A system and method for autonomous source search and exploration by robots in unknown environments

Abstract

The present invention relates to the technical field of intelligent robots, and in particular to a system and method for autonomous source search and exploration of robots in an unknown environment. An environment perception module and an environment explorer in a control module jointly explore the environment; a source search judge and a source direction guided exploration decision in the source search perception module and the control module complete the source search decision and guide the work of the environment exploration system; a motion system follows the instructions sent by the environment explorer to a path planner to go to a target exploration point; a communication component is used to receive instructions and report source search results; for autonomous source search of robots in an unknown environment, the present invention adopts source prediction to guide environmental exploration and simultaneously carry out source information judgment, adopts global and local multi-level boundary exploration methods for unknown environments, adopts boundary maximum source search value judgment for local non-source extreme point dilemmas, selects the optimal target exploration point, jumps out of the current local interval, and carries out source search and exploration in a new area until the real source point is found.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent robots, and in particular to a system and method for autonomous source-seeking exploration by robots in an unknown environment. Background Art

[0002] In industrial and mining spaces, various accidents that endanger production safety, such as dangerous gas leakage, nuclear radiation of dangerous substances, and the spread of dangerous chemicals, often occur. In order to deal with and control the scope of danger in a timely manner, it is necessary to quickly find the source of danger and make reasonable disposal. Because such scenes are harmful to people, robots are increasingly used to search and troubleshoot various sources of danger. The robot's search and search for dangerous sources essentially requires the robot to establish a map of the unknown environment in real time, determine the direction of the dangerous source in real time, and quickly navigate to the location of the dangerous source based on the exploration of feasible paths. The robot's autonomous source search technology in an unknown environment includes the robot's exploration and mapping in an unknown environment, the collection of dangerous source strength information, and path planning and navigation technology. At present, there are still the following problems for the robot's autonomous source search and exploration in an unknown environment: First, when the robot faces an unknown environment to search for sources, it needs to explore environmental information to find a feasible path and determine the source direction at the same time. At present, most methods are to search for sources while traversing the environment to establish a map, lack of source direction guidance, low efficiency, and easy to delay the accident handling time. Second, due to the limitations of the complex environment, the source strength information may have non-source local extreme values ​​that are not the true source points. Robot source search can easily fall into local non-source extreme points and be unable to jump out, resulting in source search failure. Summary of the invention

[0003] The purpose of the present invention is to address the problems existing in the background technology and to propose a system and method for autonomous source search and exploration by robots in unknown environments, so as to solve the problem that in complex environments, the source strength information may have non-source local extreme values ​​that are not the true source points, and the robot source search may easily fall into local non-source extreme points and be unable to jump out, resulting in source search failure.

[0004] The technical solution of the present invention is a method for autonomous source search and exploration by a robot in an unknown environment, comprising the following specific steps:

[0005] S1, the robot starts to explore the local area environment;

[0006] S2, screening the boundary points of multi-level environment exploration and extracting truly representative boundary points;

[0007] S3, using the robot's own source strength information sensor to collect source strength information and establish a source strength field;

[0008] S4, predicting the source direction to guide the environment to explore the target;

[0009] S5. Use target detection method to determine whether there is source material at the extreme point of source intensity;

[0010] S6. If it is the source, stop exploring and give the result; if not, jump out of the local non-source extreme point, return to S1, and continue to execute steps S1-S5 until the real source is found, and then exit the search.

