CN-121978972-A - Predictive low-altitude intelligent navigation method and system based on mine operation plan map

CN121978972ACN 121978972 ACN121978972 ACN 121978972ACN-121978972-A

Abstract

The invention discloses a predictive low-altitude intelligent navigation method and system based on a mine operation plan map, which belong to the technical field of mine safety monitoring, wherein the method comprises the following steps of S1, collecting mine multi-source plan data and constructing the mine operation plan map; S2, predicting potential conflict in a future time window by utilizing a time sequence deep learning network, S3, constructing a double-layer optimization model, solving to obtain an unmanned aerial vehicle track, S4, switching an unmanned aerial vehicle perception mode, S5, constructing a model prediction control frame, and carrying out real-time closed-loop correction on the unmanned aerial vehicle track. The invention integrates mine operation plan data with unmanned aerial vehicle flight control depth, realizes predictive, assisted and intelligent low-altitude intelligent navigation, can be widely applied to various mine scenes such as strip mine stopes, underground mine roadways, tailing reservoirs, dumping grounds and the like, and has extremely high industrial application value.

Inventors

BAO KUNYAN
Xiang Shuiyun
ZHANG JUNQUAN
SHI XIAOLI
WANG DEYOU
HU PENG

Assignees

优备科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (10)

1. A predictive low-altitude intelligent navigation method based on a mine operation plan map is characterized by comprising the following steps of: s1, acquiring mine multisource planning data and constructing a mine operation planning map; S2, predicting potential conflict in a future time window by utilizing a time sequence deep learning network based on a mine operation planning map; S3, constructing a double-layer optimization model, and solving to obtain an unmanned aerial vehicle track; S4, switching the unmanned aerial vehicle sensing mode based on potential conflict in a future time window; S5, based on the switched unmanned aerial vehicle perception mode, a model prediction control frame is constructed, and real-time closed-loop correction is carried out on the unmanned aerial vehicle track.
2. The predictive low altitude intelligent navigation method based on a mine work plan map of claim 1, wherein said S1 comprises the substeps of: S11, acquiring mine multisource planning data; S12, constructing a time-varying hypergraph structure according to mine multisource planning data; and S13, performing embedded learning on the time-varying hypergraph structure through a hypergraph neural network, generating space-time embedded of plan perception, and evolving the time-varying hypergraph structure to complete construction of a mine operation plan map.
3. The predictive low-altitude intelligent navigation method based on mine operation plan map of claim 2, wherein in S12, a time-varying hypergraph structure is adopted The expression of (2) is: ; Wherein, the As a set of vertices, For a time-varying set of hyperedges, Is a time-varying weight function; the vertex set comprises a three-dimensional airspace unit and a mine operation entity; the time-varying hyperedge set includes a planned drive hyperedge and a real-time perceived hyperedge.
4. A predictive low altitude intelligent navigation method based on a mine operation plan map as claimed in claim 3, wherein the plan driving superside comprises blasting influence superside, vehicle track superside, personnel activity superside and maintenance operation superside, and the construction method comprises the following steps: Connecting airspace units in the same blasting area and within the same blasting time range with a blasting operation entity to form blasting influence superflimit; Connecting an airspace unit covered by a vehicle driving path with a corresponding vehicle entity to form a vehicle track overrun; Connecting an airspace unit covered by the personnel activity area with a personnel entity to form a personnel activity overrun; and connecting the airspace unit covered by the overhaul operation area with the equipment entity to form an overhaul operation overrun.
5. The method for predictive low-altitude intelligent navigation based on the mine operation plan map according to claim 2, wherein in S13, the variogram self-encoder is adopted to learn the superside connection relation or the neural ordinary differential equation is adopted to describe the time-varying supergraph structure, so as to complete the evolution of the time-varying supergraph structure.
6. The predictive low-altitude intelligent navigation method based on a mine operation plan map according to claim 1, wherein in S2, the time-series deep learning network comprises a shared time-series convolutional encoder and a plurality of independent decoders.
7. The predictive low-altitude intelligent navigation method based on the mine operation plan map according to claim 1, wherein in the step S3, the double-layer optimization model comprises an upper-layer plan optimization model and a lower-layer track optimization model, and the double-layer optimization model is solved by adopting an alternate direction multiplier method; the lower-layer track optimization model is used for minimizing track cost, and adjusting the dynamic safety distance between the unmanned aerial vehicle and the operation entity to generate an unmanned aerial vehicle track; Dynamic safety distance between unmanned plane and operation entity The expression of (2) is: ; Wherein, the As a risk factor for the risk factor, As a basis for the safety distance of the vehicle, For collision probability predictors based on time-varying hypergraphs, For the relative speed between the drone and the work entity, Indicating time of day Is a time-out variable structure of (c), Indicating a conflict event.
8. The predictive low altitude intelligent navigation method based on a mine operation plan map of claim 1, wherein in S4, the sensing mode includes a vibration sensing mode, a dynamic tracking mode, and a wide area scanning mode; the vibration sensing mode is used for detecting blasting impact and geological disasters, the dynamic tracking mode is used for tracking moving vehicles and personnel, and the wide area scanning mode is used for environment inspection and anomaly detection.
9. The predictive low altitude intelligent navigation method based on mine operation plan map of claim 1, wherein in S5, the model predicts a cost function of a control framework The expression of (2) is: ; Wherein, the To be at the moment Predicted The unmanned state vector of the moment in time, For the moment of time Is used for controlling the input vector of the control system, In order to scroll through the number of time domain steps, As a first matrix of weights, As a second matrix of weights, To weight the square of the euclidean norm, For future risk prediction values based on time-varying hypergraphs, As a coefficient of the risk weight, As a reference trajectory for the reference trajectory, For future time of day Is a time-varying hypergraph structure of (a).
10. The predictive low-altitude intelligent navigation system based on the mine operation plan map is characterized by comprising a mine operation plan map construction module, a predictive conflict resolution module, a plan-flight cooperative optimization module, a perception focus dynamic adjustment module and a real-time closed-loop correction module; The mine operation plan map construction module is used for collecting mine multisource plan data and constructing a mine operation plan map; the predictive conflict resolution module predicts potential conflicts within a future time window using a time-sequential deep learning network; the plan-flight cooperative optimization module is used for constructing a double-layer optimization model and solving to obtain an unmanned aerial vehicle track; The sensing focus dynamic adjustment module is used for switching the sensing mode of the unmanned aerial vehicle; The real-time closed-loop correction module is used for constructing a model predictive control framework and carrying out real-time closed-loop correction on the unmanned aerial vehicle track.

