solve the above problems, the present invention provides:
A moving body having a host controller that creates an input command and a lower controller that controls a component according to the input command of the host controller, wherein the host controller predicts a state after a predetermined time ΔT by a primary hold calculation It has a prediction means, and the host controller has an optimization calculation means for determining an input so that the cumulative loss from the current time t0 to the future time t0 + ΔT is minimized under a given constraint condition.

Moreover, preferably of the present invention,
A moving body having a host controller for generating an input command and a lower controller for controlling components in accordance with the input command of the host controller, wherein the host controller receives a state after a predetermined time ΔT predicted by the navigation system. It is characterized by having an optimization calculation means for determining an input so that the cumulative loss from time t0 to future time t0 + ΔT is minimized under given constraint conditions.

Moreover, preferably of the present invention,
The moving body has a discrete state variable, and the optimization calculation means searches for a path with a minimum loss for an executable state that can be taken by the discrete state variable at each sampling time Δt, and the optimization calculation means The time history of the discrete state variable that minimizes the loss at time ΔT is determined.

Moreover, preferably of the present invention,
The lower controller includes a notification unit that notifies the upper controller of a constraint condition of the component.

Moreover, preferably of the present invention,
The host controller has an optimization calculation result notification means for notifying the calculation result of the constraint condition and the optimization result.

Moreover, preferably of the present invention,
The upper controller has distance calculating means for calculating a distance between a solution obtained as a result of optimization and an inequality constraint conditional expression, and the upper controller has a time when the distance becomes smaller than a predetermined threshold. It has a diagnostic means which detects the malfunction of components.

Moreover, preferably of the present invention,
The above feature is applied to a hybrid vehicle.

Moreover, preferably of the present invention,
The above feature is applied to an articulated mobile robot.

  According to the present invention, the host controller has a future state predicting means for predicting a state after a predetermined time ΔT by a primary hold calculation, and the host controller is provided with a cumulative constraint from the current time t0 to the future time t0 + ΔT. Since the optimization calculation means for determining the input so as to be the minimum under the conditions is provided, there is an effect of performing the optimization calculation even when the future trajectory is unknown to minimize the loss of the moving body.

  In addition, according to the present invention, there is provided a mobile body having a host controller that generates an input command and a lower controller that controls a component in accordance with the input command of the host controller, wherein the host controller is a predetermined time predicted by the navigation system. Since there is an optimization calculation means that determines the input so that the cumulative loss from the current time t0 to the future time t0 + ΔT is minimized under a given constraint condition after receiving the state after ΔT, it is possible to set the future trajectory. It is possible to perform an optimization operation. As a result, there is an effect of minimizing the loss of the moving body.

  Further, according to the present invention, the mobile body has a discrete state variable, and the optimization calculation means searches for a path that minimizes a loss for an executable state that the discrete state variable can take every sampling time Δt, and Since the optimization calculation means can determine the time history of the discrete state variable that minimizes the loss at the predetermined time ΔT, it is also optimized for control targets having discrete states such as engine start / stop and stepped transmission switching. Is possible.

  In addition, according to the present invention, since the lower-level controller has notification means for notifying the upper-level controller of component constraint conditions, there is an effect that it is easy to jointly develop a huge system by a plurality of manufacturers.

  Further, according to the present invention, the host controller has an optimization calculation result notification means for notifying the calculation result of the constraint condition and the optimization result, so the information sent from the optimization calculation result notification means By incorporating it into the development environment, you can expect effects such as offline diagnosis and debugging.

  Further, according to the present invention, the host controller has distance calculation means for calculating a distance between a solution obtained as a result of optimization and an inequality constraint conditional expression, and the host controller has a threshold value for which the distance is predetermined. Since it has a diagnostic means to detect a malfunction of a part when it becomes smaller, it can always monitor the failure state during operation, so it can notify the driver of the replacement time of parts and the necessity of inspection There is an effect.

  Moreover, by applying the present invention to a hybrid vehicle, it is possible to perform the most efficient driving in the target system, which is suitable for reducing fuel consumption.

  Further, by applying the present invention to an articulated mobile robot, it becomes possible to operate a desired motion with the most energy efficiency, which is suitable for reducing fuel consumption.

  Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

  FIG. 1 is an overall explanatory view of a moving body input determining means using the present invention. Here, 10 is the target mobile object. 11 is an external input, 12 is an output, 13 is a control compensator, 14 is a control object, 15 is an input determining means according to the present invention, and 16 is an internal input. For example, when the moving body 10 is a hybrid vehicle, the external input 11 is the driver's accelerator opening and brake pedaling force, and the output 12 is the engine generated torque, motor generated torque, battery current, and the like. Further, the control object 14 is an engine or a motor, and the control compensator 13 takes both the external input 11 and an internal input 16 described later as inputs, and creates an input for suitably controlling the control object 14. For example, it is configured by air control, fuel injection amount control, vector control, and the like.

  The input determining means 15 receives the external input 11 and the output 12 and generates an internal input 16 described later. The mobile object 10 has redundancy, and is intended for a plurality of methods for generating an output 12 corresponding to a certain external input 11. In the case of the hybrid vehicle described above, when the external input 11 is the accelerator opening and the output 12 is the vehicle driving force, all of the vehicle driving force may be generated by a motor, or all may be provided by an engine. The internal input 16 is an input provided to eliminate such redundancy. In the case of the hybrid vehicle described above, examples of the internal input include engine torque and motor torque.

  The internal input 16 is determined by the input determining means 15 so as to minimize or maximize a predetermined evaluation function. In the case of a hybrid vehicle, the evaluation function is, for example, fuel consumption or loss, and is determined so as to minimize loss. By multiplying the evaluation function itself by −1, the condition for maximizing the evaluation function can be easily converted to the condition for minimizing the evaluation function. Therefore, hereinafter, the evaluation function is treated as being minimized without particular notice.

  The input determination unit 15 includes a future state prediction unit 21, an optimization calculation unit 22, and a constraint condition determination unit 23. Generally, optimization calculations that minimize losses cannot be calculated unless the future operating route of the vehicle has been determined in advance, but in the case of passenger cars, the route can be determined freely by the driver, so it is optimal as it is. Unable to perform digitization operation. The future state prediction means 21 is provided to solve this problem. Details of the future state prediction means 21 will be described later.

  The constraint condition determination means 23 determines the relationship between the optimization calculation result and the constraint condition. In the case of a hybrid vehicle, there is a vehicle that reduces the physique of the motor in order to reduce the weight of the electric system and does not assume traveling by the motor alone at a high load. In such a case, a feasible solution cannot be obtained even if an optimization calculation that can be performed only by the motor is performed. The constraint condition determination means 23 is provided in view of such a case, and has functions of re-execution of an optimal solution with a changed initial value and selection of an alternative solution. Further, as will be described later, it is possible to have a function of realizing a diagnostic function from the evaluation result of the constraint condition expression.

  The input determining means 15 has an initial value map 24, an evaluation function / constraint condition database 25, and an alternative input database 26. In general, the optimization operation is a sequential convergence operation, and an initial value needs to be determined. The initial value map 24 maps in advance the initial values that are likely to converge to an executable solution in accordance with the external input 11 and the output 12. As a result, the number of computations can be reduced, and an evaluation function and constraint conditions can be handled as a linear approximation expression by selecting an approximate solution with sufficiently high accuracy, so that a significant reduction in computation time can be expected. The evaluation function / constraint condition database 25 stores loss evaluation formulas and constraint formulas for the controlled object 14, and the optimization calculation means 22 performs optimization calculations using these loss evaluation formulas and constraint formulas.

  Generally, the loss evaluation formula and the constraint condition formula are formulated as the following formula.

  In the above equation, x is a state variable and u is an internal input variable. Equation (1) is an evaluation function for finding the minimum value. Equation (2) is a state equation describing the target system, Equation (3) is an inequality constraint, and Equation (4) is an equality constraint. Equation (5) is an initial value. In addition, since both formula (2) and formula (4) can be handled mathematically equivalently, formula (2) and formula (4) are not distinguished from each other, and both are treated as equality constraints. Since the solution of the optimization problem described by the combination of Equation (1) to Equation (5) is disclosed in, for example, Non-Patent Document 1 (Hisashi Tamaki, “System Optimization”, Ohmsha (2005)). Details will not be described.

  When an evaluation function and constraint conditions are written down for a hybrid vehicle that is one of the objects of the present invention, the following expression is obtained, for example. Equation 2 corresponds to the evaluation function L, Equation 3 corresponds to the state equation, Equation 4 corresponds to the inequality constraint, and Equation 5 corresponds to the equality constraint. Further, FIG. 2 illustrates the symbols used in the equations 2 to 5.

  The input determining means 15 includes an alternative input database 26. The alternative input database 26 holds a set of executable internal inputs. The set of internal inputs held in the alternative input database 26 may not necessarily be optimal as long as it is feasible. This set of internal inputs is used by the constraint condition judging means 23 as the next best internal input when the optimization operation is not successful.

