Collision avoidance method based on real-time control multi-target fiber optic positioning unit

Abstract

This invention discloses a collision avoidance method for multi-target fiber optic positioning units based on real-time control. The method steps are as follows: S1: Predicting the trajectory of the fiber optic positioning unit; S2: Performing trajectory collision detection for each fiber optic positioning unit; S3: When a trajectory collision is detected, execute S4; when no trajectory collision is detected, execute S5; S4: Classify the collision points according to their locations, and execute waiting, backtracking, or unfolding procedures according to the collision classification to form a collision-free path, and then execute S5; S5: Perform real-time control on the units according to the corresponding collision-free path, so that all fiber optic positioning units converge to the target position within the required accuracy. This invention can ensure that the units move along the planned path and can perform real-time collision detection.

Description

Technical Field

[0001] This invention relates to the field of multi-target optical fiber technology, and in particular to a collision avoidance method for multi-target optical fiber positioning units based on real-time control. Background Technology

[0002] Currently, research into many fundamental questions in astrophysics and cosmology relies on the statistical properties of large samples. Therefore, acquiring large amounts of spectral data on celestial objects within a given timeframe—i.e., large-scale spectroscopic surveys—is becoming increasingly important. Multi-object fiber optic spectroscopic telescopes (MEASTs) are a crucial component of large-scale spectroscopic survey projects. Advances in fiber optic positioning technology have enabled scientists to simultaneously observe thousands of celestial objects within a short period. Due to the large number of celestial objects, survey efficiency is paramount to achieving the intended scientific objectives. The efficiency of large-scale spectroscopic surveys depends on the utilization rate of fiber optic positioning units and the observational completeness of celestial objects, i.e., the celestial object allocation methods relying on fiber optic positioning units and the resolution of collision problems.

[0003] To ensure there are no blind spots on the focal plane, the observation areas of adjacent fiber optic positioning units overlap, potentially leading to collisions. Several algorithms have been proposed to address this collision issue. One approach is hardware-based, adding extra sensors to the units and using changes in potential difference, waveform comparison, and pulse count to detect collisions and stop unit movement promptly. This method cannot prevent collisions entirely, but only allows colliding units to stop in time, minimizing damage. Another approach is software-based path planning. Makarem et al. proposed a distributed navigation planning method based on an artificial potential field for the DESI project, artificially constructing a near-natural potential field to guide units from areas of high potential energy to areas of low potential energy. However, this method can lead to unit locking and oscillation as the overlap area increases. Tao et al. combined this method with a priority-based finite state machine for the MOONS project, increasing the probability of units converging to the target position, but also increasing the algorithm's complexity.

[0004] To more effectively address collision issues, Zhang et al. proposed the concept of a safe zone. Within this zone, elements will not collide with adjacent elements regardless of their movement. Therefore, a retraction-rotation-expansion method was proposed by Silber et al., where element movement is divided into three steps: first, all element eccentric axes are retracted to ensure all elements are within the safe zone; then, the element's central axis is rotated to the target angle; and finally, the eccentric axes are expanded to the target position. This method can resolve most collisions, but it increases communication overhead and reduces coarse positioning accuracy by providing margin for the anti-collision path. Zhang et al. also proposed a tangent-based method within the safe zone, first moving the element to the tangent point between the target point and the safe zone, and then moving along the tangent to the target point. This method works in simulations, but in reality, due to the double-rotation characteristic, the element's movement cannot be a straight line. None of the above anti-collision methods are real-time; that is, on the one hand, these methods cannot guarantee that the element moves along the planned path, and on the other hand, they cannot stop the element's movement in time when it deviates from the planned path and collides. Summary of the Invention

[0005] Based on the technical problems existing in the background technology, the present invention proposes a collision avoidance method for multi-target fiber optic positioning units based on real-time control, which can ensure that the unit moves along the planned path and can perform real-time collision detection.

[0006] The anti-collision method for multi-target fiber optic positioning units based on real-time control proposed in this invention comprises the following steps:

[0007] S1: Prediction of the trajectory of the fiber optic positioning unit;

[0008] S2: Perform trajectory collision detection for each fiber optic positioning unit;

[0009] S3: When a trajectory collision is detected, execute S4; when no trajectory collision is detected, execute S5.