[0011] A robot autonomous source-seeking exploration system in an unknown environment, comprising a control module, an environment perception module, a source-seeking perception module, a motion module and a communication component;

[0012] The control module includes an environment explorer, a source point determiner, and a source direction guidance exploration decision maker;

[0013] The environment perception module includes an environment exploration distance sensor and an environment exploration visual sensor;

[0014] The source-seeking perception module includes a source strength information sensor and a source-judgment visual sensor;

[0015] The motion module includes a battery, a 4-wheel drive motion assembly, and a mobile robot body;

[0016] The environmental explorer in the environmental perception module and the control module jointly explores the environment; the source-seeking perception module and the source-seeking determiner and source-direction guided exploration decision in the control module complete the source-seeking decision and guide the work of the environmental exploration system; the motion system follows the instructions sent by the environmental explorer to the path planner to go to the target exploration point; the communication component is used to receive instructions and report the source-seeking results.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects:

[0018] 1. The present invention adopts a real-time source strength maximum direction guidance strategy, selects a feasible interval with the maximum source strength direction in multiple feasible intervals, and selects a maximum source strength path in a single feasible interval. Regardless of whether the feasible interval is wide or narrow, the robot will be guided to always explore the unknown environment in the direction where the source is most likely to exist, thereby enhancing the target of source search.

[0019] 2. Since the source strength information is affected by the complex environmental structure and environmental changes, there may be multiple local source strength extreme points, resulting in a false source phenomenon, causing the robot to be trapped in a local area and unable to find the real source material. The present invention can determine whether the robot is at a non-source extreme point, and find a point with the maximum source strength and greater exploration value at the boundary of the explored environment to continue searching, thereby avoiding the failure of the robot to find the source due to being trapped in a local extreme value, and improving the robustness of the robot's autonomous source finding.

[0020] 3. Compared with the traversal environment exploration and source search, the present invention adopts a real-time maximum source strength traction method, which only needs to explore the area with rich source strength information, and also selects the direction with maximum source strength information when jumping out from the local area, so that the robot does not have to traverse all environmental areas, thereby improving the source search efficiency.

[0021] 4. In terms of environmental exploration methods, a global and local multi-level multi-random tree parallel rapid expansion method is adopted. Compared with the single-level single random tree exploration method, this method explores the environmental boundaries faster. It can not only detect local obstacles at the robot's position in real time, but also detect relatively distant boundary points, and has a higher exploration efficiency.

[0022] 5. In terms of environmental exploration methods, clustering and extraction of boundary points are adopted to obtain valuable representative boundary points, which effectively reduces the number of exploration points to be evaluated and greatly reduces the computational complexity of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of the system structure of the robot for autonomous source search and exploration in an unknown environment of the present invention;

[0024] Figure 2 It is a schematic diagram of the robot's autonomous source-seeking method in an unknown environment of the present invention. DETAILED DESCRIPTION

[0025] Example 1

[0026] like Figure 1 As shown, the present invention proposes a robot autonomous source search and exploration system in an unknown environment, comprising

[0027] 1. Controller module: The specific hardware consists of a computer system CPU i7-7700@3.8GHz, 16GB memory, and the controller hardware system is equipped with Ubuntu system and ROS system, on which the source exploration controller algorithm runs.

[0028] 2. Environmental perception module: including environmental exploration distance sensor using Hokuyo laser radar URG-04LX-UG01; environmental exploration visual sensor using Realsense D435i.

[0029] 3. Source detection module: including source strength information sensor: carbon monoxide concentration detector, range 0-5000PPM, resolution 0.1 / 1PPM; source judgment visual sensor Realsense D435i.

[0030] 4. Robot body and motion module: including 4-wheel drive Macam wheel motion components, mobile robot body and battery system.

[0031] 5. The communication component uses DTU wireless communication module SCCM-4G, which has embedded TCP / UDP protocol.

[0032] A method for a robot to explore and search for sources in an unknown environment is also proposed.

[0033] The specific implementation steps are as follows:

[0034] Step 1: Starting area exploration. Place the robot at the entrance of the factory source exploration area to be explored. After giving the robot an initial direction of travel, use Nq = 2 global random trees Rq(1) and Rq(2) for global exploration. Starting from the robot's current position, Rq(1) randomly generates a coordinate point Pq0(1) with a step size of αq(1). If Pq0(1) is in the obstacle area, it is discarded and a new random coordinate point is generated. If Pq0(1) is in the unknown area, Pq0(1) is placed in the global exploration boundary point set {Qq(t)}. If Pq0(1) is in the feasible area, Pq0(1) is used as the new vertex of Rq(1), and a new random point Pq1(1) is regenerated with a step size of αq(1). Repeat the above process in sequence to obtain the global exploration boundary point set {Qq(t)}. Similarly, Rq(2) simultaneously performs global boundary exploration.

