WO2017153896A1

WO2017153896A1 - Motor-driven autonomous robot with obstacle anticipation

Info

Publication number: WO2017153896A1
Application number: PCT/IB2017/051301
Authority: WO
Inventors: Sébastien Bonnet
Original assignee: Effidence
Priority date: 2016-03-07
Filing date: 2017-03-06
Publication date: 2017-09-14
Also published as: FR3048517A1; FR3048517B1

Abstract

The invention relates to a guidance method for an autonomous robot, with obstacle avoidance, comprising steps consisting in: establishing a setpoint for the advancement of the robot; determining (51) an effective anticipation distance of the robot; receiving (52) data about possible obstacles within the anticipation distance on the initial path, in the event that the robot may hit an obstacle within the effective anticipation distance after the speed setpoint has been applied thereto; and determining (54) a plurality of avoidance paths, grading the avoidance paths and selecting the path with the best grade, the effective anticipation distance of the robot corresponding to the minimum observation distance in front of the vehicle, for the avoidance of obstacles.

Description

AUTONOMOUS MOTORIZED ROBOT WITH OBSTACLE ANTICIPATION

TECHNICAL FIELD OF THE INVENTION

The present invention relates to an autonomous robot. It relates more particularly to a robot capable of anticipating the presence of obstacles on its path.

STATE OF THE PRIOR ART

[0002] Powered autonomous robots are now well known and used in many fields such as logistics, agriculture, industrial production, etc. A large part of the robots used is intended to transport loads over short distances. To ensure their movement autonomously, the robots must include locating and guiding means. They must be able to detect possible obstacles in front of them, and provide means to avoid collisions with these obstacles. Many forms of implementation exist today, with sensors and / or cameras arranged at locations to implement the locating functions, while allowing the robot to perform its basic mission, load transport. There are several conflicting requirements between providing the robot with the equipment enabling it to be autonomous, the fact of keeping loading spaces easily accessible and the protection of the means of locating against shocks and other hazards associated with intensive use. One of the most used solutions to date is to provide a mast or other arm system for positioning one or more sensors in height. The elevated position relative to the robot and the load handling area protects the sensors. Other examples of solutions are presented below.

US2015314443 discloses a visual obstacle detection method. The method includes receiving image information from an environment sensed by at least one visual sensor to obtain a three-dimensional space and a two-dimensional plane. The method also includes reconstructing the coordinates and geometric positions of the obstacles. Finally, the method allows the drawing of a path for automatic movement of the mobile robot based on the detected obstacles.

US2013223673 describes a method for identifying objects including a warehouse. [0005] US2015066201 discloses a system for performing multi-dimensional real-time inspection of multiple objects. The document provides a plurality of real-time vision sensors and other sensors are positioned above the conveyor belt to observe and record real-time data of packets and boxes moving on the web.

US2010034422 discloses a method of tracking objects in an environment comprising acquisition sensors related to the environment.

US2007058838 discloses an object detection apparatus comprising distance sensors, each detection of an object in a detection zone provides object information indicating the existence of the object and the image capture. A distance calculation unit calculates the approximate values of the distances between the distance sensors respectively to the object.

EP2124122 discloses the generation of an environment card for a mobile robot. The mobile robot generates a set of three-dimensional position data representing an external environment using measurement information from the distance sensors. From an old environmental map, a new environmental map is generated by the integration of the region and obstacles.

EP2952993 discloses a method for calculating a pixel map of the probability of absence and / or presence of obstacles in the environment of the autonomous robot. The autonomous robot comprises at least one sensor grouped in at least one set of sensors that detect obstacles of similar types.

The solutions presented above involve complex means with several sensors, means for establishing a map of the site on which the robot is called to move, and so on. These solutions are not applicable for environments that change regularly, or even in real time, or natural environments such as fields, or construction site.

