KR101231982B1 - Navigation control method for mobile robot using gspn and mobile robot using the same - Google Patents

Abstract

The present invention relates to a traveling control method of a mobile robot using GSPN and a mobile robot using the same. The traveling control method according to the present invention includes the steps of: (a) registering a mutually different first travel control technique and a second travel control technique to the mobile robot; (b) registering a traveling control GSPN model modeled by applying a generalized stochastic petri net (GSPN) technique to the first traveling control technique and the second traveling control technique; (c) executing the travel control GSPN model to select one of the first travel control technique and the second travel control technique; (d) controlling driving of the mobile robot according to any one selected from the first travel control technique and the second travel control technique, wherein the travel control GSPN model selects the first travel control technique. A first travel selection transition for igniting a hazard, a second travel selection transition for ignition for the selection of the second travel control technique, a selection for selection of any one of the first travel control technique and the second travel control technique A place; The step (c) may include: (c1) driving a mission input to the mobile robot, (c2) moving a token to the selection place, (c3) the first driving selection transition and the second driving; Evaluating the performance of the first travel control technique and the second travel control technique based on each parameter of the selected transition; and (c4) the first travel control technique and the second travel control based on a result of the performance evaluation. Wherein any one of the techniques is selected. Accordingly, while a plurality of driving control techniques are registered in the mobile robot, a driving control technique suitable for a driving environment, for example, a static environment and a dynamic environment, is selected to allow the mobile robot to travel and to adapt to various driving environments. Driving performance can be improved.

Description

NAVIGATION CONTROL METHOD FOR MOBILE ROBOT USING GSPN AND MOBILE ROBOT USING THE SAME

The present invention relates to a traveling control method of a mobile robot using a GSPN and a mobile robot using the same. More specifically, the driving of the mobile robot is controlled by any one of a plurality of traveling control techniques using a Generalized Stochastic Petri Net (GSPN) technique. The present invention relates to a traveling control method of a mobile robot using GSPN that can adapt to various driving environments, and a mobile robot using the same.

Recently, various driving control techniques for autonomous navigation of mobile robots have been developed. Two big flows in the study of the traveling control technique of the mobile robot are the study of the model-based deliberate control method and the sensor-based reactive control method.

The contemplated driving control technique yields control commands from the environmental model and shows good performance in static environments, for example static environments with few obstacles not recorded on the environment map. On the other hand, the contemplated driving control technique is not suitable for an obstacle-rich environment, for example, a dynamic environment with frequent traffic.

In contrast, the responsive travel control technique uses sensor information obtained from various sensors to calculate a control command, thereby exhibiting fast and robust performance against uncertainty in a dynamic environment. However, responsive travel control techniques are not suitable for performing delicate tasks such as passing doors.

Due to the complementary advantages and disadvantages between the contemplated travel control technique and the reactive travel control technique, it is difficult to derive the travel control technique of a mobile robot that can be completely replaced by various travel environments with any one travel control technique.

Due to the limitations of such individual driving control techniques, studies have been proposed to integrate the deliberated driving control technique and the reactive driving control technique. For example, O.Brock and O.Khatib wrote "High speed navigation using the global dynamic window approach", In Proceedings of IEEE International Conference on Robotics & Automation, Detroit, Michigan, USA, May 10-15, 1999. In this paper, we proposed the Global Dynamic Window Approach technique through the known DWA (Dynamic Window Approach) technique and path planner (J. Latombe, "Robot Motion Planning", Kluwer, 1991.) for calculating NF (Navigation Function).

J. Ulrich and J. Borenstien also reported in the paper "VFH:" Local Obstacle Avoidance with Look-Ahead Verification ", In Proceedings of IEEE International Conference on Robotics & Automation, San Francisco, USA, April 24-28, 2000. In this paper, we propose a vector field histogram (VFH) technique, which uses look-ahead verification to avoid local minimum.

In addition, Stachniss, C. and Burgard, W., wrote, "An Integrated Approach to Goal-directed Obstacle Avoidance under Dynamic Constraints for Dynamic Environments", In Proceedings of IEEE International Conference on Intelligent Robots and Systems, Switzerland, Oct. 2-4, 2002., proposed a method that integrates path planning and sensor-based collision avoidance. Borenstein, J., Koren, and Y suggest that "The Vector Field Histogram-Fast Obstacle-Avoidance for Mobile Robots", IEEE Journal of Robotics and Automation, vol. 7, no. 3, June 1991, pp. 278-288. ”Proposed a method to solve the local minimum and cyclic-trap situations using a path monitor that monitors the movement of a mobile robot.

As described above, various research results suggest a method for solving the local minimum problem of the reactive travel control technique. It has a limit to belong.

Accordingly, the present invention is due to the advantages and disadvantages of various driving control techniques, such as the contemplated driving control technique and the reactive driving control technique, the existing driving control techniques that can not be completely replaced in various driving environments by any one driving control technique In order to solve the problem, while a plurality of driving control techniques are registered in the mobile robot, a driving control method suitable for a driving environment, for example, a static environment and a dynamic environment, is selected so that the mobile robot can travel. An object of the present invention is to provide a traveling control method of a mobile robot using GSPN that can adapt to various driving environments, and a mobile robot using the same.