  If there is an evaluation function / constraint condition database 25, the constraint condition determination means 23 can be easily realized by substituting the solution candidates of the internal input into the constraint condition expression. For example, in equation (3), if the inequality constraint condition gi (x)> 0, it can be determined that the i-th inequality constraint condition has not been realized, and it can be understood that it cannot be executed. The optimization calculation result notification means 27 is provided to notify the relationship between the solution obtained by the optimization calculation and the constraint conditions to the outside, and is realized as an LED, a liquid crystal display, a memory, or an external connector. The This information is sent to the diagnostic means 17 and the development environment 18 and can be used for diagnosis and debugging. The optimization calculation result notifying unit 27 may include a data holding unit that stores a time series of optimization calculation results. This has the effect of facilitating offline diagnosis.

  FIG. 3 is an explanatory diagram of the entire mobile body input determination means that enables online diagnosis. Here, 28 is a distance calculation means for calculating the distance between the optimum solution obtained as a result of optimization and the inequality constraint condition. Reference numeral 29 denotes a diagnostic monitor that displays a diagnostic result. The difference from FIG. 1 is that the diagnostic means 17 is mounted in a target mobile body, for example, a hybrid vehicle. As a result, the driver can always monitor the vehicle state. A diagnostic method using the distance calculation means 28 will be described later.

  FIG. 4 is an explanatory diagram of the future state predicting means 21 in FIG.

  In the figure, as an example of the future state predicting means 21 of the hybrid vehicle, a case where the vehicle speed is predicted as the horizontal axis time and the vertical axis vehicle speed is shown. It is assumed that the current time is t0, the predetermined evaluation time interval is ΔT, and the evaluation end time is tf (= t0 + ΔT). The vehicle speed at t0 is v0, and the vehicle speed increases at a constant acceleration from t0 to tf. .., Vf at intermediate times t1, t2,..., Tf are determined using this characteristic (primary hold). The acceleration at the current time t0 is determined by the past actual vehicle speed history. Very simply, a moving average of the vehicle speed over a certain period of time can be used.

  In this way, the vehicle operation at the evaluation time interval ΔT can be determined, and the optimization calculation can be executed.

  FIG. 5 is an explanatory diagram of the future state predicting means 21 including discrete states.

  In a hybrid vehicle, there are at least two states of “operating the engine” and “stopping the engine”. Since the constraint condition expressions are different, it is difficult to handle the two in a unified manner. Here, the fact that the number and contents of the constraint condition expressions change significantly is referred to as “discrete state”.

  Now, assuming that the number of discrete states is 2 and the number of divisions from the current time t0 to the evaluation end time tf is 3, the loss Li (i is the time, an index representing the discrete state) is expressed as a search tree as shown in FIG. The From these, a search is made for Li that minimizes the loss at the final time t2, and an internal input that realizes the loss Lj at the current time t0 connected to Li may be selected. For example, assuming that L211 is the minimum loss node at time t2, the node selected at the current time t0 is L2. Assuming that L1 is the running state with only the motor and L2 is the state where the engine is running, it is understood that it is better to select to operate the engine at the current time t0.

  FIG. 6 shows an application example of the present invention when a navigation system is used. Here, 19 is a navigation system. Since the navigation system 19 can predict the driving state of the vehicle to some extent by setting the departure point and the destination, the navigation system 19 can be used as the future state prediction means 21 (FIGS. 1 and 3). In addition, since the navigation system 19 often has a monitor, the navigation system 19 can also serve as a diagnostic monitor 29 (FIG. 3) in this figure.

  FIG. 7 shows a configuration example for realizing the present invention. In addition, the control system of the hybrid vehicle is assumed, and is a minimum example necessary for explanation for simplicity.

  31 is a host controller according to the present invention, 32 is a motor controller, 33 is a motor, 34 is an engine controller, and 35 is an engine. The host controller 31, the motor controller 32, and the engine controller 34 exchange information with each other through the bus connection network 36. The host controller 31 has the input determining means according to the present invention.

  In general, since the system size of the hybrid vehicle targeted in this figure is very large, the host controller 31, the motor controller 32, and the engine controller 34 are often manufactured by different manufacturers. At this time, it is expected that the manufacturer of the host controller 31 is difficult to obtain sufficient information for creating the input determining means, which makes it difficult to implement. However, as shown in FIG. 7, this difficulty is alleviated when the host controller 31 receives the function information 37 from the motor controller 32 and the engine controller 34. Here, the evaluation function and the constraint conditions are collectively referred to as function information 37. The details will be described below.