The failure to take into account the position data of the obstacles can not only lead to collisions, but may also cause immobilization of the robot if a security mechanism is implemented. For example, a robot that does not know the obstacle in front of it, will eventually be very close to the obstacle, and even if it pointing hard, the corner of the robot may hit the obstacle. The robot then risks a collision or an immobilization by security.

[0012] Another frequent case is that where a robot follows a person. The person walks towards a perpendicular wall and turns 90 ° when it is close to the wall. Even when you're aiming hard, the robot may hit the wall. Even in the case of a "tank type" vehicle that has a pivoting movement on itself, when it is too close to the wall, one of the corners of the robot may hit the wall.

Moreover, there are many situations where a robot is inadvertently too close to an obstacle and sees its automatic security mode is triggered.

For example, when the robot is very close to an obstacle, if a wheel passes through a hole, this has the effect of bringing the robot even more, triggering a safety stop.

When the robot moves in a corridor, for example following a person, if the person runs along the walls, the robot will approach very close without touching them. However, if the driver turns 90 ° to take another corridor, the robot will hang because of its size, especially if it has a tray on top. For example, if the robot rotates, the rear overhang, by lever effect, may touch the wall. A safety shutdown is then likely.

Document US2006 / 0235610 describes a conventional route calculation method between a starting point and a final destination of a user in which the potential routes in which obstacles are located are simply eliminated.

To overcome these disadvantages, the invention provides different technical means.

SUMMARY OF THE INVENTION

First, a first object of the invention is to provide an autonomous motorized robot to anticipate obstacles on its path. Another object of the invention is to provide an obstacle avoidance method without maneuver.

Another object of the invention is to provide a robot with trajectories having the smallest curvatures possible.

Another object of the invention is to provide a robot whose implementation is simple.

To do this, the invention provides a driving method for autonomous robot with obstacle bypass comprising the steps of:

establish a robot advance instruction (for example with a linear speed and a constant or variable direction);

determining an effective anticipation distance of the robot;

receive potential obstacle data within the anticipation distance on the initial trip;

in the case where the robot would hit an obstacle in the actual distance of anticipation after applying the speed instruction;

determining a plurality of bypass paths, noting the bypass paths and selecting the route with the most favorable rating;

the effective anticipation distance of the robot corresponding to the minimum distance of observation in front of the vehicle to avoid obstacles.

According to such an architecture, the robot is able to anticipate a turn early enough to obtain trajectories with the smallest curvatures possible. This makes it possible to maintain a high speed without risk of overturning or losing load. In the presence of a robot type "char" anticipation avoids the denaturation of the soil and reduce the energy required to turn, thus avoiding over-size engines. Finally, such an architecture avoids the lateral sliding phenomenon, which helps to improve the location of the robot.

According to an advantageous embodiment, the effective anticipation distance corresponds to the sum of the theoretical anticipation distance and the steering clearance distance.

According to yet another advantageous embodiment, the theoretical anticipatory distance of the robot corresponds to the minimum theoretical distance to search obstacles in front of the vehicle to be able to avoid it by turning the wheels fully (to the left or the right). For this, it is considered that the robot is theoretically able to steer its wheels instantly. This distance makes it possible to circumvent the obstacle with the maximum deflection from the beginning of the bypass. This distance is therefore relatively short because the maximum deflection is used as soon as the bypass begins. In practice, one can look for a more progressive trajectory, on higher radii of curvature.

Advantageously, the distance to obtain the steering corresponds to the distance traveled by the vehicle according to its speed of movement and the variation of curvature allowed until the desired deflection. In practice, since it is generally desired to avoid immobilizing the robot during the steering time of the wheels (because this would make the driving jerky, even brutal) the robot travels a certain distance between the time of the start of the steering and the one where the desired steering (maximum or no) is obtained. Thus, for a progressive deflection obtained by advancing, the total distance corresponds to the distance to obtain the steering and the remaining distance to circumvent with the steered wheels, possibly then reducing the steering, depending on a new obstacle to cross . Depending on the speed, to ensure the stability of the mobile robot, the radius of curvature can be increased, and thus the reduced turning radius, increasing the effective anticipation distance.

