Half-high container multimodal transport whole-process monitoring and tracking method and system

Abstract

The application provides a semi-high container multimodal transport whole-process monitoring and tracking method and system, relates to the container monitoring technical field, and comprises the following steps: when the tracking signal is interrupted, a container topology constraint graph is constructed, position inference is carried out based on a reference container position vector change, a continuous space-time trajectory is generated by combining matching of a reloading operation sequence and physical adaptability verification. The method can effectively solve the position tracking problem of a signal blind area, improve the continuity and reliability of multimodal transport whole-process monitoring, and reduce the risk of container loss.

Description

Technical Field

[0001] This invention relates to the field of container monitoring technology, and in particular to a method and system for full-process monitoring and tracking of semi-high container multimodal transport. Background Technology

[0002] Semi-high containers play a vital role in multimodal transport systems, with their unique dimensions offering significant advantages for transporting specific goods. Typically half the height of a standard container, semi-high containers provide greater flexibility in space utilization and cargo loading. As global trade expands and logistics networks become more complex, the demand for real-time monitoring and tracking of semi-high containers throughout multimodal transport is growing.

[0003] Existing container monitoring and tracking technologies primarily rely on signal transmission devices such as RFID and GPS for positioning. However, in actual multimodal transport environments, these technologies face problems such as signal interruption and tracking failure. In particular, semi-high containers often require transshipment and stacking operations during transportation, making it difficult for monitoring systems to maintain continuous tracking.

[0004] However, current technology still has shortcomings and deficiencies. It lacks an effective location inference mechanism in the event of signal interruption. When the signal equipment of a half-height container is interrupted due to environmental interference, battery depletion, or physical shielding, existing technology struggles to infer the possible location of the target container from surrounding environmental information and the relationship with adjacent containers, creating monitoring blind spots. Furthermore, it lacks the ability to monitor the transshipment process based on spatial topology. During multimodal transport transshipment, the stacking position of half-height containers changes frequently. Existing technology cannot effectively identify and record these spatial topological changes, making it difficult to construct a complete transshipment operation sequence, leading to tracking gaps. Finally, it lacks a location determination method that considers physical constraints. Existing technology often ignores the physical characteristics of half-height containers and the geometric constraints of their environment when determining container location, such as stacking stability, load-bearing capacity, and spatial adaptability. This results in discrepancies between the location information provided by the monitoring system and the actual situation, reducing tracking accuracy and reliability. Summary of the Invention

[0005] This invention provides a method and system for full-process monitoring and tracking of semi-high container multimodal transport, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a method for end-to-end monitoring and tracking of semi-high container multimodal transport, comprising:

[0007] Collect spatial layout data of containers and construct a topological constraint graph. The topological constraint graph represents containers through nodes and represents spatial adjacency and stacking dependency relationships through edges.

[0008] When the tracking signal of the target half-height container is interrupted, the reference container adjacent to the target half-height container is extracted from the topology constraint graph. The position vector change of the reference container is projected onto the constraint space of the target half-height container according to the stacking dependency relationship. The weighted fusion is iterated until convergence, and the spatial position distribution domain is formed by combining the geometric boundary of the carrying tool.

[0009] Node matching is performed on the topological constraint graphs before and after the replacement, non-invariant sets with positional changes are extracted, and a temporal dependent directed graph is constructed based on the kinematic reachability cone and vertical stacking load-bearing limit of the handling equipment and the topological sorting is performed to generate the replacement operation sequence.

[0010] For candidate locations in the spatial location distribution domain, stacking hierarchy features and neighborhood height features are extracted to obtain spatial stability indicators. Physical adaptability verification is performed according to priority, and locations are selected and determined based on remaining bearing capacity and overlapping area ratio.

[0011] The change-up sequence is used to construct a spatiotemporal alignment matrix with the determined position. Dynamic programming is then used to search for the optimal spatiotemporal alignment sequence, generating a continuous trajectory.

[0012] In an optional embodiment, when the tracking signal of the target half-height container is interrupted, reference containers adjacent to the target half-height container are extracted from the topology constraint graph. The position vector changes of the reference containers are projected onto the constraint space of the target half-height container according to the stacking dependency. Weighted fusion is iterated until convergence, and a spatial position distribution domain is formed by combining the geometric boundaries of the carrying vehicle, including:

[0013] Extract the set of reference containers adjacent to the target half-height container from the topology constraint graph;

[0014] Based on the position sequence of the reference container at multiple historical moments, the acceleration characteristics and directional stability characteristics are calculated. When the acceleration characteristics exceed the preset stability threshold or the directional stability characteristics are lower than the preset consistency threshold, it is determined that the reference container has undergone non-rigid coupling motion. The propagation weight of the reference container is adjusted to the attenuation weight value to obtain a weighted reference container set.

[0015] Based on the stacking dependency between the reference container and the target half-height container in the weighted reference container set, the vertical position vector change is projected or masked to obtain a layered projection vector set.

[0016] The hierarchical projection vector set is weighted and fused, and the uncertainty metric is calculated. When the uncertainty metric exceeds the preset convergence threshold, the extended reference container with second-order adjacency relationship is extracted, the second-order propagation weight is calculated and projected to supplement the hierarchical projection vector set, and the process is iterated until convergence.

[0017] The spatial location distribution domain is obtained by intersecting the converged set of layered projection vectors with the geometric boundary of the bearing tool.

[0018] In one optional embodiment, the hierarchical projection vector set is weighted and fused, and an uncertainty metric is calculated. When the uncertainty metric exceeds a preset convergence threshold, an extended reference container for extracting second-order adjacency relationships is expanded, second-order propagation weights are calculated, and the container is projected to supplement the hierarchical projection vector set. This process is iterated until convergence, including:

[0019] For each projection vector in the layered projection vector set, a weighted sum is performed according to the propagation weight of the corresponding reference container to obtain the fused position vector;

[0020] Calculate the spatial deviation between each projection vector in the layered projection vector set and the fused position vector, and obtain an uncertainty measure based on the weighted variance of the spatial deviation;

[0021] When the uncertainty metric exceeds the preset convergence threshold, the node adjacent to any reference container in the reference container set is found from the topology constraint graph. After excluding the nodes already in the reference container set, the extended reference container set is obtained.

[0022] For each extended reference container in the extended reference container set, obtain the first propagation weight of the adjacent reference containers in the reference container set, and multiply the first propagation weight by the propagation weight of the adjacent reference containers to obtain the second-order propagation weight of the extended reference container.

[0023] The position vector change of the extended reference container is projected onto the constraint space of the target half-height container with second-order propagation weights, and the projection result is supplemented into the hierarchical projection vector set.

[0024] Repeat the process until the uncertainty metric converges below a preset convergence threshold.

[0025] In one optional embodiment, node matching is performed on the topological constraint graphs before and after the replacement, non-invariant sets of positional changes are extracted, and a temporally dependent directed graph is constructed based on the kinematic reachability cone and the load-bearing limit of vertical stacking of the handling equipment, and topological sorting is performed to generate a replacement operation sequence including:

[0026] Based on the initial and target topology constraint graphs determined before and after the transformation, node pairs are extracted through node matching.

[0027] Calculate the position coordinate difference between node pairs. When the difference exceeds the preset position invariance threshold, mark it as a node with a changed position and summarize it to obtain a non-invariant set.

[0028] For each node with a change in position, a kinematically reachable cone is constructed with the current position of the handling equipment as the vertex. When the spatial position of the node is not within the reachable cone, the shortest path is extracted and the occluded nodes on the path are identified. Predecessor dependency edges are established from the occluded nodes to the node with a change in position.