[0035] Similarly, set Nj = 1 local random tree Rj(1), starting from the robot's current position, randomly generate coordinate point Pj0(1) with αj(1) as the step size. If Pj0(1) is in the obstacle area, discard it and regenerate a random coordinate point. If Pj0(1) is in the unknown area, put Pj0(1) into the local exploration boundary point set {Qj(t)} and clear Pj0(1); if Pj0(1) is in the feasible area, take Pj0(1) as the new vertex of Rj(1), and regenerate a new random point Pj0(1) with αj(1) as the step size. Repeat the above process, each time exploring from the robot's current position, and obtain the local exploration boundary point set {Qj(t)}.

[0036] Step 2: Screening of boundary points for multi-level environmental exploration. Take the union of the current global exploration boundary point set {Qq(t)} and the local exploration boundary point set {Qj(t)} to obtain {Qqj(t)}. Set the radius of the boundary point clustering to find the centroid according to the environmental complexity estimate. When the environmental complexity cannot be estimated, the clustering radius scale can be estimated based on the robot body scale, R = k*r, (r is the width of the machine body, k is the coefficient, for example, k can be 0.1). Perform clustering screening through the mean shift method. To control the complexity, constrain the number of elements in the effective boundary point set {Q(t)} obtained after boundary point screening. For example, when the robot control computing power is limited, it can be controlled to N≤1024.

[0037] Step 3: While the robot is exploring the unknown environment, it uses the carbon monoxide concentration detector it carries to collect source strength information (carbon monoxide concentration) and records it to form a carbon monoxide concentration map H(Gp(t)), where Gp(t) is the map of the explored area at time t.

[0038] Step 4: Source direction guides environment exploration.

[0039] When the robot detects that there are n0 (n0>1) feasible intervals separated by obstacles,

[0040] Compare the carbon monoxide concentration at the current position with that at the previous position. When H(P(t))≥H(P(t-1)), the robot takes the nearest boundary point in the feasible region {Q(t)}[n] from P(t-1) to P(t) as the target exploration point, that is, P(t+1)<=Qnear-0(t)[n], Qnear-0(t)[n]∈{Q(t)}[n], Qnear-0(t)[n] is the nearest boundary point within the range of β angle (β can be set to ±15°) in the robot's travel direction.

[0041] If H(P(t))<H(P(t-1)), let the direction from P(t-1) to P(t) be 0 degrees. Considering the efficiency of source search, the periphery is divided into M=6 intervals with P(t) as the center. The interval angle of each interval is γ=60°.

[0042] The robot turns to an angle of γ with P(t) as the center, so that the robot deviates from the direction of decreasing carbon monoxide concentration, and takes the nearest boundary point Qnear-0(t)[n] in the feasible interval {Q(t)}[n] in this direction as the sampling point for carbon monoxide concentration sampling. Repeat the above process until the exploration point P(t+1) with a larger carbon monoxide concentration value than P(t) is obtained as the new exploration direction, and the feasible interval along this direction is explored.

[0043] When the carbon monoxide concentration cannot exceed point P(t) after 6 attempts, record H(P(t)) as the local extreme value and record point P(t) into the set {Ppeak}.

[0044] When n0=1,

[0045] If H(P(t))≥H(P(t-1)), the robot continues to move in the direction from P(t-1) to P(t), P(t+1)<=Qnear-0(t), Qnear-0(t)(0)∈{Q(t)}, Qnear-0(t) is the nearest boundary point within the range of β angle (β is the adjustable direction coefficient set by the system) in the direction of robot movement.