  In general, the optimization operation needs to acquire the constraint condition in order to minimize the evaluation function under the constraint condition. For example, when the manufacturer of the host controller 31 (hereinafter referred to as “Company A”) is different from the manufacturer of the engine controller 34 (hereinafter referred to as “Company B”), the Company “S” does not have information on an evaluation function such as a fuel consumption rate map. . In order to create the host controller 31, it is necessary to obtain the fuel consumption rate map from B. However, the fuel consumption rate map is data that is the core of engine performance, and in most cases is a trade secret. Therefore, it is customary that Otsu-sha does not want to disclose the fuel consumption rate map information if possible.

  There is also a problem that it is difficult to achieve consistency between the host controller 31 and the engine controller 34 when the software is modified due to engine improvement or component review.

  The above problem is solved by sending the function information 37 to the host controller 31 by the motor controller 32 or the engine controller 34. As a result, the host controller 31 can import the evaluation functions and constraint functions of the motor controller 32 and the engine controller 34 as necessary, and can be installed and executed by the host controller 31, thereby ensuring consistency between the controllers. The problem of difficulty is solved. Further, by encrypting the function information 37 and sending it in a form that can only be executed, problems associated with the disclosure of confidential information can be solved.

  FIG. 8 shows an example of the failure diagnosis method mentioned in FIGS. Note that the input determination means for a moving object according to the present invention generally solves a multivariable optimization problem, but here it is expressed as a two-dimensional optimization of variables 1 and 2 for simplicity. Also, the equality constraint condition is not shown for simplicity.

  41 is a first inequality constraint, 42 is a second inequality constraint, 43 is a contour display of the value of the evaluation function, 44 is a variable 1 defined by the first inequality constraint 41 and the second inequality constraint 42 This is the executable area of variable 2. In FIG. 8, two constraint conditions are illustrated as second inequality constraint conditions, which are denoted by 42a and 42b. 42a is the second constraint before the change with time, and 42b is the second constraint after the change with time.

  45 is the optimum point obtained as a result of the optimization calculation, 45a is the optimum point before the change with time, and 45b is the optimum point after the change with time. In the figure, m1a is the distance between the optimum point 45a and the first inequality constraint 41 before the change with time, and m2a is the distance between the optimum point 45a and the second inequality restriction 42a before the change with time.

  When the second inequality constraint condition changes from 42a to 42b due to the change over time, the optimum point derived by the optimization calculation also changes from 45a to 45b. Furthermore, the optimum point 45b after the change with time is located on the boundary of the second inequality constraint condition 42b after the change with time. This means that the optimum point 45 is affected by the change over time. At this time, if the distance between the optimal point 45b after the change with time and the second inequality constraint condition 42b is m2b, m2b is zero.

  In this way, at a predetermined operating point, for example, in a driving state where the hybrid vehicle has a certain speed and certain acceleration, the constraint condition evaluation formula is monitored, and if the positional relationship between the inequality constraint condition 42 and the optimum point 45 can be grasped, performance degradation will occur. It can be detected. Specifically, the distance between the expression representing the inequality constraint condition 42 and the optimum point 45 may be monitored. When this becomes 0, it means that the performance has changed due to deterioration over time, so it can be determined as a sign of failure. In addition, by monitoring the rate of change of the distance between the inequality constraint condition 42 and the optimum point 45, the time margin until the constraint condition 42 affects the optimum point 45 can be understood, so a failure sign can be diagnosed. It becomes possible.

  Although the distance between the optimum point 45 and the inequality restriction condition 42, various known calculation means such as a city area distance, an Euclidean distance, and a Mahalanobis distance can be applied. Since the method for calculating the distance is known, the description thereof is omitted here.

  FIG. 9 shows a method for estimating a failure site by using a trigger when the distance between the inequality sign constraint condition 42 and the optimum point 45 becomes zero. Here, 51 is a diagnostic manager, 52 is a diagnostic database, and 29 is a diagnostic monitor. The diagnosis manager 51 receives the identifier of the inequality constraint condition expression whose distance from the optimum point is 0, and issues a search query to the diagnosis database 52. Here, a case where SQL is used as a search query is illustrated. The diagnosis database 52 sends related part information obtained as a search result to the diagnosis manager 51. The diagnosis manager 51 converts the related component information into a displayable format and displays it on the diagnosis monitor 29.