[0029] For each position-changing node in the non-invariant set, extract the upper load-bearing node of the vertical stack from the initial topology constraint graph, calculate the ratio of the mass of the upper load-bearing node to the load-bearing limit of the position-changing node, and when the ratio exceeds the preset load-bearing safety factor, establish a load-bearing dependency edge from the upper load-bearing node to the position-changing node.

[0030] The nodes with changes in position in the non-invariant set are used as nodes in the temporal dependent directed graph. The predecessor dependency edge and the bearing dependency edge are added to the temporal dependent directed graph. The topological sorting is performed to obtain the linear permutation sequence, and the replacement operation sequence is determined.

[0031] In one optional embodiment, for each location-changed node, a kinematically reachable cone is constructed with the current position of the transport equipment as the vertex. When the spatial position of a node is not within the reachable cone, the shortest path is extracted and occluded nodes on the path are identified. The predecessor dependency edges from the occluded nodes to the location-changed nodes are established, including:

[0032] Using the current position of the handling equipment as the cone apex, the cone opening angle is determined based on the joint angle range of the handling equipment, and the cone height is determined based on the extension radius of the end effector. The cone geometry range of the kinematically reachable cone is constructed. The distance of the position change node from the cone apex to the distance and the deflection angle from the cone central axis are calculated in the initial topology constraint diagram. When the distance exceeds the cone height or the deflection angle exceeds the cone opening angle, it is determined that the position change node is not within the kinematically reachable cone range.

[0033] The initial topology constraint graph is converted into a spatial connectivity graph. In the spatial connectivity graph, the shortest path search is performed with the current position of the transport equipment as the starting point and the spatial position of the position change node as the ending point to obtain the path node sequence. The path node sequence is traversed, and the height difference between each path node and the position change node in the vertical direction is calculated. When the height difference is greater than the preset occlusion judgment threshold, the path node is marked as an occlusion node.

[0034] Create a predecessor dependency edge in a temporally dependent directed graph, pointing from an occluded node to a node whose position changes.

[0035] In one optional embodiment, for candidate locations in the spatial location distribution domain, stacking hierarchy features and neighborhood height features are extracted to obtain spatial stability indices. Physical adaptability verification is performed according to priority. The location is determined by screening based on remaining bearing capacity and overlapping area ratio, including:

[0036] For each candidate location in the spatial location distribution domain, the stacking level number of the candidate location in the topological constraint graph is extracted to determine the stacking level feature, the vertical height range of the nodes in the neighborhood of the candidate location is extracted to determine the neighborhood height feature, the stacking level feature and the neighborhood height feature are weighted and combined to obtain the spatial stability index, and the candidate locations are prioritized according to the spatial stability index.

[0037] Physical compatibility verification is performed on candidate locations in order of priority. The difference between the remaining load capacity of the load-bearing node below the candidate location and the mass of the half-height container is calculated. The overlap area ratio between the load-bearing node below the candidate location and the bottom of the half-height container is calculated. When the difference in remaining load capacity is positive and the overlap area ratio exceeds the preset area ratio threshold, the candidate location is marked as a confirmed location and the verification process for subsequent candidate locations is terminated.

[0038] In an optional embodiment, a spatiotemporal alignment matrix is ​​constructed by combining the changing operation sequence with the determined position, and dynamic programming search is performed to obtain the spatiotemporally optimal alignment sequence, generating a continuous trajectory including:

[0039] Extract the timestamp of each costume change operation node from the costume change operation sequence, and extract the spatial coordinates of each specific location from the set of specific locations;

[0040] Construct a spatiotemporal alignment matrix, using timestamps as row indices and spatial coordinates as column indices. For each combination of a changeover operation node and a determined location, calculate the shortest path length in the topology constraint graph as the topological distance, and calculate the Euclidean distance between the spatial coordinates of the determined location and the corresponding spatial location of the changeover operation node as the spatial distance. Perform a weighted sum of the topological distance and the spatial distance to obtain the spatiotemporal matching cost, and fill the spatiotemporal alignment matrix with the spatiotemporal matching cost.

[0041] Dynamic programming search is performed on the spatiotemporal alignment matrix. Starting from the start timestamp of the costume change operation sequence, the determined positions that minimize the cumulative spatiotemporal matching cost are selected sequentially in time order. At the same time, the shortest path length between the currently selected determined position and the previously selected determined position in the initial topology constraint graph is constrained to be less than a preset reachable threshold, so as to obtain the spatiotemporally optimal alignment sequence.

[0042] By sequentially connecting the timestamps and spatial coordinates in the optimally aligned spatiotemporal sequence, a continuous trajectory is generated.

[0043] A second aspect of the present invention provides a half-height container multimodal transport end-to-end monitoring and tracking system, comprising:

[0044] The first unit is used to collect spatial layout data of containers and construct a topological constraint graph. The topological constraint graph represents containers through nodes and represents spatial adjacency and stacking dependency relationships through edges.

[0045] The second unit is used to extract reference containers adjacent to the target half-height container from the topology constraint graph when the tracking signal of the target half-height container is interrupted. Based on the stacking dependency, the position vector change of the reference containers is projected onto the constraint space of the target half-height container, and the weighted fusion is iterated until convergence. Combined with the geometric boundary of the carrying tool, a spatial position distribution domain is formed.

[0046] The third unit is used to perform node matching on the topological constraint graphs before and after the replacement, extract the non-invariant set of position changes, construct a time-dependent directed graph based on the kinematic reachability cone and the bearing limit of vertical stacking of the handling equipment, and perform topological sorting to generate the replacement operation sequence.

[0047] The fourth unit is used to extract stacking hierarchy features and neighborhood height features from candidate locations in the spatial location distribution domain to obtain spatial stability indicators, perform physical adaptability verification according to priority, and screen and determine locations based on remaining bearing capacity and overlapping area ratio.

[0048] The fifth unit is used to construct a spatiotemporal alignment matrix by combining the changing operation sequence with the determined position, perform dynamic programming search to obtain the spatiotemporally optimal alignment sequence, and generate a continuous trajectory.

[0049] A third aspect of the present invention provides an electronic device, comprising:

[0050] processor;

[0051] Memory used to store processor-executable instructions;

[0052] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0053] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0054] In this embodiment of the invention, by constructing a topological constraint graph and a temporally dependent directed graph, signal tracking recovery and spatial positioning of half-height containers are achieved, effectively solving the problem of container position loss caused by tracking signal interruption during multimodal transport, and improving the continuity and reliability of logistics tracking. Physical adaptability verification is performed using stacking dependency relationships and spatial stability indicators to ensure accurate positioning of half-height containers. Weighted fusion iteration and carrying capacity analysis improve the system's positioning accuracy and physical feasibility, making the spatial position inference of containers during transshipment more consistent with actual operating conditions. The use of a spatiotemporal alignment matrix and dynamic programming algorithm to generate continuous trajectories enables optimized sorting of transshipment operation sequences and accurate prediction of container positions, enhancing the integrity and real-time performance of multimodal transport monitoring, and significantly reducing logistics management costs and the risk of cargo loss. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the method for full-process monitoring and tracking of semi-high container multimodal transport according to an embodiment of the present invention.

[0056] Figure 2 A flowchart is constructed for the temporal dependency of kinematic reachability cone and spatial occlusion in semi-high container multimodal transport. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0059] Figure 1 This is a flowchart illustrating the method for end-to-end monitoring and tracking of semi-high container multimodal transport according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0060] Collect spatial layout data of containers and construct a topological constraint graph. The topological constraint graph represents containers through nodes and represents spatial adjacency and stacking dependency relationships through edges.

[0061] When the tracking signal of the target half-height container is interrupted, the reference container adjacent to the target half-height container is extracted from the topology constraint graph. The position vector change of the reference container is projected onto the constraint space of the target half-height container according to the stacking dependency relationship. The weighted fusion is iterated until convergence, and the spatial position distribution domain is formed by combining the geometric boundary of the carrying tool.