According to an aspect of the present invention, there is provided a traveling control method of a mobile robot, the method comprising: (a) registering a mutually different first travel control technique and a second travel control technique to the mobile robot; (b) registering a traveling control GSPN model modeled by applying a generalized stochastic petri net (GSPN) technique to the first traveling control technique and the second traveling control technique; (c) executing the travel control GSPN model to select one of the first travel control technique and the second travel control technique; (d) controlling driving of the mobile robot according to any one selected from the first travel control technique and the second travel control technique, wherein the travel control GSPN model selects the first travel control technique. A first travel selection transition for igniting a hazard, a second travel selection transition for ignition for the selection of the second travel control technique, a selection for selection of any one of the first travel control technique and the second travel control technique A place; The step (c) may include: (c1) driving a mission input to the mobile robot, (c2) moving a token to the selection place, (c3) the first driving selection transition and the second driving; Evaluating the performance of the first travel control technique and the second travel control technique based on each parameter of the selected transition; and (c4) the first travel control technique and the second travel control based on a result of the performance evaluation. Any one of the techniques is achieved by the traveling control method of the mobile robot, characterized in that it comprises the step of selecting.

Wherein the first travel control technique includes a deliberate travel control technique; The second travel control technique may include a reactive travel control technique.

The deliberate driving control technique includes a trajectory tracking driving control technique; The reactive driving control technique may include a dynamic window approach (DWA) driving control technique.

Wherein the first travel selection transition includes a tracking travel selection transition that ignites to select the Trajectory tracking travel control technique; The second travel selection transition may include a DWA travel selection transition that ignites to select the dynamic window approach traveling control technique.

Here, the step (c3) is a first transition throughput rate for the trajectory tracking driving control technique based on the steady state probability of the tracking driving selection transition and the trajectory tracking driving control technique. Calculating a second transition throughput for the DWA driving control scheme based on the DWA driving selection transition and the steady state probability of the dynamic window approach traveling control technique. Including; In the step (c4), any one of the first transition throughput and the second transition throughput may be selected.

The driving control GSPN model includes a tracking driving completion transition for the trajectory tracking driving control technique, a DWA driving completion transition for the dynamic window approach driving control technique, and the trajectory tracking driving. A tracking driving warning transition for a control technique, a DWA driving warning transition for the dynamic window approach (DWA) driving control technique, a tracking normal place in which driving control according to the trajectory tracking driving control technique is normal, and the DWA (Dynamic Window Approach) DWA normal place in which the driving control according to the driving control technique is in a normal state, a tracking alarm place transitioning according to the ignition of the tracking driving warning transition, and a DWA transitioned according to the ignition of the DWA driving warning transition. It may include an alarm place.

(여기서, λ_i는 상기 트래킹 주행 완료 트랜지션 또는 상기 DWA 주행 완료 트랜지션의 파라미터이고, v_e는 경험 속도이고, dist_opt는 상기 궤적 추종(Trajectory tracking) 주행 제어 기법에 의해 결정되는 예상 주행 경로의 길이이다)에 의해 산출되며; 상기 경험 속도는 수학식

(여기서, t_navi와 dist_{opt_s}는 상기 궤적 추종(Trajectory tracking) 주행 제어 기법과 상기 DWA(Dynamic Window Approach) 주행 제어 기법이 단독으로 시뮬레이션 되었을 때의 주행 경로 및 주행 시간으로 다수 회의 시뮬레이션을 통한 평균값이다)에 의해 산출될 수 있다.In addition, the initial values of the parameters of the tracking travel complete transition and the DWA travel complete transition may be expressed by the following equation.

(Where λ _i is a parameter of the tracking travel complete transition or the DWA travel complete transition, v _e is an experience speed, and dist _opt is the length of an expected travel path determined by the Trajectory tracking travel control technique) Is calculated; The experience rate is the equation

Here, t _navi and dist _{opt_s} are the average values obtained from the multiple simulations, which are driving paths and driving times when the trajectory tracking driving control technique and the dynamic window approach driving control technique are simulated alone. Can be calculated by

In the step (c), when the DWA driving warning transition ignites while driving is controlled according to the dynamic window approach (DWA) driving control technique, the trajectory tracking driving control technique is selected. Becoming a step; (c6) if the tracking driving warning transition ignites in a driving control state according to the trajectory tracking driving control technique, the dynamic window approach (DWA) driving control technique may be further selected. .

The tracking driving warning transition ignites when the number of rerouting is greater than a predetermined first transition threshold during a preset reference time in a driving control state according to the Trajectory tracking driving control technique. The DWA driving warning transition may be ignited when the number of times that the speed of the mobile robot falls below a preset reference speed in a predetermined period in a state of driving control according to the dynamic window approach (DWA) driving control technique is greater than a preset second threshold. Can be.

At this time, the first transition threshold and the second transition threshold is expressed by the equation

Where N _f is the first transition threshold or the second transition threshold, Pr {Pi} is the token probability of the tracking alarm place or the DWA alarm place, and the F _p {} function is cumulative with frequency λ _j Poisson distribution function, λ _j is a parameter of a tracking driving warning transition or a DWA driving warning transition).

Then, (e) when the driving task is completed, the parameters of each transition of the driving control GSPN model may be updated according to a Bayesian update rule having a reference prior.

Here, the parameter of the transition of the driving control GSPN model is updated according to the Poisson model of the Bayesian update rule when counting is the measurement target; The parameter of the transition of the driving control GSPN model may be updated according to the exponential model of the Bayesian update rule when time is a measurement target.

(여기서, T_fail은 상기 주행 실패 시간이고, v_fail은 기 설정된 실패 속도이다)에 의해 산출될 수 있다.
Further, when the driving task completion time of the mobile robot exceeds the driving failure time, it is determined that the driving robot has failed to travel; The driving failure time is equation

(Where T _fail is the driving failure time and v _fail is a predetermined failure speed).

Through the above configuration, according to the present invention, a traveling control technique suitable for a driving environment, for example, a static environment and a dynamic environment, may be selected when a plurality of traveling control techniques are registered in the mobile robot, and thus the mobile robot may travel. The present invention provides a travel control method of a mobile robot using GSPN and a mobile robot using the same, which can adapt to various driving environments and improve driving performance.