[0010] S4: Classify the collision points according to their locations, and execute the waiting, backtracking, or unfolding procedure according to the collision classification to form a collision-free path, and then execute S5;

[0011] S5: Real-time control of the units is performed according to the corresponding collision-free path to ensure that all fiber optic positioning units converge to the target position within the required accuracy.

[0012] Preferably, the prediction method in S1 is as follows: calculate the rotation angle of the central axis and the eccentric axis based on the starting position and target position of the fiber optic positioning unit, and the trajectory points obtained after equal division are the predicted unit trajectories.

[0013] Preferably, the starting position is achieved by illuminating the optical fiber carried by the optical fiber positioning unit using a back-illuminated method, and then capturing a focal plane image using a pre-calibrated camera system.

[0014] Preferably, the camera system uses a telecentric lens and calibrates the camera through an external calibration target to obtain fourth-order polynomial transformation parameters between world coordinates and pixel coordinates.

[0015] Preferably, the method for trajectory collision detection in S2 is as follows: if the minimum value of the eccentric axis distance from a certain trajectory point of a certain fiber optic positioning unit to all trajectory points of a certain adjacent fiber optic positioning unit is less than the collision threshold, then a certain trajectory point of a certain fiber optic positioning unit is regarded as a collision point.

[0016] Preferably, the collision classification in S4 includes the following cases: the collision point does not coincide with the starting position and the target position of the unit; the collision point coincides with the starting position of one unit; the collision point coincides with the target position of one unit; the collision point coincides with the starting positions of two units; the collision point coincides with the starting position of one unit and coincides with the target position of another unit; the collision point coincides with the target positions of two units; the collision point coincides with the starting positions of two units and coincides with the target position of one unit; the collision point coincides with the target positions of two units and coincides with the starting position of one unit.

[0017] Preferably, when the collision point does not coincide with the starting position and target position of the unit, the collision point coincides with the starting position of one of the units, the collision point coincides with the target position of one of the units, or the collision point coincides with the starting position of one unit and the target position of another unit, the unit with the collision point later is selected to execute the waiting procedure; when the collision point coincides with the starting positions of two units, or the collision point coincides with the starting positions of two units and the target position of one unit, the unit with the collision termination point later is selected to execute the rollback procedure; when the collision point coincides with the target positions of two units, or the collision point coincides with the target positions of two units and the starting position of one unit, the unit with the collision point earlier is selected to execute the unfolding procedure.

[0018] Preferably, the real-time control method in S5 is as follows: the camera system acquires back-illuminated images of the optical fiber on the focal plane, the hardware acceleration platform on the FPGA development board processes the images to obtain the position of the optical fiber, and the control system performs real-time collision detection based on the position of the optical fiber; if the collision threshold is reached, the control unit stops moving; if the collision threshold is not reached, an adjustment command is sent to the unit, and the adjustment process is repeated until the unit reaches the target accuracy.

[0019] Beneficial technical effects of the present invention:

[0020] This invention divides the rotation angles of the central axis and eccentric axis of the optical fiber into equal parts and performs trajectory collision detection. Then, it classifies collisions according to the starting and ending points of the collision unit pairs and selects the corresponding waiting unit, retreat unit, or unfolding unit. Finally, it performs real-time control of the units according to the corresponding collision-free path. This method can ensure that the units move along the planned collision-free path while performing real-time collision detection to avoid unit collisions caused by deviation from the path or other unexpected situations. Attached Figure Description

[0021] Figure 1 This is a flowchart of the anti-collision method for fiber optic positioning units based on real-time control proposed in this invention.

[0022] Figure 2 This is a classification diagram of the collision of two unit trajectories proposed in this invention;

[0023] Figure 3 This is a schematic diagram of the collision-free path for the waiting unit proposed in this invention;

[0024] Figure 4 This is a schematic diagram of the collision-free path of the back-up unit proposed in this invention;

[0025] Figure 5 This is a schematic diagram of the collision-free path of the unfolding unit proposed in this invention. Detailed Implementation

[0026] The present invention will be further explained below with reference to specific embodiments.