[0062] Node matching is performed on the topological constraint graphs before and after the replacement, non-invariant sets with positional changes are extracted, and a temporal dependent directed graph is constructed based on the kinematic reachability cone and vertical stacking load-bearing limit of the handling equipment and the topological sorting is performed to generate the replacement operation sequence.

[0063] For candidate locations in the spatial location distribution domain, stacking hierarchy features and neighborhood height features are extracted to obtain spatial stability indicators. Physical adaptability verification is performed according to priority, and locations are selected and determined based on remaining bearing capacity and overlapping area ratio.

[0064] The change-up sequence is used to construct a spatiotemporal alignment matrix with the determined position. Dynamic programming is then used to search for the optimal spatiotemporal alignment sequence, generating a continuous trajectory.

[0065] In one optional implementation, when the tracking signal of the target half-height container is interrupted, reference containers adjacent to the target half-height container are extracted from the topology constraint graph. The position vector changes of the reference containers are projected onto the constraint space of the target half-height container according to the stacking dependency. Weighted fusion iterations are then performed until convergence, and a spatial position distribution domain is formed by combining the geometric boundaries of the carrying vehicle, including:

[0066] Extract the set of reference containers adjacent to the target half-height container from the topology constraint graph;

[0067] Based on the position sequence of the reference container at multiple historical moments, the acceleration characteristics and directional stability characteristics are calculated. When the acceleration characteristics exceed the preset stability threshold or the directional stability characteristics are lower than the preset consistency threshold, it is determined that the reference container has undergone non-rigid coupling motion. The propagation weight of the reference container is adjusted to the attenuation weight value to obtain a weighted reference container set.

[0068] Based on the stacking dependency between the reference container and the target half-height container in the weighted reference container set, the vertical position vector change is projected or masked to obtain a layered projection vector set.

[0069] The hierarchical projection vector set is weighted and fused, and the uncertainty metric is calculated. When the uncertainty metric exceeds the preset convergence threshold, the extended reference container with second-order adjacency relationship is extracted, the second-order propagation weight is calculated and projected to supplement the hierarchical projection vector set, and the process is iterated until convergence.

[0070] The spatial location distribution domain is obtained by intersecting the converged set of layered projection vectors with the geometric boundary of the bearing tool.

[0071] In one specific implementation, when tracking a container, if the tracking signal of the target half-height container is interrupted, its location needs to be inferred by utilizing the adjacency relationships and stacking dependencies in the topological constraint graph.

[0072] Identify all reference containers directly adjacent to the target half-height container from a pre-constructed topology constraint graph. For example, for the target half-height container with ID "HLXU8012345", by retrieving the adjacency relationships in the topology constraint graph, the set of reference containers adjacent to it can be extracted, including the three containers "CMAU7654321" on the left, "BSIU3210987" on the right, and "MSDU4567890" above.

[0073] For each extracted reference container, its sampled position sequence data over the past 10 seconds is analyzed to calculate acceleration and directional stability characteristics. Acceleration characteristics are obtained by calculating the rate of change of position between adjacent time points; directional stability characteristics are measured by the consistency of the angle between the position change vectors. For example, the position data of the most recent 10 sampling points for reference container "CMAU7654321" shows an acceleration characteristic value of 0.05 m / s². 2 The speed is below the preset stability threshold of 0.1 m / s. 2 The directional stability eigenvalue is 0.92, which is higher than the preset consistency threshold of 0.85, therefore it is determined to be a stable reference and assigned a propagation weight of 1.0. The acceleration eigenvalue of the reference container "MSDU4567890" is 0.18 m / s². 2 The value exceeds the stability threshold, and the directional stability eigenvalue is 0.76, which is lower than the consistency threshold, indicating that non-rigid coupling motion may have occurred. The system adjusts its propagation weight to a decay value of 0.3.

[0074] Based on the stacking dependencies recorded in the topology constraint diagram, the position vector changes of each reference container are projected or masked. For example, for reference container "MSDU4567890" located above the target half-height container, its horizontal (XY plane) position change can be directly projected onto the target container, but the vertical (Z axis) change needs to consider stacking constraints. Specifically, if the displacement of "MSDU4567890" in the XY plane is (0.15m, 0.08m) and the Z-direction displacement is 0.02m, then only the XY plane displacement (0.15m, 0.08m) is projected onto the target container, while the Z-direction remains unchanged. For horizontally adjacent reference containers "CMAU7654321" and "BSIU3210987", their complete three-dimensional position change vectors can be directly projected, which are (0.12m, -0.05m, 0.00m) and (0.14m, -0.04m, 0.00m), respectively.

[0075] All projected position vectors are weighted and fused, with weights derived from the calculated propagation weights. Specifically, a weighted average is calculated for each dimension, and a weighted standard deviation is also calculated as an uncertainty measure. In this embodiment, the weighted average in the X direction is 0.134m, in the Y direction it is -0.019m, and in the Z direction it is 0.000m; the uncertainty measure value is 0.027, which is lower than the preset convergence threshold of 0.05, indicating that the information provided by the reference container in the current first-order adjacency relationship is sufficiently reliable.

[0076] If the uncertainty metric exceeds a preset convergence threshold, the process expands to second-order adjacency relationships, extracting extended reference containers adjacent to the first-order reference container. For example, if the uncertainty metric is 0.06, exceeding the threshold of 0.05, the container "TRLU2468013," adjacent to "CMAU7654321," is further extracted as a second-order reference. The propagation weight of the second-order reference container is calculated by multiplying its propagation weight by the propagation weight of the first-order reference container and then multiplying by a decay factor of 0.8; in this example, it is 0.8 × 1.0 = 0.8. The above projection and fusion steps are repeated until the uncertainty metric falls below the convergence threshold.

[0077] The converged target half-height container position vector is intersected with the geometric boundary of the carrying vehicle (such as a ship deck or dock area). Assuming the carrying vehicle is a container ship, its deck boundary is defined as a rectangular area with a length of 200 meters, a width of 35 meters, and a height limit of 25 meters, with the origin located on the port side of the aft deck. The estimated position of the target half-height container is (75.134m, 18.981m, 2.500m). After boundary constraints, this position is confirmed to be within the valid range. The spatial distribution domain of the target half-height container is determined as a three-dimensional ellipsoid centered at this point, considering uncertainty metrics, with a semi-axis length of (0.027m, 0.027m, 0.000m).

[0078] In this embodiment, even when the target half-height container signal is interrupted, the spatial distribution of the target container can be accurately inferred by utilizing topological constraints and motion information of the reference container, providing effective support for subsequent recovery and tracking.

[0079] In one optional implementation, the hierarchical projection vector set is weighted and fused, and an uncertainty metric is calculated. When the uncertainty metric exceeds a preset convergence threshold, an extended reference container for extracting second-order adjacency relationships is expanded, second-order propagation weights are calculated, and the container is projected to supplement the hierarchical projection vector set. This process is iterated until convergence, including:

[0080] For each projection vector in the layered projection vector set, a weighted sum is performed according to the propagation weight of the corresponding reference container to obtain the fused position vector;

[0081] Calculate the spatial deviation between each projection vector in the layered projection vector set and the fused position vector, and obtain an uncertainty measure based on the weighted variance of the spatial deviation;

[0082] When the uncertainty metric exceeds the preset convergence threshold, the node adjacent to any reference container in the reference container set is found from the topology constraint graph. After excluding the nodes already in the reference container set, the extended reference container set is obtained.

[0083] For each extended reference container in the extended reference container set, obtain the first propagation weight of the adjacent reference containers in the reference container set, and multiply the first propagation weight by the propagation weight of the adjacent reference containers to obtain the second-order propagation weight of the extended reference container.