1 is a diagram illustrating a traveling control GSPN model of a traveling control method of a mobile robot according to the present invention;
2 to 8 are diagrams for explaining a simulation result for the driving control method of the mobile robot according to the present invention.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

In the traveling control method of the mobile robot according to the present invention, mutually different first travel control techniques and second travel control techniques are registered in the mobile robot. In addition, the travel control GSPN model is registered in the mobile robot for selecting the first travel control technique and the second travel control technique.

Here, the driving control GSPN model is modeled by applying the Generalized Stochastic Petri Net (GSPN) technique to the first driving control technique and the second driving control technique. That is, the driving control GSPN model is modeled to select one of the first driving control technique and the second driving control technique according to the static driving environment or the dynamic driving environment.

As such, when the mobile robot travels in a state in which the first travel control technique, the second travel control technique, and the travel control GSPN model are registered in the mobile robot, the first travel control technique and the first travel control technique and the first travel control technique are performed according to the execution of the travel control GSPN model. One of two travel control techniques is selected, and the travel of the mobile robot is controlled according to the selected travel control technique.

In the traveling control method of the mobile robot according to the present invention, the driving control technique registered in the mobile robot includes a deliberate driving control technique and a reactive driving control technique. That is, the driving control method of the mobile robot according to the present invention uses a deliberate driving control technique showing a high performance in a static environment and a reactive driving control technique showing a high performance in a dynamic environment. In this regard, complementary advantages can be exerted.

Here, the trajectory tracking driving control technique is applied as the deliberate driving control technique, and the dynamic window approach driving control technique is applied as the reactive driving control technique. .

In addition, as a trajectory tracking driving control technique, Kanayama, Y. Kimura, Y. Miyazaki, F. Noguchi, T., "A stable tracking control method for an autonomous mobile robot", In Proceedings of IEEE International The Trajectory tracking motion controller disclosed in Conference on Robotics & Automation, Cincinnati, OH, USA, May 13-18, 1990. is applied.DWA (Dynamic Window Approach) driving control technique . Fox, W. Burgard, S. Thrun, "The dynamic window approach to collision avoidance", IEEE Robotics and Automation Magazine, vol. 4, no. 1, pp. 23-33, 1997. The application of the DWA motion controller disclosed in the example.

The trajectory tracking driving control technique according to the present invention may include a path planner, a trajectory generator, and a trajectory tracking controller.

The route planner generates a route to a target point on an environment map in the form of a grid map. At this time, the route generated by the route planner includes only location information on the grid map. Here, in the present invention, a gradient path planner is applied as the path planner, for example, and thus a path without collision can be generated on the grid map. For a gradient path planner, the paper `` Konolige, K, "A gradient method for realtime robot control", In Proceedings of IEEE International Conference on Intelligent Robots and Systems, Takamatsu, Japan, October 30, November 5, 2000 Is specifically disclosed, the detailed description thereof will be omitted.

The trajectory generator generates a smooth curve path using the path generated by the path planner. This creates a smooth curve path that satisfies non-holonomic constraints. In the present invention, a cubic B-spline courve technique using a bubble-band algorithm is applied to the trajectory generator as an example. The Bubble-Bands algorithm is specifically described in Sean Quinlan, "Real-time Modification of Collision-free paths", Ph.D Thesis, 1994. .

On the other hand, DWA (Dynamic Window Approach) driving control technique shows a high performance in a dynamic driving environment, as described above, by dividing the path generated by the path planner to the way-point Create This waypoint generation algorithm reduces the risk of local-minimum problems.

In the present invention, the driving control GSPN model for selecting one of a plurality of driving control techniques registered in the mobile robot, that is, a trajectory tracking driving control technique and a dynamic window approach traveling control technique as described above, Generalized Stochastic Petri Net (GSPN) is applied to the Trajectory Tracking and DWA (Dynamic Window Approach) driving controls.

1 is a diagram illustrating a traveling control GSPN model of a traveling control method of a mobile robot according to the present invention, and detailed descriptions of a place and a transition constituting the traveling control GSPN model are given in [Table 1]. As shown.

[Table 1]

Referring to FIG. 1 and Table 1, the travel control GSPN model includes 10 places, 13 timed transitions, and 4 imide transitions.

Place is a ready-to-place (P0) indicating the readiness of the mobile robot, a trajectory tracking (Trajectory tracking) driving control technique, and a selection place (P1) and trajectory tracking (DWA) driving control technique. tracking) The tracking selection place (P2) with the driving control selected by the driving control technique, the DWA selection place (P3) with the driving control selected according to the dynamic window approach (DWA) driving control technique, and the trajectory tracking driving control technique. Driving according to the tracking normal place (P4), DWA (Dynamic Window Approach) driving control technique in which the driving control is normal, driving according to the DWA normal place (P6) and trajectory tracking driving control technique in which the driving control is normal Tracking alarm place (P5) where control is in alarm state, DWA alarm where driving control is in alarm state according to Dynamic Window Approach (DWA) driving control technique Race (P7), comprises a total running this place is finished running success to success (P8), and the whole drive has ended in failure drive fails Place (P9).

In addition, the timed transition is a tracking driving selection transition (T1), a dynamic window approach (DWA), which ignites to select a driving start transition (T0) that fires at the start of a driving mission, a trajectory tracking driving control technique. DWA driving warning transitions for DWA driving selection transitions (T2), trajectory tracking driving control techniques for selecting control techniques, driving warning transitions (T5), and dynamic window approach (DWA) driving control techniques. (T7), tracking driving return transition (T6) for trajectory tracking driving control technique, DWA driving return transition (T8), trajectory tracking driving control for Dynamic Window Approach (DWA) driving control technique Tracking driving return completion transition (Ta6) for the technique, DWA driving recovery completion transition (Ta8) and trajectory tracking (Tr) for the Dynamic Window Approach (DWA) driving control technique ajectory tracking) Tracking driving completion transition (T9) for driving control technique, DWA driving completion transition (T10) for trajectory control technique (DWA) and tracking driving failure transition for trajectory tracking driving control technique (T11), and a DWA driving failure transition (T12) for a dynamic window approach (DWA) driving control technique. Here, the tracking travel selection transition is a kind of first travel selection transition, and the DWA travel selection transition is a kind of second travel selection transition.