Description

Predictive low-altitude intelligent navigation method and system based on mine operation plan map Technical Field The invention belongs to the technical field of mine safety monitoring, and particularly relates to a predictive low-altitude intelligent navigation method and system based on a mine operation plan map. Background In the open air and underground mine exploitation process, unmanned aerial vehicle inspection has become the important means of guaranteeing production safety, and wide application is in scenes such as side slope monitoring, blasting effect evaluation and equipment inspection. However, the existing mine unmanned aerial vehicle inspection technology still has fundamental defects. In the aspect of flight and operation planning, the current unmanned aerial vehicle mostly executes inspection tasks according to a preset route, cannot sense dynamic operation planning of mine sites, such as blasting operation time schedule, transportation vehicle scheduling, personnel positioning track and the like, and can face safety risks of blasting impact, vehicle collision and the like when the unmanned aerial vehicle flies through an operating area, so that the situation of 'flying without awareness, flying with danger' is caused. In the aspect of collision avoidance, most of collision avoidance in the prior art is in a 'perception-obstacle avoidance' passive mode, namely, an unmanned aerial vehicle adopts emergency avoidance after detecting an obstacle, the post-response mechanism does not react in a mine scene of high-speed flight, the mine operation is characterized by strong planning and space-time predictability (such as clear time of blasting and fixed route of vehicles), but the prior art cannot realize predictive avoidance by utilizing the prior knowledge. In the aspect of multi-source data fusion, a data island exists between unmanned aerial vehicle perception data and mine planning data, real space-ground coordination cannot be realized, multi-source heterogeneous data exists in a mine site, wherein the multi-source heterogeneous data comprises geological exploration data, production scheduling data, equipment state data, personnel positioning data and the like, and the data are stored independently and lack of a unified space-time correlation model. In terms of sensing resource allocation, the existing unmanned aerial vehicle generally adopts a fixed sensing mode (such as uniform scanning) in the inspection process, and cannot dynamically adjust a sensing focus according to front risks, so that insufficient sensing is caused in a key area and calculation force is wasted in a safety area. Therefore, a predictive intelligent navigation method capable of fusing a mine operation plan with the unmanned aerial vehicle flight depth is needed. Disclosure of Invention The invention provides a predictive low-altitude intelligent navigation method and system based on a mine operation plan map. The technical scheme of the invention is that the predictive low-altitude intelligent navigation method based on the mine operation plan map comprises the following steps: s1, acquiring mine multisource planning data and constructing a mine operation planning map; S2, predicting potential conflict in a future time window by utilizing a time sequence deep learning network based on a mine operation planning map; S3, constructing a double-layer optimization model, and solving to obtain an unmanned aerial vehicle track; S4, switching the unmanned aerial vehicle sensing mode based on potential conflict in a future time window; S5, based on the switched unmanned aerial vehicle perception mode, a model prediction control frame is constructed, and real-time closed-loop correction is carried out on the unmanned aerial vehicle track. Further, S1 comprises the following sub-steps: S11, acquiring mine multisource planning data; S12, constructing a time-varying hypergraph structure according to mine multisource planning data; and S13, performing embedded learning on the time-varying hypergraph structure through a hypergraph neural network, generating space-time embedded of plan perception, and evolving the time-varying hypergraph structure to complete construction of a mine operation plan map. Further, in S12, a time-varying hypergraph structureThe expression of (2) is: ; Wherein, the As a set of vertices,For a time-varying set of hyperedges,Is a time-varying weight function; The vertex set comprises a three-dimensional airspace unit and a mine operation entity; the time-varying hyperedge set includes a plan-driven hyperedge and a real-time perceived hyperedge. Further, the planned driving overrun comprises blasting influence overrun, vehicle track overrun, personnel activity overrun and maintenance operation overrun, and the construction method comprises the following steps: Connecting airspace units in the same blasting area and within the same blasting time range with a blasting operation entity to form blasting influence