[0084] The position vector change of the extended reference container is projected onto the constraint space of the target half-height container with second-order propagation weights, and the projection result is supplemented into the hierarchical projection vector set.

[0085] Repeat the process until the uncertainty metric converges below a preset convergence threshold.

[0086] In one specific implementation, a hierarchical projection vector set is obtained. This set contains multiple projection vectors, each corresponding to a reference container, and each reference container has a corresponding propagation weight. For each projection vector in the set, a weighted sum is calculated based on the propagation weight of its corresponding reference container. For example, assuming there are three reference containers A, B, and C with propagation weights of 0.5, 0.3, and 0.2 respectively, and corresponding projection vectors (10, 20, 30), (15, 25, 35), and (5, 15, 25) respectively, the fused position vector is calculated as (10, 20, 30) × 0.5 + (15, 25, 35) × 0.3 + (5, 15, 25) × 0.2 = (10.5, 21, 30.5).

[0087] After calculating the fused position vector, the reliability of this position estimate needs to be evaluated by calculating an uncertainty metric. Specifically, the spatial deviations between each projected vector and the fused position vector are calculated, and then the uncertainty metric is obtained based on the weighted variance of these deviations. Continuing the example above, the spatial deviations are calculated as follows: the deviation between vector A and the fused position vector is [(10-10.5)]. 2 +(20-21) 2 +(30-30.5) 2 ] 1 / 2 =0.87; The deviation between vector B and the fused position vector is [(15-10.5)]. 2 +(25-21) 2 +(35-30.5) 2 ] 1 / 2 =7.09; The deviation between vector C and the fused position vector is [(5-10.5)]. 2 +(15-21) 2 +(25-30.5) 2 ] 1 / 2 =8.95. Then, calculate the weighted variance according to their respective weights: 0.5 × 0.87 2 +0.3×7.09 2 +0.2×8.95 2 =26.8, the uncertainty measure is the square root of 26.8, which is approximately 5.18.

[0088] Assuming a preset convergence threshold of 3.0, since the current uncertainty metric of 5.18 is greater than the threshold of 3.0, the reference container set needs to be expanded. The node adjacent to any reference container in the reference container set is found in the topology constraint graph. After excluding nodes already in the reference container set, the expanded reference container set is obtained. For example, reference container A is adjacent to containers D and E, B is adjacent to containers F and G, and C is adjacent to container H. After excluding existing reference containers, the expanded reference container set is {D, E, F, G, H}.

[0089] For each extended reference container in the extended reference container set, obtain the first propagation weight of its neighboring reference containers in the set. Multiply the first propagation weight by the propagation weight of the neighboring reference containers to obtain the second-order propagation weight of the extended reference container. For example, assuming the first propagation weight of container D and A is 0.6, then the second-order propagation weight of D is 0.6 × 0.5 = 0.3; assuming the first propagation weight of container E and A is 0.4, then the second-order propagation weight of E is 0.4 × 0.5 = 0.2; similarly, if the first propagation weights of F, G, and B are calculated to be 0.7 and 0.3 respectively, then the second-order propagation weights of F and G are 0.7 × 0.3 = 0.21 and 0.3 × 0.3 = 0.09 respectively; if the first propagation weight of H and C is 0.8, then the second-order propagation weight of H is 0.8 × 0.2 = 0.16.

[0090] The position vector changes of the extended reference container are projected onto the constraint space of the target half-height container using second-order propagation weights, and the projection results are supplemented into the hierarchical projection vector set. For example, assuming the position vectors of extended reference containers D, E, F, G, and H are (12, 22, 32), (8, 18, 28), (16, 26, 36), (14, 24, 34), and (6, 16, 26) respectively, then their projections onto the constraint space of the target half-height container are as follows: the projection of D is (12, 22, 32) × 0.3 = (3.6, 6.6, 9.6), the projection of E is... The projection of F is (8, 18, 28) × 0.2 = (1.6, 3.6, 5.6), the projection of F is (16, 26, 36) × 0.21 = (3.36, 5.46, 7.56), the projection of G is (14, 24, 34) × 0.09 = (1.26, 2.16, 3.06), and the projection of H is (6, 16, 26) × 0.16 = (0.96, 2.56, 4.16).

[0091] The new projection vectors are added to the original hierarchical projection vector set to form the expanded hierarchical projection vector set {(10, 20, 30), (15, 25, 35), (5, 15, 25), (3.6, 6.6, 9.6), (1.6, 3.6, 5.6), (3.36, 5.46, 7.56), (1.26, 2.16, 3.06), (0.96, 2.56, 4.16)}, with the corresponding weight set being {0.5, 0.3, 0.2, 0.3, 0.2, 0.21, 0.09, 0.16}.

[0092] Re-execute the weighted fusion calculation to obtain the updated fused position vector, and calculate the new uncertainty metric. Iterate the above process until the uncertainty metric converges to below the preset convergence threshold. Assuming that after multiple iterations, the uncertainty metric drops to 2.8, which is below the threshold of 3.0, the iteration terminates, and the final fused position vector is output as the position estimation result of the target half-height container.

[0093] In this embodiment, by gradually expanding the range of reference containers under consideration and utilizing the topological constraints of second-order adjacency relationships, the accuracy and robustness of location estimation are effectively improved. This method is particularly suitable for positioning tasks in container yard environments and can overcome the problem of inaccurate positioning in complex environments by traditional methods.

[0094] In one optional implementation, node matching is performed on the topological constraint graphs before and after the replacement, non-invariant sets with positional changes are extracted, and a time-dependent directed graph is constructed based on the kinematic reachability cone and the load-bearing limit of vertical stacking of the handling equipment. Topological sorting is then performed to generate the replacement operation sequence, including:

[0095] Based on the initial and target topology constraint graphs determined before and after the transformation, node pairs are extracted through node matching.

[0096] Calculate the position coordinate difference between node pairs. When the difference exceeds the preset position invariance threshold, mark it as a node with a changed position and summarize it to obtain a non-invariant set.

[0097] For each node with a change in position, a kinematically reachable cone is constructed with the current position of the handling equipment as the vertex. When the spatial position of the node is not within the reachable cone, the shortest path is extracted and the occluded nodes on the path are identified. Predecessor dependency edges are established from the occluded nodes to the node with a change in position.

[0098] For each position-changing node in the non-invariant set, extract the upper load-bearing node of the vertical stack from the initial topology constraint graph, calculate the ratio of the mass of the upper load-bearing node to the load-bearing limit of the position-changing node, and when the ratio exceeds the preset load-bearing safety factor, establish a load-bearing dependency edge from the upper load-bearing node to the position-changing node.

[0099] The nodes with changes in position in the non-invariant set are used as nodes in the temporal dependent directed graph. The predecessor dependency edge and the bearing dependency edge are added to the temporal dependent directed graph. The topological sorting is performed to obtain the linear permutation sequence, and the replacement operation sequence is determined.

[0100] In one specific implementation, initial and target topology constraint diagrams are obtained before and after the transposition. The initial topology constraint diagram represents the current state of the object to be transpositioned, and the target topology constraint diagram represents the expected final state. Each topology constraint diagram contains multiple nodes, their spatial coordinates, and the constraint relationships between nodes. Each node represents an object, and node attributes include the object's three-dimensional coordinate position, size parameters, mass characteristics, and load-bearing capacity limits. Nodes in the initial and target topology constraint diagrams are matched to extract node pairs with the same identifier. For example, in a semi-high container multimodal transport scenario, a node identified as container number A12345 is located at coordinates (3, 4, 0) in the initial topology constraint diagram and at coordinates (8, 4, 0) in the target topology constraint diagram. These two nodes are matched as a node pair using the identifier A12345.