According to an advantageous embodiment, the theoretical distance of anticipation is measured from the remarkable point of the robot, this corresponding distance:

- for a robot type char (with caterpillars), it is its pivot point;

- for a robot with front steering axle and fixed rear, this is the middle of the rear axle;

- for a robot with front and rear axle steering, it is the middle of the two axles.

According to another advantageous embodiment, the robot driving method also comprises the steps of:

for the route chosen, determine an obstacle avoidance envelope;

perform the chosen route taking into account the obstacle avoidance envelope.

According to an advantageous variant, the step of determining an obstacle avoidance envelope comprises the following steps:

determining a first obstacle avoidance envelope;

evaluate, with this first envelope, the possibility of circumvention;

if the test with the first envelope is positive, determine a second larger envelope; evaluate the possibilities of circumvention for the second envelope;

- if the test with the second envelope is positive, use the modified elected path taking into account this obstacle avoidance envelope.

Advantageously, the test is performed for at least one other envelope of greater width.

The invention also provides a driving system for autonomous robot for the implementation of the previously described driving method, comprising a tracking device able to communicate with a driver module provided with a microprocessor and instructions for use. implementation of a setpoint module, an obstacle detection module, an actual anticipation distance determining module of the robot, a bypass path determination module, designed to determine a plurality of bypass paths , note the bypass paths and select the route with the most favorable rating; a scoring module, the actual anticipation distance determining module of the robot being provided to determine the effective anticipation distance of the robot corresponding to the minimum distance of observation in front of the vehicle to avoid obstacles.

[0031] Advantageously, the autonomous robot driving system comprises an avoidance envelope determination module, designed to determine an obstacle avoidance envelope and perform the chosen path taking into account the envelope of the obstacle. obstacle avoidance,

The invention finally provides a motorized autonomous robot, comprising a body mounted on wheels or tracks, also comprising a driving system as previously described.

Advantageously, the tracking device comprises a single sensor, preferably a Lidar, advantageously adapted to emit a rotating beam over an angular range of 360 °.

DESCRIPTION OF THE FIGURES All the details of implementation are given in the description which follows, supplemented by FIGS. 1 to 4, presented solely for purposes of non-limiting examples, and in which:

FIG 1A schematically illustrates a robot moving in a mode to anticipate the obstacles present in front of him;

FIG. 1B schematically illustrates the dimension corresponding to the Instantaneous Center of Rotation (CIR);

FIG. 2 is a functional flowchart illustrating a part of the key steps of the robot driving method according to the invention (for the obstacle anticipation part);

FIG. 3 is a functional flowchart illustrating the following key steps of the robot driving method according to the invention (for the delimiting part of an optimal avoidance envelope);

FIG. 4 is a schematic representation of an exemplary driving system.

DETAILED DESCRIPTION OF THE INVENTION

Figure 1A illustrates an embodiment of a robot 1 comprising a body 2 mounted on wheels 3. In this example, the body 2 is rectangular in shape, and substantially flat, to maintain the center of gravity near the ground and facilitate loading and unloading operations by the operator. Alternatively, the body can be designed according to a wide range of shapes and profiles, depending on the intended uses, and aesthetic qualities required. In a conventional manner, the robot comprises at least one motor, electric or thermal, and means for managing the movements autonomously.

The robot is designed to advance in at least one direction, preferably two, and preferably several angular directions. The change of angular direction is ensured either by pivoting the wheels (two or four directional wheels) or by relative angular velocity variation between the wheels on each side of the robot. For this purpose, the robot is advantageously equipped with four electric motors, located in the axes of the wheels. The body 2 can accommodate one or more batteries and the electronic elements required for the management and guidance of the robot. Still in the example of FIGS. 1A and 1B, the tracking device 4 comprises a radar laser (or LIDAR) adapted to make it possible to locate the surrounding objects over an angular range of 360 °.