The transition transition is a trajectory tracking driving control technique, and when the tracking driving warning transition T5 ignites during driving, the immediate transition is directly switched to a dynamic window approach driving control technique (DWA). When the DWA driving warning transition (T7) ignites during driving with the DWA (Dynamic Window Approach) driving control technique, the tracking switching transition (T4) is directly switched to the trajectory tracking driving control technique, and the vehicle is ready for driving after successful driving. The driving success termination transition T13 for switching to the place P0 and the driving failure ending transition T14 for switching to the driving preparation place P0 after the driving failure are included.

Herein, the parameter of the driving start transition T0, that is, the firing rate λ ₀ , is set to a constant, and in the present invention, λ ₀ = 1.0 is set as an example. And, the ignition rate of the remaining transitions are updated after the completion of the driving task, a description thereof will be described later.

And in the travel control GSPN model, performance evaluation of each travel control technique is performed when the token is present in P1, i.e., the selection place P1. In addition, performance evaluation of each driving control technique is performed in consideration of an expected completion time and a failure rate of the driving task.

For example, when the mobile robot moves in a static driving environment, a trajectory tracking driving control technique suitable for moving along an optimal path is selected by the driving control GSPN model. On the other hand, when the mobile robot is faced with a dynamic driving environment, the route planner continuously generates a route without collision, and when the route is changed frequently, the driving control GSPN model recognizes the dynamic environment. Tracking alarm place P5 is activated. As a result, the driving control GSPN model activates the Dynamic Window Approach (DWA) driving control technique to enable the mobile robot to travel in a dynamic environment. In addition, when the mobile robot is in a local-trap situation, the DWA driving warning transition T7 is ignited. 1 is a diagram illustrating control logic of the driving control GSPN model.

On the other hand, the travel control GSPN model according to the present invention calculates the parameters of the tracking travel completion transition T9 and the DWA travel completion transition T10, that is, the initial values of the ignition rates λ ₉ and λ ₁₀ when the travel pregnancies are given to the mobile robot. do. In the present invention, it is assumed that the initial values of the ignition rates λ ₉ and λ ₁₀ are calculated using an empirical velocity based on a plurality of simulation results. This is calculated through [Equation 1].

[Equation 1]

Here, λ _i means the ignition rates λ ₉ and λ ₁₀ , v _e is the experiential speed, and dist _opt is the length of the expected driving route determined by the route planner of the trajectory tracking driving control technique.

Then, the experience speed is calculated through [Equation 2]. Here, the experience speed is calculated by applying a trajectory tracking driving control technique and a dynamic window approach driving control technique to a driving simulation alone, and calculating the average speed of a plurality of simulations.

&Quot; (2) "

Here, t _navi and dist _{opt_s} are driving paths and driving times when the trajectory tracking (Trajectory tracking) driving control method and the DWA (Dynamic Window Approach) driving control method are simulated alone, and are average values of values obtained in a plurality of simulations. .

On the other hand, the driving failure time is calculated through [Equation 3]. In Equation 3, T _fail is a driving failure time, and v _fail is a failure speed. Here, V _fail has a predetermined constant value in consideration of the speed, the driving environment, and the like of the mobile robot, in the present invention, it is _assumed that V _fail is set to 5 cm / s.

&Quot; (3) "

If the total driving task completion time of the mobile robot is longer than the T _fail which is the driving failure time, it is determined as a driving failure, and the failed driving completion time is recorded as T _fail again.

In the present invention, since a trajectory tracking driving control technique and the DWA driving control technique may be selectively used among driving missions, a trajectory tracking driving control technique and a dynamic window approach traveling technique are used. ), The failure rate of the driving control technique, that is, the parameters λ ₁₁ and λ ₁₂ of the tracking driving failure transition T11 and the DWA driving failure transition T12 determine the ratio of the driving time of each technique and the total driving mission completion time determined by the driving failure. It is calculated by standardization.

For example, it is 200 seconds of the completion time of the entire driving mission determined as driving failure, and for 200 seconds, the robot moves 150 seconds using the dynamic window approach (DWA) driving control technique and 50 seconds using the trajectory tracking driving control technique. When traveling, it becomes (lambda) ₁₁ = 1 / (150/200) and becomes (lambda) ₁₂ = 1 / (50/200).

On the other hand, the traveling control method of the mobile robot according to the present invention uses the calculation of throughput as a method of predicting the performance of the trajectory tracking driving control technique and the dynamic window approach (DWA) driving control technique. Here, the concept of calculating throughput for performance prediction is described in the paper `` J. Wang, "Timed Petri Nets Theory and Application", Norwell, MA: Kluwer, 1998., the detailed description of which is omitted.

Further, in the present invention, as a method of predicting the performance of the trajectory tracking driving control technique and the dynamic window approach driving control technique, the tracking driving completion transition T9 corresponding to the driving success and the DWA driving completion transition ( The ignition rate of T10), i.e., λ ₉ and λ ₁₀ .

Here, the throughputs of the trajectory tracking driving control technique and the dynamic window approach driving control technique are the firing rates of the tracking driving completion transition T9 and the DWA driving completion transition T10 and the respective driving control techniques, respectively. For example, it is calculated by multiplying the steady-state probability of.