  In this figure, for the sake of simplicity, an example using a database is shown, but other means may be used as long as information can be searched. For example, backtracking using a logic language such as prolog The search may be used, or pattern matching may be performed in order from the top. The implementation means is determined as appropriate according to available computing resources.

  As described above, the example of the input determining means mainly applied to the hybrid vehicle has been described. However, the present invention can be effectively applied to other systems in which it is difficult to predict the future state as long as it is a redundant input system. For example, in a disaster-relief vehicle, road information known in advance cannot be used due to road breaks or sudden changes in road surface conditions. In order to make the best use of the limited energy under such circumstances, it is necessary to rely on the future state prediction means shown in the present invention. In addition, for example, in a welfare / nursing care robot, the target is a human, and there are a wide variety of variables that need to be considered, such as parts that require support, physique, and health status. In such a system, assuming a usage pattern in advance is not a good idea from the viewpoint of storage capacity, and on-demand future state prediction works effectively. In particular, an articulated device such as a robot is an object to which the optimization calculation according to the present invention is advantageous because there are almost infinite combinations of joint angles when realizing the same hand position.

It is a whole explanatory drawing of the input determination means of a moving body. It is description of the symbol used in the type | formula. It is a whole explanatory drawing of the input determination means of the moving body which enables online diagnosis. 5 is an explanatory diagram of future state prediction means 21. FIG. 6 is an explanatory diagram of a future state prediction means 21 including a discrete state. FIG. It is an example of application of this invention at the time of using a navigation system. It is a structural example which implement | achieves this invention. It is an example of a failure diagnosis method. It is illustrated about the method of estimating a failure location.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 ... Mobile body, 11 ... External input, 12 ... Output, 13 ... Control compensator, 14 ... Control object, 15 ... Input decision means, 16 ... Internal input, 17 ... Diagnosis means, 18 ... Development environment, 19 ... Navigation system , 21 ... Future state prediction means, 22 ... Optimization calculation means, 23 ... Restriction condition determination means, 24 ... Initial value map, 25 ... Evaluation function / constraint condition database, 26 ... Alternative input database, 27 ... Optimization calculation result notification Means 28 ... Distance calculation means 29 ... Diagnostic monitor 31 ... Host controller 32 ... Motor controller 33 ... Motor 34 ... Engine controller 35 ... Engine 36 ... Bus connection network 37 ... Function information 41 ... No. First inequality constraint, 42 ... Second inequality constraint, 43 ... Contour value of evaluation function, 44 ... Executable region of variable 1 and variable 2, 45 ... Optimal point, 51 ... Diagnostic manager, 52 ... Diagnosis Database.

Claims (8)

  A moving body having a host controller that creates an input command and a lower controller that controls a component according to the input command of the host controller, wherein the host controller predicts a state after a predetermined time ΔT by a primary hold calculation And a high-order controller having optimization calculation means for determining an input so that a cumulative loss from a current time t0 to a future time t0 + ΔT is minimized under a given constraint condition. Body input decision means.   A moving body having a host controller that creates an input command and a lower controller that controls components in accordance with the input command of the host controller, the host controller receiving a state after a predetermined time ΔT predicted by a navigation system. An input determining means for a moving body comprising an optimization calculating means for determining an input so that a cumulative loss from a time t0 to a future time t0 + ΔT is minimized under a given constraint condition.   3. The mobile unit according to claim 1, wherein the moving body has a discrete state variable, and the optimization calculating unit searches for a path that minimizes a loss for an executable state that can be taken by the discrete state variable at each sampling time Δt. The mobile unit input determining means, wherein the optimization calculating means determines the time history of the discrete state variable that minimizes the loss at the predetermined time ΔT.   3. The moving body input determination means according to claim 1, wherein the lower controller includes notification means for notifying the upper controller of a constraint condition of the component.   3. The moving body input determination means according to claim 1, wherein the host controller includes an optimization calculation result notification means for notifying the calculation result of the constraint condition and the optimization result.   3. The high-order controller according to claim 1, wherein the high-order controller has distance calculation means for calculating a distance between a solution obtained as a result of optimization and an inequality constraint conditional expression, and the high-order controller has the distance determined by a predetermined threshold. A moving body input determining means comprising diagnostic means for detecting a malfunction of a part as it becomes smaller.   A hybrid vehicle comprising the input determining means according to any one of claims 1 to 6.   An articulated mobile robot comprising the input determining means according to any one of claims 1 to 6.

Images