Alternatively, other types of sensors may be used, such as one or more cameras, one or more inductive sensors, or others. Hybrid solutions, with several types of sensors, can also be implemented. DRIVING SYSTEM

FIG. 4 schematically illustrates an exemplary autonomous robot driving system comprising a tracking device 4 as previously described, capable of communicating by wired or wireless connection with a driving module 100. For its implementation, FIG. implemented, the driving module comprises at least one microprocessor and instructions 101 for implementation by the microprocessor of the various modules.

The driving module also provides a setpoint module 102, adapted to receive or give and manage a driving instruction of the robot with a linear speed and a curvature. An obstacle detection module 103 makes it possible to detect, in connection with the tracking device 4, the obstacles likely to be on the intended path, within the limit of the effective anticipation distance of the robot.

A robotic anticipation distance determining module 104 is designed to evaluate the minimum distance of observation in front of the vehicle to avoid obstacles. A bypass path determination module 105 makes it possible to evaluate the paths likely to enable the robot to bypass the identified obstacle (s). A scoring module 106 makes it possible to give a more or less favorable notation to each of the possible bypass paths. An avoidance envelope determining module 107 is provided to add a second level of security based on a virtual robot of greater width than that of the actual robot.

The driving module also provides obstacle databases 109, path data 1 10, data of bypass paths 1 January 1 and avoidance envelope data 1 12. Finally, a bus 108 makes it possible, in a conventional manner, to exchange data between the different modules.

DRIVING METHOD WITH ANTICIPATION DEFINITIONS

The term "Instantaneous Center of Rotation" (CIR): the position of the center of a virtual circle passing through the center of the vehicle. Any vehicle (including a tank) has an "Instant Center of Rotation". The position of the CIR changes according to the steering angle of the wheels, as shown in Figure 1 B.

The term "Theoretical Anticipation Distance of the Robot" (or Dist Ant Théo) the minimum theoretical distance to search for obstacles in front of the vehicle to be able to avoid it by turning the wheels fully to the left or right to avoid it (this formula is true if we consider that the vehicle can point instantly and runs perpendicular to a wall). To evaluate this distance, the wheels of the robot are turned to the bottom, which allows to deduce the associated CIR. The theoretical distance of anticipation is the distance between the corner of the furthest robot of the CIR and the CIR associated with the maximum deflection.

The theoretical distance of anticipation is measured from the remarkable point of the robot:

- for a robot type char (with caterpillars), it is its pivot point;

The term "steering distance function" or Fct Dist Bra (speed, variation_courbure), the function that returns the distance traveled by the vehicle according to its speed of movement and the variation of curvature allowed.

The term "Effective Anticipation Distance of the Robot", or Dist Ant Eff (speed, variation_curbure), the function that returns the minimum distance of observation in front of the vehicle to avoid obstacles:

-Dist_Ant Eff (speed, variation_curbure) = Dist Ant Theo + Fct Dist Bra (velocity, variation_curbure).

The anticipation aims to have trajectories with the smallest curvatures possible (or with a large radius). The greater the anticipation distance, the more likely the trajectory of the robot will be to have a small curvature.

To take into account this technical constraint, the term "Curvature Variation" means:

for a robot with directional trains: the variation of curvature defines how much it is allowed to rotate more or less the steering angle of the wheels;

for a robot of the char type: the variation of curvature defines how much it is allowed to modify more or less the angular velocity of the robot.

IMPLEMENTATION OF THE DRIVING METHOD WITH ANTICIPATION Figures 2 and 3 illustrate the main steps for the implementation of the method and the driving system with anticipation of obstacles.

In step 50, a driving instruction is received, such as a linear speed of advance and a curvature.