The physical meaning of the throughput calculated in this way means the success frequency of the driving task. Therefore, when the throughput of the trajectory tracking driving control technique is higher than that of the DWA driving control technique, the estimated driving time according to the trajectory tracking driving control technique is DWA (Dynamic Window Approach). It becomes shorter than the expected travel time according to the travel control technique, and as a result, a trajectory tracking travel control technique is selected. In the opposite case, the Dynamic Window Approach (DWA) driving control technique will be chosen.

Here, in the present invention, the firing rate of the tracking travel selection transition T1 and the DWA travel selection transition T2 indicating the selection of the trajectory tracking travel control method and the dynamic window approach travel control method, λ ₁ and λ ₂ , ignition rate λ ₁ and λ of tracking drive selection transition (T1) and DWA drive selection transition (T2) for performance evaluation of trajectory tracking travel control and dynamic window approach travel control Throughput is calculated using ₂ . At this time, the ignition rates of the above-described tracking travel completion transition T9 and the DWA travel completion transition T10, that is, λ ₉ and λ _10, are the ignition rates λ ₁ of the tracking travel selection transition T1 and the DWA travel selection transition T2. since reflected in the λ _2, it is selected in the present invention, a running control technique by monitoring the firing rate, λ ₁ and λ ₂ of the tracking driving selected transition (T1) and the driving DWA selected transition (T2).

On the other hand, in the traveling control method of the mobile robot according to the present invention for the selection of the trajectory tracking (Trajectory tracking) driving control method and the DWA (Dynamic Window Approach) driving control technique, the trajectory tracking (Trajectory tracking) driving control technique and DWA (Dynamic Window) Approach The internal state of each driving control technique is evaluated. Here, the internal states of the trajectory tracking driving control technique and the dynamic window approach driving control technique include a driving warning state and a normal driving state, respectively.

The driving warning state of the trajectory tracking driving control technique refers to a state in which a path generation occurs frequently due to an obstacle or a path closure, and the normal driving state is a driving warning state of the trajectory tracking driving control technique. It means no state. That is, the driving warning state of the trajectory tracking driving control technique means a state in which the tracking driving warning transition T5 ignites.

On the other hand, the driving warning state of the DWA (Dynamic Window Approach) driving control technique means a state in which the mobile robot cannot proceed due to a local-trap situation, etc., and the normal driving state is a DWA (Dynamic Window Approach) driving. It means a state that is not a driving warning state of the control technique. That is, the driving warning state of the DWA driving control technique refers to a state in which the DWA driving warning transition T7 ignites.

In the present invention, in order to evaluate the internal state of the trajectory tracking driving control technique, the number of route resets (Np) for a preset reference time (Tc) in a state of driving control according to the trajectory tracking driving control technique. Count) On the other hand, in order to evaluate the internal state of the dynamic window approach (DWA) driving control technique, the number Nl of the speed of the mobile robot falling below a predetermined reference speed is counted at a predetermined cycle.

In the present invention, a method of calculating a transition probability is modeled using a Poisson model. Here, the Poisson model is approximated with an exponential density function that can be used for timed transition of the GSPN model through Poisson limit theorem.

Here, in the driving control method of the mobile robot according to the present invention, since one token is given to each subsystem, the token probability of each subsystem corresponds to the probability of the internal state of each subsystem.

The ignition of the tracking driving warning transition T5 and the DWA driving warning transition T7 is N to Poisson (λ _i T _c ). That is, the ignition of the tracking driving warning transition T5 and the DWA driving warning transition T7 means that N follows a Poisson probability model having a parameter λ _i T _c . In addition, when the number of rerouting (Np) of the trajectory tracking driving control technique and the number (Nl) of the dynamic window approach (DWA) driving control technique are larger than their transition threshold Nf, the tracking driving warning transition (T5). ) And DWA driving warning transition (T7) will be ignited.

Here, the transition threshold Nf and its probability according to the present invention are calculated through Equation 4 and Equation 5, respectively.

&Quot; (4) "

[Equation 5]

Pr {Pi} is the token probability of the state Pi. For example, the token probability of tracking alarm place P5. And the Fp {} function is a cumulative Poisson distribution function with frequency λ _j . For example, in a state where a driving warning of the trajectory tracking driving control technique occurs, it is λ ₅ . In this case, the Fp {} function of Equation 4 may be more simply calculated by integrating the method using Equation 5 a plurality of times.

Here, Np and Nl values are proportional to the probability of the normal driving state of each of the trajectory tracking driving control technique and the dynamic window approach driving control technique. That is, when the driving environment of the mobile robot is a dynamic environment, the calculation result of the Np value is low because the probability of the normal driving state of the trajectory tracking driving control technique is low. This result means that the trajectory tracking driving control technique can be converted to the dynamic window approach driving control technique in a dynamic environment.

Meanwhile, in the traveling control method of the mobile robot according to the present invention, the tracking travel return transition T6 and the DWA travel return transition T8 are respectively returned from the time at which the tracking travel warning transition T5 and the DWA travel warning transition are ignited. Fires later Here, the firing event of the tracking travel return transition T6 and the DWA travel return transition T8 follows an exponential probability density function with t to Exp (−λ _i ), that is, the return time t is a parameter λ _i .

Here, the return time is proportional to the alarm state, that is, the probability of tracking alarm place P5 and DWA alarm place P7 and λ _j respectively. This means that when the tracking driving warning transition T5 or the DWA driving warning transition T7 ignites frequently, the return time increases.

Such a return time can be calculated using [Equation 6].

&Quot; (6) "

Here, in [Equation 6], t _f is the return time, λ _j is the firing rate (λ ₅ , λ ₇ ) of the tracking driving warning transition T5 and the DWA driving warning transition T7, and t _a is the running of the mobile robot. The time required to adjust the direction of heading.