[0046] If H(P(t))<H(P(t-1)), the robot turns to an angle of γ with P(t) as the center, so that the robot deviates from the direction of decreasing carbon monoxide concentration, and takes the nearest boundary point Qnear-1(t) in this direction as the sampling point for carbon monoxide concentration sampling. Repeat the above process until the exploration point P(t+1) with a larger carbon monoxide concentration value than P(t) is obtained as the new exploration direction, and the feasible interval along this direction is explored.

[0047] When the carbon monoxide concentration cannot exceed point P(t) after 6 attempts, record H(P(t)) as the local extreme value and record point P(t) into the set {Ppeak}.

[0048] The above process is repeated until at least one local extreme point is found, and then the next step is entered;

[0049] Step 5:

[0050] Determine whether there is a leakage source at the extreme point of source strength. The Realsense D435i source judgment visual sensor is used to collect the environmental image at the extreme point of source strength {Ppeak}. The YOLOv5-target detection algorithm trained with the source image set features recognizes the environmental image at the extreme point to determine whether there is a leakage point. At the same time, the current environmental image is transmitted to the remote operator through the communication module for auxiliary judgment. Finally, the remote operator gives a confirmation result after judgment. If it is determined that there is a leakage source, the search ends, the source search result is given through the communication component, and the source search work is exited. If the leakage source is not found, proceed to the next step.

[0051] Step 6: If the robot fails to find the carbon monoxide leakage source at the extreme point {Ppeak} of the carbon monoxide concentration source intensity after step 5, it means that the robot is trapped in a local non-source extreme point in the known area and needs to continue searching in the unknown area. Select the 10 boundary points Qfar[n] (n=1,2...10) farthest from the current position P(t) of the robot in the boundary point set {Q(t)} as candidate direction points. Specifically, in order to strengthen the exploration of the boundary with the highest carbon monoxide concentration at the far boundary, the source intensity coefficient can be set to 3-5 times the unknown area coefficient. More specifically, the source intensity coefficient δ in the value gain model formula (1) is set to 1, and the unknown area coefficient λ can be set to 0.2. Formula (2), R0 is calculated by the system according to the exploration boundary scale, the coefficient η is set to 3, and θ is set to 1, and formula (3) and formula (4) are obtained.

[0052] X=δH near +λS=H near +0.2S (3)

[0053]

[0054] According to formula (3), the exploration value gains of the 10 boundary points are obtained. The boundary point corresponding to the maximum exploration value gain is selected as the exploration target point P(t+1) to start exploring the new area.

[0055] Go back to step 1, start exploring new unknown areas, and repeat steps 1-6.

[0056] Example 2

[0057] On the basis of Example 1, the robot system can be equipped with multiple sensors or all-in-one sensors according to the specific characteristics of the source in the source search scenario. For example, in the scenario of source search for leaks of dangerous gases with complex components, the sensor uses all-in-one gas detection (combustible gas / liquid vapor TVOC / toxic and harmful gas), and adopts multiple source information variable combinations at the same time. For example, an all-in-one gas detection sensor can be used to detect sulfur dioxide SO2 and carbon monoxide CO at the same time. The multivariate information data is weighted and normalized to improve the accuracy of source information detection.

[0058] Example 3

[0059] On the basis of the first embodiment, in terms of the source search method, a flexible combination of various methods in the present invention can be adopted.

[0060] More specifically, in terms of the number of fast-expanding random trees for global detection and local detection, the number is adjusted according to the complexity of the environment;

[0061] More specifically, in terms of source-direction-guided environmental exploration, the rotation angle γ of each step in finding the extreme point can be selected according to the current environmental complexity of the robot. For example, the environmental complexity is divided into three levels. The value γ = 360° / 6 is taken in high-complexity scenarios, the value γ = 360° / 5 is taken in medium-complexity scenarios, and the value γ = 360° / 4 is taken in low-complexity scenarios, which can moderately improve the detection efficiency.

[0062] More specifically, in terms of jumping out of the local non-source extreme point, the values ​​of the coefficients in equations (1) and (2) can be flexibly selected based on experimental experience.