Step 51 concerns the determination of an effective distance of anticipation of the robot. This step can be done beforehand, before the robot makes a journey. It does not have to be repeated for each trip, as long as the master data does not change.

In the next step 52, the data corresponding to possible obstacles are obtained. This step makes it possible to predetermine a path that the robot will perform if it is applied to the setpoint over an effective anticipation distance (speed, variation_curbure), in which:

- the speed is the set speed;

- Curvature variation is the variation of the curvature allowed for the vehicle. If the variation is 0, then the robot is not allowed to change the steering angle of its wheels in the case of a steered vehicle;

the robot has a given shape (for example a rectangle);

- to predict the position, we define the position where the robot will be located 100 ms later, then we define the position where the robot will be 200 ms later, and so on, until the robot has walked the effective anticipation distance. This distance is a function of the speed and the variation of curvature.

For the predetermination of the path, it is verified that for each predicted position, there is no obstacle, otherwise it is considered that there is a collision.

In steps 53 and 54, a collision test with the obstacle is performed: we list all authorized paths, that is to say we list all curvatures of authorized paths. The requested bending setpoint of the robot is used plus or minus a function of variation of curvature.

Then, for each possible path with its curvature, it deduces the bending setpoint that should be sent to the robot so that it carries out the path with the required curvature. A prediction of the path taking account of the latency of the vehicle, the rise time to achieve the curvature setpoint is made. The prediction extends over the actual distance of anticipation from

According to a preferred embodiment, the purpose of the obstacle avoidance is not to turn around. Also, during the prediction, when the robot has rotated 90 ° from its starting position, the trajectory is extended in a straight line.

Finally, in step 55, a notation is performed. For each possible path with its curvature, we measure the distance traveled by the robot before hitting an obstacle. The curvilinear abscissa is preferably used. The curvature control that has the most favorable obstacle rating is elected at step 56.

In the absence of obstacle (step 57), the robot continues the initial path (step 58).

Figure 3 illustrates, by way of example, the steps for taking into account an obstacle avoidance envelope. Step 60 corresponds to the determination of an obstacle avoidance envelope. Step 61 corresponds to the determination for an obstacle of the bypass for a robot with an obstacle avoidance envelope of a width for example 10% larger than the actual envelope of the robot. If the bypass path with the robot's avoidance envelope does not hit the obstacle (step 62), a bypass path for an obstacle clearance envelope of a surface area of 20% larger is determined. than the actual envelope of the robot (step 63). If this 20% envelope bypass does not hit the obstacle (step 64) then a bypass path is determined for an envelope 30% larger than the actual envelope of the robot (step 65). If the path does not hit the obstacle then the robot will continue this journey (step 67). Nevertheless, at each step of testing the path, if it hits an obstacle (step 68) then the robot continues the initial path if no path having a larger envelope than the actual envelope is satisfactory, or it continues the last path not hitting an obstacle having an avoidance envelope larger than the actual envelope of the robot (step 69). The percentage values given as examples may vary depending on the case. The number of iterations can also vary.

It should be noted that this method and this system are particularly advantageous for robots operating according to a driving mode in which the robot precedes a person who serves as a driver.

In an exemplary embodiment, the obstacle avoidance envelope is determined by successive iterations, from a minimum envelope, to a larger envelope. large, in order to find the optimal envelope ensuring to circumvent the obstacle without coming too close, and by using a weak curvature (or large radius).

[0057] Example of implementation:

- Advance obstacle avoidance is applied and a curvature control solution is found. The list of valid curvatures that are not separated by an obstacle is delimited: optimization zone at 0 cm.

- Advance obstacle avoidance is applied with a virtually larger vehicle width of 10 cm, only on the curvatures of the 0 cm optimization zone. There is a solution. The optimization zone associated with 10 cm is delimited.

- Advance obstacle avoidance is applied with a virtually larger vehicle width of 20 cm, only on the curvatures of the 10cm optimization zone. There is a solution.