In addition, the F _e {} function means a cumulative density function of an exponential probability density having λ _j , and is calculated through Equation 7 below. Here, the cumulative density and F _e {} value is represented by Equation 7, where λ _j is the ignition rate λ ₅ , of the tracking driving warning transition T5 and the DWA driving warning transition T7. λ ₇ ) and the alarm state, that is, the probability of tracking alarm place P5 or DWA alarm place P7.

[Equation 7]

Meanwhile, in the driving control method of the mobile robot according to the present invention, when the driving task is completed, the parameters of each transition of the driving control GSPN model are updated. In the present invention, a Bayesian update rule having a reference prior may be used. For example, update according to rule). Here, the basic concept of the Bayesian update rule is disclosed in the paper "Athanasios Papoulis," Probability, Random Variables, and Stochastic Processes ", McGraw-Hill Companies, 3rd edition, 1991.

Here, in the traveling control method of the mobile robot according to the present invention, among the transition parameters, λ ₁ , λ ₂ , λ ₅ , and λ ₇ are updated through a Poisson model of the Bayesian update rule, and among the transition parameters [lambda] ₆ , [lambda] ₈ , [lambda] ₉ , [lambda] ₁₀ , [lambda] ₁₁ , and [lambda] ₁₂ are updated using an exponential model of the Bayesian update rule. The update through the Poisson model is performed through [Equation 8] and [Equation 9], the update through the exponential model is performed through [Equation 10] and [Equation 11].

[Equation 8]

&Quot; (9) "

&Quot; (10) "

&Quot; (11) "

Equation (8) represents the posterior of a Poisson model with a reference prior considering each driving result. Here, N _c represents the number of times measured as the tracking driving warning transition T5 or the DWA driving warning transition T7. For example, as described above, Nf and Nl mean the number of measurements.

In addition, α in Equation 9 is an existing Nc value calculated through previous experience, and β in Equation 9 represents the number of driving. In addition, [alpha] of [Equation 11] means the number of driving, and [beta] of [Equation 11] means the existing λ value calculated through previous experience. And? C represents a return time for returning to the normal driving state from the driving warning state of the currently completed driving task.

More specifically, when updating λ ₁ and λ ₂ by using Equations 8 and 9, the number of trajectory tracking driving control methods are selected and DWA (DWA) until the driving task is completed. Dynamic Window Approach) The number of times the driving control technique is selected becomes N value of [Equation 9]. Similarly, in the update of λ ₅ and λ _7, the number of occurrences of the tracking driving warning transition and the DWA driving warning transition becomes N value.

In addition, when λ ₆ and λ ₈ are updated using Equations 10 and 11, the reciprocal of the mean of the time periods in which the tracking driving warning state and the DWA driving warning state persisted, respectively. It is calculated as the value of λ _{c of} . That is, the lambda _c value in [Equation 10] is determined by the inverse of the time.

In the above manner, the transition parameters λ ₁ , λ ₂ , λ ₅ , λ ₇ are updated through Equation 9, and the transition parameters λ ₆ , λ ₈ , λ ₉ , λ ₁₀ , λ ₁₁ , and λ ₁₂ are It is updated through Equation 11.

Here, when the driving environment of the mobile robot is fixed to either a dynamic or static environment, each transition parameter converges to a value suitable for the corresponding environment by Equation 9 and Equation 11, as a result. The driving control method suitable for the driving environment of the mobile robot is selected. This is because the β value of [Equation 9] and [alpha] of [Equation 11] corresponding to previous driving experiences continuously increase through repetition of driving.

At this time, when the β value of [Equation 9] and the α value of [Equation 11] are limited to a predetermined value or less, for example, if the β value of [Equation 9] is limited to 100 or less, 100 times The previous driving experience is eliminated so that the convergence of the transition parameters is cut off to some extent so that it can be set to actively cope with changes in the driving environment.

Hereinafter, the driving simulation result of the mobile robot using the traveling control method of the mobile robot according to the present invention will be described. Here, the simulation of the mobile robot uses a player / stage tool that evaluates the performance of the driving control technique in a standard state.

2 is a diagram illustrating a map of a driving environment applied to a simulation. The width of the simulation environment was 43m and the length was set to 22m. In addition, the speed of the moving obstacle was randomly set within the range of 0.8 to 1.4 m / s with reference to the average walking speed of the person, and the diameter of the obstacle was set to 0.4 m.

In addition, the diameter of the mobile robot applied to the simulation was set to 0.5m, and the maximum translational velocity of the DWA driving control technique was set to 0.5m / s. In addition, the experience speed is calculated as the average of 20 times the G8 position at the G1 position in the absence of a moving obstacle, and the experience speed of the DWA (Dynamic Window Approach) driving control technique is 0.28 m / s. The experiential speed of the trajectory tracking driving control technique was set to 0.35 m / s. Then, the initial values of the parameters of each transition of the travel control GSPN model are set to λ ₁ to λ ₈ as 0.5 and λ ₁₁ and λ ₁₂ as 0.01.

Simulation in a static driving environment sets the mobile robot to start at G7 and drive to destinations G1, G2, G3 and G7 sequentially without dynamic obstacles. The simulation in a dynamic environment was carried out to evaluate the change in performance in the dynamic environment. Here, the simulation in the dynamic environment was set to reciprocate between two destinations G1 and G8 in the presence of 20 moving obstacles. 28 repetitions were performed for each simulation.

3 is a graph showing the results of simulation in a static environment and a dynamic environment, and Table 2 shows the simulation results.

[Table 2]

Referring to FIG. 3 and [Table 2], in the static driving environment, the average driving time of the trajectory tracking driving control technique was about 38% shorter than the dynamic window approach driving control technique. This means that the Trajectory tracking driving control technique is suitable for the static driving environment. However, in the dynamic driving environment, the performance of the tracking (tracking) tracking control method is lower than that of the dynamic window approach (DWA) driving control method.