[0027] Reference Figure 1 The anti-collision method for multi-target fiber optic positioning units based on real-time control proposed in this invention comprises the following steps:

[0028] S1: Based on the starting position and target position of the unit, the rotation angles of the central axis and eccentric axis can be calculated. Divide both angles into 2000 equal parts to obtain 2000 trajectory points, which are the predicted unit trajectories.

[0029] S2: When unit a is at trajectory point a i When i=1, 2…2000, calculate the trajectory point b from ai to the adjacent cell b. j The eccentricity distance (j=1, 2…2000). For each ai, iterate through all b. j Get the minimum distance s ai (i=1, 2…2000), if s ai If the value is less than the collision threshold, the trajectory point is marked as a collision point. Among the collision points, the one with the smallest value i is the collision start point, and the one with the largest value i is the collision end point. Similarly, trajectory collision detection is performed for each unit.

[0030] S3: When a trajectory collision is detected, execute S4; when no trajectory collision is detected, execute S5.

[0031] S4: As Figure 2 The collision starting point S of the collision unit pair is shown. 1、 S2 and termination point E 1、 E2 classifies collisions. A collision is labeled M when the collision point does not coincide with either the starting or target position of a cell; H when the collision point coincides with the starting position of one of the cells; and T when the collision point coincides with the target position of one of the cells. Similarly, the remaining cases HH, HT, TT, HHT, and HTT can be derived.

[0032] Based on collision classification, corresponding collision-free paths are obtained: when the collision type is M, H, T, or HT, the cell where the collision point occurs later is selected as the waiting cell, such as... Figure 3 As shown, the other unit moves first until it passes the collision termination point, then waits for the unit to move, and eventually both units reach the target position; when the collision type is HH or HHT, the unit with the collision termination point later is selected as the retreat unit, as shown in the example. Figure 4 As shown, the retracting unit first retracts its eccentric shaft to allow another unit to move. After the other unit passes the collision termination point, the retracting unit then deploys its eccentric shaft back to its initial position. The retracting unit then moves towards the target position, and eventually both units reach the target position. When the collision type is TT or HTT, the unit with the earlier collision point is selected as the deploying unit, as shown below. Figure 5 As shown, the unfolding unit first moves to the target position, which is also the collision termination point, and then retracts the eccentric axis. Next, another unit moves to the target position, and finally the unfolding unit unfolds the eccentric axis back to the target position. After forming a collision-free path, S5 is executed.

[0033] S5: The unit is controlled in real time according to the corresponding collision-free path. The photogrammetric device acquires the back-illuminated image of the optical fiber on the focal plane. The image is processed by the hardware acceleration platform on the FPGA development board to obtain the position of the optical fiber. The control system detects collisions in real time according to the position of the optical fiber. If the collision threshold is reached, the unit stops moving. Otherwise, an adjustment command is sent to the unit. The adjustment process is repeated until the unit reaches the target accuracy.

[0034] The photogrammetry device in this embodiment uses a dual telecentric lens with a magnification of 0.355 and a 2048x2448 area array CCD camera. The fiber optic positioning unit is positioned at the optimal imaging distance from the photogrammetry device. The FPGA development board used is a Zynq Ultra+ MPSoC ZU3EG, which can achieve an image processing speed of 17ms for the used resolution.

[0035] The devices and structures involved in the above-described solutions of this invention are not complex, and the experimental costs required to implement the entire solution are relatively low. Based on real-time control of the units, it can be ensured that the units move along the planned collision-free path, while real-time collision detection can be performed to avoid unit collisions due to deviation from the path or other unexpected situations.