- Advance obstacle avoidance is applied with a vehicle width of virtually 30 cm, only on the curvatures of the 20cm optimization zone. There is no solution.

- Also, we retain the last solution found namely in this example the setpoint found for a vehicle width of 20 cm.

It is observed that the driving method according to the invention allows a reactive obstacle avoidance and not a method to find a way to reach a position without touching obstacles. The goal is to see enough in front of the vehicle, to shoot early enough. On the other hand, the goal is not to change the speed of advancement. Also, if you ask the robot to advance in a deadlock it will hang. On the other hand, if the robot is asked to move towards a wall (and perpendicular to it), the robot will avoid the collision by pointing at the right moment. The contribution of the method is to steer before it is too late. This is an obstacle avoidance method that avoids having to intervene on a robot that is in a blocking situation either following a collision or following a trip.

Claims

A driving method for an autonomous robot with an obstacle bypass comprising the steps of:

Establish (50) an instruction of advancement of the robot;

determining (51) an effective anticipation distance of the robot;

receiving (52) the potential obstacle data within the anticipation distance on the initial path;

in the case (53) or the robot would hit an obstacle in the actual distance of anticipation after applying the speed instruction,

determining (54) a plurality of bypass paths, noting the bypass paths, and selecting the route with the most favorable rating;

characterized in that the effective anticipation distance of the robot corresponds to the minimum distance of observation in front of the vehicle to avoid obstacles.

2. Robot driving method according to claim 1, wherein the effective anticipation distance corresponds to the sum of the theoretical distance of anticipation and the distance of obtaining the steering.

3. Robot driving method according to claim 2, wherein the theoretical anticipatory distance of the robot corresponds to the minimum theoretical distance to search for obstacles in front of the vehicle in order to be able to avoid it by fully turning the wheels.

4. A method of driving a robot according to claim 2, wherein the distance of obtaining the steering corresponds to the distance traveled by the vehicle according to its speed of movement and the variation of curvature allowed until the desired deflection is obtained. .

5. Robot driving method according to one of claims 2 to 4, wherein the theoretical distance of anticipation is measured from the remarkable point of the robot, this distance corresponding:

for a robot type char, it is its pivot point;

Robot driving method according to one of claims 1 to 5, also comprising the steps of:

for the chosen path, determining (60) an obstacle avoidance envelope;

perform the chosen route taking into account the obstacle avoidance envelope.

The method of driving a robot according to claim 6, wherein the obstacle avoidance envelope determining step comprises the following steps:

determining a first obstacle avoidance envelope (61);

evaluate, with this first envelope (62), the possibility of circumvention;

if the test with the first envelope is positive, determining a second, larger envelope (63);

evaluate the bypass capabilities (64) for the second envelope;

Robot driving method according to claim 7, wherein the test is performed for at least one other envelope of greater width.

9. autonomous robot driving system (1) for implementing the driving method according to one of claims 1 to 8, comprising a tracking device (4) adapted to communicate with a driving module (100) provided with a microprocessor (101) and implementation instructions, a setpoint module (102), an obstacle detection module (103), an actual anticipation distance determination module of the robot ( 104), a bypass path determination module (105) for determining a plurality of bypass paths, noting the bypass paths, and selecting the most favorable path, a scoring module (106), characterized in that the actual anticipation distance determining module of the robot (104) is provided to determine the effective anticipation distance of the robot corresponding to the minimum observation distance in front of the vehicle to avoid obstacles.

The autonomous robot driving system (1) according to claim 9, comprising an avoidance envelope determining module (107), designed to determine an obstacle avoidance envelope and perform the elected path taking into account the obstacle avoidance envelope,

1 1. Robot (1) autonomous motorized, comprising a body (2) mounted on wheels (3) or tracks, also comprising a driving system according to one of claims 9 or 10.