Here, the performance of the Dynamic Window Approach (DWA) driving control technique is shown to decrease when entering a narrow entrance, as shown in FIG. In addition, driving failure in the Dynamic Window Approach (DWA) driving control technique occurred when the mobile robot encountered a local minimum situation.

In a static driving environment, driving failure of a trajectory tracking driving control technique is sometimes caused by an error in path generation around an entrance and exit due to a localization error. In addition, the performance of the trajectory tracking driving control technique is reduced by the path regeneration according to the moving obstacle.

Simulation results in a static and dynamic driving environment clearly show that the control techniques for driving show different performances when applied individually.

On the other hand, the simulation in a mixed environment of the static driving environment and the dynamic driving environment for evaluating the performance of the driving control method according to the present invention is a static driving in the state where the 20 mobile obstacles exist in the driving environment Driving to the same destination as the environmental simulation was performed. 5 is a graph illustrating simulation results in the driving environment as described above, and the simulation results are shown in [Table 3].

[Table 3]

Referring to FIG. 5 and [Table 3], the driving control method to which the driving control GSPN model is applied according to the present invention is a trajectory tracking driving control technique and a dynamic window approach driving control technique applied separately. You can see that it shows better performance than that. That is, the average driving time was 43% shorter than the DWA (Dynamic Window Approach) driving control technique alone, and 47% shorter than the Trajectory tracking driving control technique.

6 is a diagram illustrating a driving trajectory of the traveling control method of the mobile robot according to the present invention. FIG. 6 (a) is a view showing the entire driving trajectory of the mobile robot from the G1 position to the G2 position and the traveling control technique selected from the driving trajectories, and FIG. 6 (b) is the position ⓐ of FIG. 6 (a). FIG. 6C is a diagram illustrating a driving environment at the ⓑ position of FIG. 6A. The red dots in FIGS. 6B and 6C are moving obstacles detected by sensors of the mobile robot.

Referring to FIG. 6, the mobile robot travels at the starting point G1 position through a dynamic window approach (DWA) driving control technique. This is because the throughput of the dynamic window approach traveling control technique calculated by the parameter initial value of the transition of the traveling control GSPN model is larger than the throughput of the trajectory tracking traveling control technique. .

In addition, when the mobile robot reaches the point ⓐ shown in FIG. 6A in the process of driving with the DWA (Dynamic Window Approach) driving control technique, as shown in FIG. 6B, the moving obstacle It is surrounded by. At this time, the mobile robot enters a trapped situation, and the DWA driving warning transition T7 ignites.

Accordingly, the driving control GSPN model converts the DWA (Dynamic Window Approach) driving control technique to a trajectory tracking driving control technique, and track trajectory tracking from point ⓐ to point ⓑ of FIG. 6 (a). ) The mobile robot travels according to the travel control technique.

As shown in (c) of FIG. 6, only a few moving obstacles exist at point ⓑ, and at this time, the tracking driving warning transition of the driving control GSPN model is ignited, so as to use the DWA (Dynamic Window Approach) driving control technique. Is switched.

In addition, when the mobile robot travels by the DWA (Dynamic Window Approach) driving control technique from the point ⓑ of FIG. 6 (a), and reaches the point ⓒ, the lobby area, the DWA driving warning transition of the driving control GSPN model is generated. The vehicle is ignited and then switched to a trajectory tracking driving control technique to control driving to the target point G2.

On the other hand, Figure 7 shows a change in the parameters of the tracking driving selection transition and the DWA driving selection transition, that is, the ignition rates λ ₁ and λ ₂ in the simulation in the driving environment in which the static driving environment and the static driving environment and the dynamic driving environment are mixed. Is a graph. Here, the ignition rates λ ₁ and λ ₂ are the performances of the trajectory tracking driving control technique and the dynamic window approach driving control technique, respectively, as described above. The ignition rates λ ₁ and λ ₂ are updated in accordance with the above-described update rule after completion of the traveling task.

After as shown in FIG.'S 7 (a), eight completed the running task in a static environment can be found to λ ₁ is very jyeoteum greater than λ _2. This indicates that the trajectory tracking driving control technique is applied to the driving control in the static driving environment, and the driving control GSPN model is adapted to the static environment through the update of λ ₁ and λ ₂ .

On the other hand, as shown in (b) of FIG. 7, in the driving environment in which the static driving environment and the dynamic driving environment are mixed, λ ₁ is slightly larger than λ ₂ , but the difference is not large. . This means that the trajectory tracking driving control technique and the dynamic window approach driving control technique are selectively applied to the driving control in a mixed driving environment.

Meanwhile, FIG. 8 is a graph showing a result of 55 simulations in a dynamic driving environment according to the driving control technique of the mobile robot according to the present invention. Here, the dynamic environment is set to an environment that travels between the G1 point and the G8 point shown in FIG. 6, and simulates a reciprocating run.

FIG. 8A is a graph illustrating a change in throughput, that is, an update result of a trajectory tracking driving control technique and a dynamic window approach driving control technique calculated through the simulation. In the dynamic driving environment, the throughput of the trajectory tracking driving control technique decreases and the throughput of the dynamic window approach driving control technique increases. This means that in a dynamic driving environment, the tracking (Trajectory tracking) driving control technique is not appropriate due to moving obstacles, and as a result, the dynamic window approach (DWA) driving control technique is mainly selected by the driving control GSPN model.

8 (b) is a graph showing the change in the total running time measured through each simulation. As shown in (b) of FIG. 8, it can be seen that the running time gradually decreases as the simulation is repeated. This is because the driving control GSPN model according to the present invention reflects the previous driving experience in updating the parameters of the transition, thereby improving the overall performance of the driving control method according to the present invention.