[0101] Calculate the difference between the position coordinates of each node in the initial and target topology constraint graphs to determine if the node's position has changed. A position invariance threshold of 0.5 meters is set; when the Euclidean distance of the coordinate difference exceeds this threshold, the node is marked as a position-changed node. For example, if a container moves from position (10, 15, 0) in the initial topology constraint graph to position (10, 15.3, 0) in the target topology constraint graph, the calculated Euclidean distance is 0.3 meters, which is less than the 0.5-meter threshold, thus indicating a position invariance. However, from the initial position (10, 15, 0) to the target position (10, 16, 0), the calculated Euclidean distance is 1 meter, which is greater than the 0.5-meter threshold, thus indicating a position-changed node. All position-changed nodes are aggregated into a non-invariant set. Specifically, the non-invariant set refers to the set of all nodes marked as having changed positions. Nodes in the non-invariant set need to be moved during the transshipment process.

[0102] For each position-changing node in the non-invariant set, a kinematic reachability cone is constructed with the current position of the handling equipment as its vertex. Specifically, the kinematic reachability cone refers to a three-dimensional spatial range constructed based on the technical parameters of the handling equipment (such as maximum reach radius, angle range, and height limit), with the current position of the handling equipment as its vertex. This range represents the spatial area that the handling equipment can directly reach from its current position. The kinematic reachability cone is determined based on the technical parameters of the handling equipment, including the maximum reach radius, angle range, and height limit. In practical applications, a kinematic reachability cone with a radius of 25 meters and a height of 15 meters can be constructed for the gantry crane equipment in a container yard. It is then determined whether the spatial position of the position-changing node in the target topology constraint graph is within the reachability cone range. If it is not within the reachability range, occlusion nodes need to be identified and predecessor dependencies established.

[0103] When the target location of a node undergoing a position change is outside the reachable cone, a breadth-first search algorithm is used to extract the shortest path from the current location of the transport equipment to the target location. The path search employs a three-dimensional grid discretization space, with a grid size of 0.5m × 0.5m × 0.5m, and each grid point records the passability status. Based on the position and volume information of nodes in the initial topology constraint graph, grid points occupied by objects are marked as impassable. Starting from the location of the transport equipment, the search expands along the passable grid points until a path to the target location is found or it is confirmed that no feasible path exists.

[0104] After determining the shortest path, identify the occupying nodes on the path. An occupying node is a node whose space in the initial topology constraint graph intersects with the shortest path. For example, when a handling device at position (5, 5, 0) needs to reach a container at position (20, 5, 0), if there is another container at position (12, 5, 0) on the path, that container is identified as an occupying node. For each occupying node, establish a predecessor dependency edge from the occupying node to the node whose position has changed, indicating that the occupying node must be moved before the node whose position has changed can be moved.

[0105] For each node with a changing position in the non-invariant set, extract the upper carrying node for vertical stacking from the initial topology constraint graph. Determine if a vertical stacking relationship exists between nodes by comparing their planar coordinates (x, y) and height coordinates (z). A vertical stacking relationship is confirmed when the difference between the planar coordinates of the upper node and the lower node is less than half the node size, and the height coordinate of the upper node is greater than that of the lower node. For example, in the initial topology constraint graph, a container located at (8, 6, 2.5) is below a container located at (8, 6, 0), both having the same planar coordinates and a height difference of 2.5 meters, thus confirming a vertical stacking relationship.

[0106] For the identified vertical stacking relationships, the ratio of the mass of the upper load-bearing node to the load-bearing limit of the node with changing position is calculated. A load-bearing safety factor is set to 0.85. When the ratio exceeds the safety factor, a load-bearing dependency edge is established from the upper load-bearing node to the node with changing position. For example, if the upper container's mass is 20 tons and the lower container's load-bearing limit is 25 tons, the calculated ratio is 0.8, which is less than the safety factor of 0.85, so no load-bearing dependency edge is established. However, if the upper container's mass is 22 tons, the calculated ratio is 0.88, which is greater than the safety factor of 0.85, so a load-bearing dependency edge is established.

[0107] After establishing the predecessor dependency edges and the carrier dependency edges, the nodes with position changes in the non-invariant set are used as nodes in the temporally dependent directed graph, and all dependency edges are added to this graph. The Kahn algorithm is used to perform topological sorting, generating a linear permutation sequence that satisfies all dependencies. This sequence is the optimized replacement operation sequence, ensuring that no physical conflicts or overload situations occur during the operation.

[0108] In one optional implementation, for each location-changed node, a kinematically reachable cone is constructed with the current position of the transport equipment as the vertex. When the spatial position of a node is not within the reachable cone, the shortest path is extracted and occluded nodes on the path are identified. The predecessor dependency edges from the occluded nodes to the location-changed nodes are established, including:

[0109] Using the current position of the handling equipment as the cone apex, the cone opening angle is determined based on the joint angle range of the handling equipment, and the cone height is determined based on the extension radius of the end effector. The cone geometry range of the kinematically reachable cone is constructed. The distance of the position change node from the cone apex to the distance and the deflection angle from the cone central axis are calculated in the initial topology constraint diagram. When the distance exceeds the cone height or the deflection angle exceeds the cone opening angle, it is determined that the position change node is not within the kinematically reachable cone range.

[0110] The initial topology constraint graph is converted into a spatial connectivity graph. In the spatial connectivity graph, the shortest path search is performed with the current position of the transport equipment as the starting point and the spatial position of the position change node as the ending point to obtain the path node sequence. The path node sequence is traversed, and the height difference between each path node and the position change node in the vertical direction is calculated. When the height difference is greater than the preset occlusion judgment threshold, the path node is marked as an occlusion node.

[0111] Create a predecessor dependency edge in a temporally dependent directed graph, pointing from an occluded node to a node whose position changes.

[0112] In one specific implementation, the handling device has a current position coordinate (x0, y0, z0) in the workspace, which serves as the apex of the kinematically reachable cone. Based on the robotic arm structural parameters of the handling device, its joint angle range is determined. For example, for a six-axis robotic arm, the shoulder joint angle range is ±170°, the elbow joint angle range is ±120°, and the wrist joint angle range is ±180°. The opening angle of the reachable cone is calculated based on these angle ranges. Typically, twice the minimum joint angle limit value is taken as the cone opening angle θ, which is 240° in this embodiment.

[0113] The maximum extension radius R of the end effector is determined based on the sum of the lengths of the links of each joint of the robotic arm; in this embodiment, R = 1.8 meters. The height h of the reachable cone is then determined, with a value of R = 1.8 meters. The geometric range of the reachable cone is constructed as a cone with (x0, y0, z0) as its vertex, an opening angle of θ = 240°, and a height of h = 1.8 meters.

[0114] For each position-changing node in the initial topology constraint graph, calculate its spatial position (x1, y1, z1) relative to the cone vertex, the distance d, and the deflection angle α relative to the cone's central axis. The distance d is calculated using the three-dimensional Euclidean distance formula, which is the straight-line distance between the position-changing node and the cone vertex. The cone's central axis direction is the default extension direction of the handling equipment's robotic arm, for example, taking a value of (0, 0, -1), representing vertically downwards. The deflection angle α is calculated as the angle between the line connecting the position-changing node and the cone vertex and the central axis direction.

[0115] When the calculated distance d exceeds the cone height h (i.e., d>1.8 meters) or the deflection angle α exceeds half of the cone opening angle (i.e., α>120°), it is determined that the node with the position change is not within the range of the kinematically reachable cone, and the occlusion relationship needs to be further processed.