Claims

1. A method for autonomous source search and exploration by a robot in an unknown environment, characterized in that: The specific steps include: S1, the robot starts to explore the local area environment; S2, screening the boundary points of multi-level environment exploration and extracting truly representative boundary points; S3, using the robot's own source strength information sensor to collect source strength information and establish a source strength field; S4, predicting the source direction to guide the environment to explore the target; S5. Use target detection method to determine whether there is source material at the extreme point of source intensity; S6, if it is the source, stop exploring and give the result; if not, jump out of the local non-source extreme point, return to S1, and continue to execute steps S1-S5 until the real source is found, and then exit the search; In S6, the N0 boundary points Qfar[n] farthest from the current position P(t) of the robot in the boundary point set {Q(t)} are selected, where n=1, 2...N0 as candidate direction points; wherein N0 is a finite number that can be set by the system; Calculate the value gain X, select the maximum value gain point in Qfar[n], n=1,2...N0 as the target exploration point P(t+1) with the most exploration value; the value gain model is as follows: X=δH near +λS,δ>0,λ>0 (1) Let X be the value gain of N0 candidate points, Hnear be the nearest point in the searched area to Qfar[n], n=1,2…N0 with source strength information, S be the size of the unknown exploration area of ​​Qfar[n], d be the distance from Qfar[n] to the current position P(t) of the robot, δ be the source strength coefficient, and λ be the unknown area coefficient; Return to S1, start exploring new unknown areas, and repeat steps S1-S6.

2. The method for autonomous source search and exploration by a robot in an unknown environment according to claim 1, characterized in that: In S1, the robot first conducts local environment exploration, conducts real-time positioning and mapping through environment exploration visual sensors and environment exploration distance sensors, gradually establishes local maps, and adopts a global and local multi-level environment exploration architecture based on rapidly expanding random trees to conduct environmental exploration.

3. The method for autonomous source search and exploration by a robot in an unknown environment according to claim 2, characterized in that: The method for global and local multi-level environment exploration based on rapidly expanding random trees is as follows: S11. Set Nq global random trees Rq(i)(i∈[1,Nq]), starting from the current position of the robot, with αq(i) as the step length, randomly generate coordinate points Pq0(i), if Pq0(i) is in the obstacle area, discard and regenerate random coordinate points, if Pq0(i) is in the unknown area, put Pq0(i) into the global exploration boundary point set {Qq(t)}, if Pq0(i) is in the feasible area, take Pq0(i) as the new vertex of Rq(i), and regenerate new random points Pq1(i) with αq(i) as the step length; repeat the above process in sequence to obtain the global exploration boundary point set {Qq(t)}; S12, set Nj local random trees Rj(i)(i∈[1,Nj]), starting from the current position of the robot, randomly generate coordinate points Pj0(i) with a step size of αj(i). If Pj0(i) is in the obstacle area, discard it and regenerate a random coordinate point. If Pj0(i) is in the unknown area, put Pj0(i) into the local exploration boundary point set {Qj(t)} and clear Pj0(i); if Pj0(i) is in the feasible area, take Pj0(i) as the new vertex of Rj(i), and regenerate a new random point Pj0(i) with a step size of αj(i); repeat the above process in sequence to obtain the local exploration boundary point set {Qj(t)}; Among them, t is the current time point in the source search process.

4. The method for autonomous source search and exploration by a robot in an unknown environment according to claim 1, characterized in that: S2 uses the mean shift method to perform cluster analysis to obtain the area where the boundary points are highly concentrated, calculates the centroid of the boundary points in the circular area with a radius of R, and screens out the effective boundary point set {Q(t)}.

5. The method for autonomous source search and exploration by a robot in an unknown environment according to claim 1, characterized in that: In S3, while the robot is exploring the unknown environment, a source strength field H(p(t)) is established for the area where the robot has traveled, where P(t) is the position of the robot at time t, and P(t) belongs to the searched area D0.