Although several embodiments of the present invention have been shown and described, those skilled in the art will appreciate that various modifications may be made without departing from the principles and spirit of the invention . The scope of the alias will be defined by the appended claims and their equivalents.

Claims (14)

In the traveling control method of a mobile robot,
(a) registering a mutually different first travel control technique and a second travel control technique to the mobile robot;
(b) registering a traveling control GSPN model modeled by applying a generalized stochastic petri net (GSPN) technique to the first traveling control technique and the second traveling control technique;
(c) executing the travel control GSPN model to select one of the first travel control technique and the second travel control technique;
(d) controlling driving of the mobile robot according to any one selected from the first travel control technique and the second travel control technique;
The driving control GSPN model may include a first driving selection transition for igniting the selection of the first driving control technique, a second driving selection transition for igniting the selection of the second driving control technique, and the first driving control technique. And a selection place for selecting any one of the second travel control technique;
The step (c)
(c1) inputting a driving task to the mobile robot;
(c2) the token is moved to the selection place;
(c3) evaluating the performance of the first travel control technique and the second travel control technique based on the parameters of each of the first travel selection transition and the second travel selection transition;
and (c4) selecting one of the first travel control technique and the second travel control technique according to the performance evaluation result.
The method of claim 1,
The first travel control technique includes a deliberate travel control technique;
The second travel control method includes a travel control method of the mobile robot, characterized in that the reactive travel control method. The method of claim 2,
The deliberate driving control technique includes a trajectory tracking driving control technique;
The reactive driving control technique includes a dynamic window approach (DWA) driving control technique. The method of claim 3,
The first travel selection transition includes a tracking travel selection transition that ignites for selection of the Trajectory tracking travel control technique;
And the second travel selection transition includes a DWA travel selection transition that ignites to select the dynamic window approach traveling control technique.
5. The method of claim 4,
The step (c3)
Calculating a first transition throughput for the trajectory tracking driving control technique based on the steady state probability of the tracking driving selection transition and the trajectory tracking driving control technique;
Calculating a second transition throughput for the Dynamic Window Approach (DWA) driving control technique based on the steady state probability of the DWA driving selection transition and the Dynamic Window Approach (DWA) driving control technique;
In the step (c4), any one of the first transition throughput (Throughput) and the second transition throughput (Throughput) is selected one of the larger travel control method of the mobile robot. The method of claim 5,
The driving control GSPN model is a tracking driving completion transition for the trajectory tracking driving control technique, a DWA driving completion transition for the DWA driving control technique, and the trajectory tracking driving control technique. A tracking driving warning transition for the DWA driving warning transition for the dynamic window approach (DWA) driving control technique, a tracking normal place in which driving control according to the trajectory tracking driving control technique is normal, and the DWA Window Approach) DWA normal place in which the driving control according to the driving control technique is in a normal state, a tracking alarm place transitioning according to the ignition of the tracking driving warning transition, and a DWA alarm place transitioning according to the ignition of the DWA driving warning transition. Of the mobile robot comprising a Driving control method. The method according to claim 6,
The initial values of the parameters of the tracking travel completion transition and the DWA travel completion transition are represented by equations.

(Where λ _i is a parameter of the tracking travel complete transition or the DWA travel complete transition, v _e is an experience speed, and dist _opt is the length of an expected travel path determined by the Trajectory tracking travel control technique) Is calculated;
The experience speed is the equation

Here, t _navi and dist _{opt_s} are the average values obtained from the multiple simulations, which are driving paths and driving times when the trajectory tracking driving control technique and the dynamic window approach driving control technique are simulated alone. Traveling control method of a mobile robot, characterized in that it is calculated by. The method according to claim 6,
The step (c)
(c5) selecting the trajectory tracking driving control method when the DWA driving warning transition ignites in a driving control state according to the dynamic window approach driving control technique;
(c6) when the tracking driving warning transition is ignited in a state of driving control according to the trajectory tracking driving control technique, the dynamic window approach (DWA) driving control technique is further selected. Travel control method of the mobile robot. The method according to claim 6,
The tracking driving warning transition ignites when the number of rerouting is greater than a predetermined first transition threshold during a preset reference time in a driving control state according to the trajectory tracking driving control technique;
The DWA driving warning transition ignites when the number of times that the speed of the mobile robot falls below a preset reference speed in a predetermined period while driving is controlled according to the dynamic window approach (DWA) driving control technique is greater than a preset second threshold. Travel control method of a mobile robot, characterized in that. 10. The method of claim 9,
Wherein the first transition threshold and the second transition threshold are

Where N _f is the first transition threshold or the second transition threshold, Pr {Pi} is the token probability of the tracking alarm place or the DWA alarm place, and the F _p {} function is cumulative with frequency λ _j A Poisson distribution function, and λ _j is a parameter of a tracking travel warning transition or a DWA travel warning transition). The method according to any one of claims 7 to 10,
(e) When the driving task is completed, the driving control of the mobile robot, wherein the parameters of each transition of the driving control GSPN model are updated according to a Bayesian update rule having a reference prior. Way. The method of claim 11,
The parameter of the transition of the traveling control GSPN model is updated according to the Poisson model of the Bayesian update rule when counting is the measurement target;
And the parameter of the transition of the traveling control GSPN model is updated according to the exponential model of the Bayesian update rule when time is the measurement target. The method of claim 11,
When the driving task completion time of the mobile robot exceeds the driving failure time, it is determined that the driving robot has failed to travel;
The driving failure time is equation

Wherein T _fail is the travel failure time and v _fail is a predetermined failure speed. A mobile robot characterized in that the driving control according to the driving control method according to claim 11.

Images