[0116] The initial topological constraint graph is transformed into a spatial connectivity graph. Each node in the topological constraint graph retains its three-dimensional spatial coordinates, and the connections between nodes represent directly reachable paths in space. For example, for parts A and B in an assembly task, if they are adjacent in physical space and there are no obstacles in between, then connecting edges are established between the corresponding nodes in the spatial connectivity graph.

[0117] In the transformed spatial connectivity graph, starting from the current position of the transport equipment (x0, y0, z0) and ending at the position of a node whose location changes outside the reachable cone (x1, y1, z1), Dijkstra's shortest path search algorithm is executed. This algorithm iteratively calculates the shortest distance from the starting point to every node in the graph until the shortest path to the destination is found. The search result yields a sequence of path nodes P = (p0, p1, ..., p...). n ), where p0 is the starting point, pn The endpoint.

[0118] Traverse each node p in the path node sequence P i Calculate its relationship with the endpoint node p. n The height difference Δz = p in the vertical direction (i.e., the z-axis direction) i ·zp n .z. Set the occlusion detection threshold T = 0.15 meters. When the height difference Δz is greater than this threshold (i.e., Δz > 0.15 meters), the current path node p is... i Marked as an occluded node.

[0119] For example, if the current position of the handling equipment is (2.5, 3.0, 1.2), and the position of a node that changes location is (4.0, 2.0, 0.5), the calculated shortest path node sequence is [(2.5, 3.0, 1.2), (3.0, 2.8, 1.0), (3.5, 2.5, 0.8), (4.0, 2.0, 0.5)]. The height difference between the second node (3.0, 2.8, 1.0) and the endpoint (4.0, 2.0, 0.5) is 0.5 meters, which is greater than the threshold of 0.15 meters. Therefore, node (3.0, 2.8, 1.0) is marked as an occluded node. Similarly, the height difference of the third node (3.5, 2.5, 0.8) is 0.3 meters, and it is also marked as an occluded node.

[0120] In a temporally dependent directed graph, a predecessor dependency edge is created from each marked occluded node to a node whose position has changed. The dependency edge indicates that the occluded node must be processed or removed before the transport equipment can reach the node whose position has changed. For example, in the example above, a dependency edge is created from node (3.0, 2.8, 1.0) to the node whose position has changed (4.0, 2.0, 0.5), and a dependency edge is created from node (3.5, 2.5, 0.8) to the node whose position has changed (4.0, 2.0, 0.5).

[0121] This dependency ensures that during assembly, the handling equipment will first address obstructions before attempting to reach the obstructed node, thus avoiding collisions and improving the success rate of assembly tasks. An optimized assembly sequence is generated based on the constructed temporal dependency directed graph, effectively handling assembly challenges in complex spatial environments.

[0122] In this embodiment, the spatial occlusion relationship during the assembly process can be automatically identified, and a reasonable temporal dependency can be established, providing key technical support for the planning of automated assembly tasks.

[0123] like Figure 2 As shown, a flowchart illustrating the temporal dependency construction process based on kinematic reachability cone and spatial occlusion in semi-high container multimodal transport is presented.

[0124] In one optional implementation, for candidate locations in the spatial location distribution domain, stacking hierarchy features and neighborhood height features are extracted to obtain spatial stability indices. Physical adaptability verification is performed according to priority, and the locations are screened and determined based on remaining bearing capacity and overlapping area ratio, including:

[0125] For each candidate location in the spatial location distribution domain, the stacking level number of the candidate location in the topological constraint graph is extracted to determine the stacking level feature, the vertical height range of the nodes in the neighborhood of the candidate location is extracted to determine the neighborhood height feature, the stacking level feature and the neighborhood height feature are weighted and combined to obtain the spatial stability index, and the candidate locations are prioritized according to the spatial stability index.

[0126] Physical compatibility verification is performed on candidate locations in order of priority. The difference between the remaining load capacity of the load-bearing node below the candidate location and the mass of the half-height container is calculated. The overlap area ratio between the load-bearing node below the candidate location and the bottom of the half-height container is calculated. When the difference in remaining load capacity is positive and the overlap area ratio exceeds the preset area ratio threshold, the candidate location is marked as a confirmed location and the verification process for subsequent candidate locations is terminated.

[0127] In one specific implementation, the optimal stacking location for containers needs to be determined at the freight yard. The space to be arranged is divided into a spatial location distribution domain, which contains multiple candidate locations. Each candidate location represents a three-dimensional spatial coordinate point where a half-height container may be placed, described in meters. For example, a candidate location can be represented as coordinates (12.5, 8.0, 2.8), indicating a location point 12.5 meters laterally, 8.0 meters longitudinally, and 2.8 meters high in the freight yard.

[0128] For each candidate location in the spatial distribution domain, the stacking level number of the candidate location in the topological constraint graph is extracted to determine the stacking level feature. The topological constraint graph contains the spatial distribution and constraint relationships of all objects in the yard, where the stacking level number increases sequentially from the bottom layer. Objects directly in contact with the ground at the bottom layer have a stacking level of 1, objects placed on top of objects at the bottom layer have a stacking level of 2, and so on. For example, a candidate location with a stacking level of 3 means that the container at that location will be placed on top of the second layer of containers, becoming the third layer in the stack.

[0129] Stacking hierarchy features are closely related to spatial stability; the higher the stacking level, the worse the stability. To quantify this relationship, a stacking hierarchy influence factor is defined, converting the stacking hierarchy into a stability influence value. The influence factor values ​​were determined experimentally: the influence factor for level 1 is 0.1, for level 2 it is 0.3, for level 3 it is 0.6, and for level 4 it is 0.9. For example, a candidate position with a stacking hierarchy of 3 has a stacking hierarchy feature value of 0.6.

[0130] The vertical height feature of a candidate location is determined by extracting the vertical height range of nodes within its neighborhood. The neighborhood is defined as a cylindrical spatial region extending horizontally for 5 meters from the candidate location. Within this region, the top height values ​​of all object nodes are extracted, and the difference between the highest and lowest points is calculated as the vertical height range. For example, if there are three object nodes within the neighborhood of a candidate location with top heights of 0 meters (ground level), 2.8 meters, and 5.6 meters, the calculated vertical height range is 5.6 meters.

[0131] The vertical height range reflects the height non-uniformity within a neighborhood. The larger the range, the more drastic the height changes in the surrounding environment, and the worse the stability. The vertical height range is standardized to a value between 0 and 1 as a neighborhood height characteristic. The standardization method uses a piecewise function: when the range is less than 2 meters, the characteristic value is 0.1; when the range is between 2 and 6 meters, the characteristic value is (range - 2) / 4 × 0.6 + 0.1; and when the range is greater than 6 meters, the characteristic value is 0.7. In the example above, the range is 5.6 meters, and the calculated neighborhood height characteristic value is (5.6 - 2) / 4 × 0.6 + 0.1 = 0.64.

[0132] The spatial stability index is obtained by weighting the stacking hierarchy feature and the neighborhood height feature. The weight of the stacking hierarchy feature is 0.6, and the weight of the neighborhood height feature is 0.4. For example, if the stacking hierarchy feature value is 0.6 and the neighborhood height feature value is 0.64, the calculated spatial stability index is 0.6 × 0.6 + 0.4 × 0.64 = 0.616. The lower the spatial stability index, the better the stability of the location, and the more suitable it is for placing half-height containers.

[0133] Candidate locations are prioritized based on spatial stability indices, with lower indices indicating higher priority. In the sorted candidate location list, assuming the spatial stability indices for the first three locations are 0.32, 0.45, and 0.52 respectively, the physical fit of these three locations is then verified sequentially.