6. The method for autonomous source search and exploration by a robot in an unknown environment according to claim 5, characterized in that: In S4, the boundary point set {Q(t)} is divided into n feasible intervals according to the obstacle interval, where n is an integer and n≥1, denoted as {Q(t)}[n], n∈[1,n0], where n is an integer; Let the direction from P(t-1) to P(t) be the 0° direction, and divide the periphery into M intervals with P(t) as the center, and the interval angle of each interval is γ = 360° / M; When n0>1, That is, when the robot can detect the existence of multiple feasible intervals at its current location, it first selects the direction of the feasible interval; If H(P(t))≥H(P(t-1)), the robot takes the nearest boundary point in the feasible region {Q(t)}[n] from P(t-1) to P(t) as the target exploration point, that is, P(t+1)<=Qnear-0(t)[n], Qnear-0(t)[n]∈{Q(t)}[n], Qnear-0(t)[n] is the nearest boundary point within the angle range β in the robot's travel direction; If H(P(t))<H(P(t-1)), the robot turns to an angle γ with P(t) as the center, so that the robot deviates from the direction of source intensity reduction, and takes the nearest boundary point Qnear-0(t)[n] in the feasible interval {Q(t)}[n] in this direction as the sampling point for source intensity sampling; Repeat the above process until you get an exploration point P(t+1) with a greater source strength than P(t). This point is used as a new exploration direction, and the feasible interval is explored along this direction. When the source intensity cannot exceed point P(t) after M attempts, record H(P(t)) as the local extreme value and record point P(t) into the set {Ppeak}; When n0=1, the robot can detect the existence of a feasible interval at its current location, and searches for the path closest to the source direction in this feasible interval; If H(P(t))≥H(P(t-1)), the robot continues to move in the direction from P(t-1) to P(t), P(t+1)<=Qnear-0(t), Qnear-0(t)(0)∈{Q(t)}, Qnear-0(t) is the nearest boundary point within the angle range β in the robot's moving direction; If H(P(t))<H(P(t-1)), the robot turns to an angle γ with P(t) as the center, so that the robot deviates from the direction of source intensity reduction, and takes the nearest boundary point Qnear-1(t) in this direction as the sampling point for source intensity sampling; Repeat the above process until you get an exploration point P(t+1) with a greater source strength than P(t), which is used as a new exploration direction, and explore along this direction. When the source strength beyond point P(t) cannot be obtained after M attempts, H(P(t)) is recorded as the local extreme value, and point P(t) is recorded in the set {Ppeak}; the above process is executed repeatedly until at least one local extreme value point is found, and then step S5 is entered; In S4, β is the adjustable directivity coefficient set by the system.

7. The method for autonomous source search and exploration by a robot in an unknown environment according to claim 1, characterized in that: In S5, the source judgment visual sensor collects the environmental image at the extreme point {Ppeak} of the source intensity. The YOLOv5-target detection algorithm trained with the source image set features recognizes the environmental image at the extreme point and autonomously determines whether there is a real source. The current environmental image is transmitted to the remote operator through the communication module for auxiliary judgment. Finally, the remote operator gives a confirmation result after judgment. If it is determined that there is a leakage source, the search is terminated, the source search result is given through the communication component, and the source search work is exited; If the real source is not found, go to the next step.

8. A robot autonomous source search and exploration system in an unknown environment, executing the method according to any one of claims 1 to 7, characterized in that: It includes control module, environment perception module, source perception module, motion module and communication component; The control module includes an environment explorer, a source point determiner, and a source direction guidance exploration decision maker; The environment perception module includes an environment exploration distance sensor and an environment exploration visual sensor; The source-seeking perception module includes a source strength information sensor and a source-judgment visual sensor; The motion module includes a battery, a 4-wheel drive motion assembly, and a mobile robot body; The environmental perception module and the environmental explorer in the control module jointly explore the environment; the source-seeking perception module and the source-seeking determiner and source-direction-guided exploration decision in the control module complete the source-seeking decision and guide the work of the environmental exploration system; the motion system follows the instructions sent by the environmental explorer to the path planner to go to the target exploration point; the communication component is used to receive instructions and report the source-seeking results.

Images