Claims

1. A collision avoidance method based on a multi-target fiber optic positioning unit with real-time control, characterized in that, The steps are as follows: S1: Based on the starting position and target position of the fiber optic positioning unit, calculate the rotation angles of the central axis and eccentric axis, divide both angles into 2000 equal parts, and obtain 2000 trajectory points, which are the predicted trajectory of the fiber optic positioning unit. S2: When fiber optic positioning unit a is at trajectory point a i When i = 1, 2, ..., 2000, calculate a. i To the adjacent fiber optic positioning unit b at trajectory point b j The eccentric axis distance (j=1, 2…2000), for each a i Traverse all b j Get the minimum distance s ai (i=1, 2…2000), if s ai If the value is less than the collision threshold, the trajectory point is marked as a collision point. Among the collision points, the one with the smallest value i is the collision start point, and the one with the largest value i is the collision end point. Similarly, trajectory collision detection is performed on each fiber optic positioning unit. S3: When a trajectory collision is detected, execute S4; when no trajectory collision is detected, execute S5. S4: Based on the collision initiation point S of the fiber optic positioning unit pair. 1、 S2 and termination point E 1、 E2. Collisions are classified into the following categories: the collision point does not coincide with the starting position and the target position of the fiber optic positioning unit; the collision point coincides with the starting position of one fiber optic positioning unit; the collision point coincides with the target position of one fiber optic positioning unit; the collision point coincides with the starting position of two fiber optic positioning units; the collision point coincides with the starting position of one fiber optic positioning unit and the target position of another fiber optic positioning unit; the collision point coincides with the target position of two fiber optic positioning units; the collision point coincides with the starting position of two fiber optic positioning units and the target position of one fiber optic positioning unit; the collision point coincides with the target position of two fiber optic positioning units and the starting position of one fiber optic positioning unit. Based on collision classification, the corresponding collision-free path is obtained: When the collision point does not coincide with the starting position and target position of the fiber optic positioning unit, coincides with the starting position of one of the fiber optic positioning units, coincides with the target position of one of the fiber optic positioning units, or coincides with both the starting position and target position of one fiber optic positioning unit, the fiber optic positioning unit with the later collision point is selected as the waiting fiber optic positioning unit. The other fiber optic positioning unit moves first until it passes the collision termination point, then waits for the fiber optic positioning unit to move again, and eventually both fiber optic positioning units can reach the target position; when the collision point coincides with the starting positions of two fiber optic positioning units, or coincides with both the starting positions and target position of one fiber optic positioning unit, the fiber optic positioning unit with the later collision termination point is selected as the waiting fiber optic positioning unit. The retracting fiber optic positioning unit first retracts its eccentric shaft to allow another fiber optic positioning unit to move. Once the other fiber optic positioning unit passes the collision termination point, the retracting fiber optic positioning unit then unfolds its eccentric shaft back to its initial position. The retracting fiber optic positioning unit then moves towards the target position, until both fiber optic positioning units reach the target position. When the collision point coincides with the target positions of both fiber optic positioning units, or when the collision point coincides with both the target positions of the two fiber optic positioning units and the initial position of one fiber optic positioning unit, the fiber optic positioning unit with the earlier collision point is selected as the unfolding fiber optic positioning unit. The unfolding fiber optic positioning unit first moves to the target position, which is also the collision termination point, then retracts its eccentric shaft. Next, the other fiber optic positioning unit moves to the target position, and finally, the unfolding fiber optic positioning unit unfolds its eccentric shaft back to the target position. After a collision-free path is formed, execute S5; S5: Real-time control of the fiber optic positioning unit according to the corresponding collision-free path, so that all fiber optic positioning units converge to the target position within the required accuracy. The real-time control method is as follows: the camera system acquires back-illuminated images of the optical fiber on the focal plane, and the hardware acceleration platform on the FPGA development board processes the images to obtain the position of the optical fiber. The control system detects collisions in real time based on the position of the optical fiber. If the collision threshold is reached, the optical fiber positioning unit is controlled to stop moving. If the collision threshold is not reached, an adjustment command is sent to the optical fiber positioning unit. The adjustment process is repeated until the optical fiber positioning unit reaches the target accuracy.

2. The anti-collision method for a multi-target fiber optic positioning unit based on real-time control according to claim 1, characterized in that, The starting position is obtained by illuminating the optical fiber carried by the optical fiber positioning unit using the back-illumination method, and then capturing the focal plane image using a pre-calibrated camera system.

3. The anti-collision method for a multi-target fiber optic positioning unit based on real-time control according to claim 2, characterized in that, The camera system uses a telecentric lens and calibrates the camera through an external calibration target to obtain fourth-order polynomial transformation parameters between world coordinates and pixel coordinates.

Images