[0134] Physical compatibility verification is performed on candidate locations sequentially according to priority, calculating the difference between the remaining load-bearing capacity of the supporting node below the candidate location and the mass of the half-height container. A supporting node is an object node that directly supports the candidate location. The maximum load-bearing capacity and the currently loaded mass of the supporting node are extracted from the topology constraint graph; the difference between the two is the remaining load-bearing capacity. For example, if a supporting node has a maximum load-bearing capacity of 30 tons, currently carries 15 tons, and has a remaining load-bearing capacity of 15 tons, and the mass of the half-height container to be placed is 12 tons, the calculated difference is 15 - 12 = 3 tons.

[0135] Calculate the overlap area ratio between the load-bearing node below the candidate location and the bottom of the half-height container. The bottom area of ​​a standard half-height container is 14.8 square meters (6.1 meters × 2.43 meters). Calculate the overlap area between the load-bearing node and the container bottom through projection. If the overlap area is 12.5 square meters, then the overlap area ratio is 12.5 / 14.8 = 0.84. The preset area ratio threshold is 0.75, meaning that at least 75% of the bottom area needs to be supported to ensure stability.

[0136] When the remaining load-bearing capacity difference is positive and the overlap area ratio exceeds a preset area ratio threshold, the candidate location is marked as the confirmed location, and the verification process for subsequent candidate locations is terminated. In the example above, the remaining load-bearing capacity difference is positive (3 tons), and the overlap area ratio (0.84) is greater than the threshold (0.75). Therefore, this candidate location meets the physical compatibility requirements and is marked as the confirmed placement location for the half-height container.

[0137] If the first priority location does not meet the physical compatibility requirements, the second priority location is then verified, and so on, until a location that meets the requirements is found or all candidate locations have been verified. For example, if the remaining load-bearing capacity of the bearing node at the first priority location is 8 tons, and the mass of a half-height container is 12 tons, the difference is -4 tons, which does not meet the requirements; then the second priority location is verified, and its remaining load-bearing capacity of the bearing node is 14 tons, the difference is 2 tons, and the overlap area ratio is 0.82, which meets the requirements, and it is marked as the confirmed location.

[0138] The placement of the semi-high containers determined by the above methods takes into account both spatial stability and physical adaptability, which can effectively ensure the stacking safety of containers in multimodal transport, improve the efficiency of cargo yard space utilization, and reduce the risk of container damage.

[0139] In one optional implementation, a spatiotemporal alignment matrix is ​​constructed by combining the changing operation sequence with the determined position, and dynamic programming search is performed to obtain the spatiotemporally optimal alignment sequence, generating a continuous trajectory including:

[0140] Extract the timestamp of each costume change operation node from the costume change operation sequence, and extract the spatial coordinates of each specific location from the set of specific locations;

[0141] Construct a spatiotemporal alignment matrix, using timestamps as row indices and spatial coordinates as column indices. For each combination of a changeover operation node and a determined location, calculate the shortest path length in the topology constraint graph as the topological distance, and calculate the Euclidean distance between the spatial coordinates of the determined location and the corresponding spatial location of the changeover operation node as the spatial distance. Perform a weighted sum of the topological distance and the spatial distance to obtain the spatiotemporal matching cost, and fill the spatiotemporal alignment matrix with the spatiotemporal matching cost.

[0142] Dynamic programming search is performed on the spatiotemporal alignment matrix. Starting from the start timestamp of the costume change operation sequence, the determined positions that minimize the cumulative spatiotemporal matching cost are selected sequentially in time order. At the same time, the shortest path length between the currently selected determined position and the previously selected determined position in the initial topology constraint graph is constrained to be less than a preset reachable threshold, so as to obtain the spatiotemporally optimal alignment sequence.

[0143] By sequentially connecting the timestamps and spatial coordinates in the optimally aligned spatiotemporal sequence, a continuous trajectory is generated.

[0144] In one specific implementation, the timestamp of each transshipment operation node is extracted from the transshipment operation sequence. The transshipment operation sequence is an ordered list of operations generated according to topological sorting order. Each operation node contains information such as operation start time, operation type, operation object, and estimated completion time. For example, a semi-height container transshipment task includes 5 operation nodes, with their start timestamps being 08:15:30, 08:23:45, 08:37:20, 08:52:15, and 09:08:40 respectively. The timestamps are recorded in the format of hours:minutes:seconds, and can also be converted to seconds relative to a base time point for easy subsequent calculation. If the base time is set to 08:00:00, the above timestamps can be represented as 930 seconds, 1425 seconds, 2240 seconds, 3135 seconds, and 4120 seconds.

[0145] The spatial coordinates of each determined location are extracted from the set of determined locations. A determined location refers to a spot confirmed as suitable for placing a half-height container after spatial stability assessment and physical adaptability verification. Each determined location is represented by three-dimensional coordinates (x, y, z), in meters. For example, there are 6 determined locations within the yard, with spatial coordinates of (12.5, 8.0, 0.0), (18.3, 8.0, 0.0), (24.1, 8.0, 0.0), (12.5, 14.2, 0.0), (18.3, 14.2, 0.0), and (24.1, 14.2, 0.0). These coordinates represent container storage locations in different areas of the yard.

[0146] Construct a spatiotemporal alignment matrix, using timestamps as row indices and spatial coordinates as column indices. For the example of 5 timestamps and 6 spatial coordinates, construct a 5x6 matrix. For each combination of a transshipment operation node and a given location, calculate the shortest path length in the topology constraint graph as the topological distance. The topology constraint graph represents the spatial distribution and constraint relationships of objects within the yard. By searching for the shortest path in the graph, the shortest movement distance from the transshipment operation node to the given location is determined. For example, the topological distance from the operation node location (10.2, 5.5, 0.0) at timestamp 08:15:30 to the given location (12.5, 8.0, 0.0) is calculated to be 4.8 meters.

[0147] The spatial distance is calculated by taking the Euclidean distance between the spatial coordinates of the determined location and the corresponding spatial location of the costume change operation node. The Euclidean distance is calculated by taking the square root of the sum of the squares of the differences in the coordinates of the two points in each dimension. Using the coordinates in the example above, the spatial distance is 3.4 meters. The spatiotemporal matching cost is obtained by weighted summation of the topological distance and the spatial distance, with the topological distance weight set to 0.7 and the spatial distance weight set to 0.3. Continuing with the example above, the spatiotemporal matching cost is 0.7 × 4.8 + 0.3 × 3.4 = 4.38. The calculated spatiotemporal matching cost is then filled into the spatiotemporal alignment matrix, with the element in the first row and first column corresponding to the value 4.38.

[0148] The values ​​of all elements in the matrix are calculated using the same method. For example, the first row and second column represent the spatiotemporal matching cost between the operating node at timestamp 08:15:30 and the determined position (18.3, 8.0, 0.0). Assuming a topological distance of 10.6 meters and a spatial distance of 8.3 meters, the calculated spatiotemporal matching cost is 0.7 × 10.6 + 0.3 × 8.3 = 9.91. After completing all calculations, the complete spatiotemporal alignment matrix is ​​obtained.

[0149] A dynamic programming search is performed on the spatiotemporal alignment matrix, starting from the beginning timestamp of the costume change sequence, and sequentially selecting the positions that minimize the cumulative spatiotemporal matching cost. A cumulative cost array is defined, with a length equal to the number of positions. Initially, the values ​​in the cumulative cost array are set to the spatiotemporal matching costs of each position corresponding to the first timestamp. For example, the initial values ​​of the cumulative cost array are [4.38, 9.91, 15.42, 7.25, 12.84, 18.33].

[0150] For each subsequent timestamp, calculate the cumulative cost of moving from each defined location at the previous timestamp to each defined location at the current timestamp. Simultaneously, constrain the shortest path length between the currently selected defined location and the previously selected defined location in the initial topology constraint graph to be less than a preset reachability threshold. The reachability threshold is set to 20 meters, indicating that the container movement distance between adjacent time points does not exceed 20 meters, ensuring the continuity and feasibility of the trajectory.

[0151] For example, calculate the transition from timestamp 08:15:30 to 08:23:45. Assume the spatiotemporal matching cost of the determined location corresponding to the second timestamp is [8.12, 5.26, 9.73, 12.38, 7.45, 11.92]. For the first determined location of the second timestamp, calculate the cumulative cost of transitioning from each determined location of the first timestamp to this location and select the minimum value. Assume the shortest path length in the topological constraint graph from the first determined location of the first timestamp to the first determined location of the second timestamp is 6.8 meters, which is less than the reachability threshold of 20 meters, and the cumulative cost is 4.38 + 8.12 = 12.50. Similarly, calculate the cumulative cost of transitioning from other determined locations of the first timestamp to the first determined location of the second timestamp and select the minimum value as the cumulative cost of the first determined location of the second timestamp.

[0152] Repeat the above calculation for all determined positions of the second timestamp to obtain the updated cumulative cost array. Simultaneously, record the optimal predecessor position for each determined position, i.e., the determined position of the previous timestamp that generated the minimum cumulative cost. Continue performing the same operation for subsequent timestamps until the last timestamp has been processed.

[0153] After processing all timestamps, find the minimum value in the cumulative cost array corresponding to the last timestamp. The determined position corresponding to this value is the optimal position of the last timestamp. Based on the recorded predecessor information, backtrack from the end to the beginning to obtain the optimal determined position of each timestamp, forming a spatiotemporally optimal alignment sequence. For example, the obtained optimal alignment sequence is the determined position [(12.5, 8.0, 0.0), (18.3, 8.0, 0.0), (18.3, 14.2, 0.0), (24.1, 14.2, 0.0), (24.1, 8.0, 0.0)] corresponding to the timestamps [08:15:30, 08:23:45, 08:37:20, 08:52:15, 09:08:40].

[0154] A continuous trajectory is generated by sequentially concatenating the timestamps and spatial coordinates in the optimally aligned spatiotemporal sequence. This continuous trajectory represents the movement path of a semi-high container during multimodal transport, containing both temporal and spatial information. To enhance the continuity of the trajectory, intermediate points can be inserted between adjacent timestamps and spatial coordinates. For example, between 08:15:30 and 08:23:45, two points, 08:18:15 and 08:21:00, are inserted at equal intervals, and their corresponding spatial coordinates are obtained through linear interpolation [(14.4, 8.0, 0.0), (16.3, 8.0, 0.0)]. The complete continuous trajectory contains the original 5 key points and the interpolated 8 intermediate points, totaling 13 spatiotemporal points, describing the complete movement process of the semi-high container within the freight yard.

[0155] The semi-high container trajectory generated by the above method combines the transshipment operation sequence and location information, takes into account topological constraints and spatial distance, and ensures the continuity and optimality of the trajectory through dynamic programming optimization, providing accurate location information for the whole-process monitoring and tracking of semi-high container multimodal transport.

[0156] The semi-high container multimodal transport end-to-end monitoring and tracking system of this invention includes:

[0157] The first unit is used to collect spatial layout data of containers and construct a topological constraint graph. The topological constraint graph represents containers through nodes and represents spatial adjacency and stacking dependency relationships through edges.

[0158] The second unit is used to extract reference containers adjacent to the target half-height container from the topology constraint graph when the tracking signal of the target half-height container is interrupted. Based on the stacking dependency, the position vector change of the reference containers is projected onto the constraint space of the target half-height container, and the weighted fusion is iterated until convergence. Combined with the geometric boundary of the carrying tool, a spatial position distribution domain is formed.

[0159] The third unit is used to perform node matching on the topological constraint graphs before and after the replacement, extract the non-invariant set of position changes, construct a time-dependent directed graph based on the kinematic reachability cone and the bearing limit of vertical stacking of the handling equipment, and perform topological sorting to generate the replacement operation sequence.

[0160] The fourth unit is used to extract stacking hierarchy features and neighborhood height features from candidate locations in the spatial location distribution domain to obtain spatial stability indicators, perform physical adaptability verification according to priority, and screen and determine locations based on remaining bearing capacity and overlapping area ratio.

[0161] The fifth unit is used to construct a spatiotemporal alignment matrix by combining the changing operation sequence with the determined position, perform dynamic programming search to obtain the spatiotemporally optimal alignment sequence, and generate a continuous trajectory.

[0162] A third aspect of the present invention provides an electronic device, comprising:

[0163] processor;

[0164] Memory used to store processor-executable instructions;

[0165] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0166] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0167] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for full-process monitoring and tracking of semi-high container multimodal transport, characterized in that, include: Collect spatial layout data of containers and construct a topological constraint graph. The topological constraint graph represents containers through nodes and represents spatial adjacency and stacking dependency relationships through edges. When the tracking signal of the target half-height container is interrupted, the reference container adjacent to the target half-height container is extracted from the topology constraint graph. The position vector change of the reference container is projected onto the constraint space of the target half-height container according to the stacking dependency relationship. The weighted fusion is iterated until convergence, and the spatial position distribution domain is formed by combining the geometric boundary of the carrying tool. Node matching is performed on the topological constraint graphs before and after the refit, and the non-invariant set of positional changes is extracted. Based on the kinematic reachability cone and bearing limit of the vertical stacking of the handling equipment, a temporal dependent directed graph is constructed and topological sorting is performed to generate a refit operation sequence, including: Based on the initial and target topology constraint graphs determined before and after the transformation, node pairs are extracted by matching nodes. Calculate the position coordinate difference between node pairs. When the difference exceeds the preset position invariance threshold, mark it as a node with a changed position and summarize it to obtain a non-invariant set. For each node with a change in position, a kinematically reachable cone is constructed with the current position of the handling equipment as the vertex. When the spatial position of the node is not within the reachable cone, the shortest path is extracted and the occluded nodes on the path are identified. Predecessor dependency edges are established from the occluded nodes to the node with a change in position. For each position-changing node in the non-invariant set, extract the upper load-bearing node of the vertical stack from the initial topology constraint graph, calculate the ratio of the mass of the upper load-bearing node to the load-bearing limit of the position-changing node, and when the ratio exceeds the preset load-bearing safety factor, establish a load-bearing dependency edge from the upper load-bearing node to the position-changing node. The nodes with changes in position in the non-invariant set are used as nodes in the temporal dependent directed graph. The predecessor dependency edge and the bearing dependency edge are added to the temporal dependent directed graph. The topological sorting is performed to obtain the linear permutation sequence, and the replacement operation sequence is determined. For candidate locations in the spatial location distribution domain, stacking hierarchy features and neighborhood height features are extracted to obtain spatial stability indicators. Physical adaptability verification is performed according to priority, and locations are selected and determined based on remaining bearing capacity and overlapping area ratio. The change-up sequence is used to construct a spatiotemporal alignment matrix with the determined position. Dynamic programming is then used to search for the optimal spatiotemporal alignment sequence, generating a continuous trajectory.

2. The method according to claim 1, characterized in that, When the tracking signal of the target half-height container is interrupted, reference containers adjacent to the target half-height container are extracted from the topology constraint graph. Based on the stacking dependency, the position vector changes of the reference containers are projected onto the constraint space of the target half-height container. Weighted fusion iterations are then performed until convergence. Combined with the geometric boundaries of the carrying vehicle, a spatial location distribution domain is formed, including: Extract the set of reference containers adjacent to the target half-height container from the topology constraint graph; Based on the position sequence of the reference container at multiple historical moments, the acceleration characteristics and directional stability characteristics are calculated. When the acceleration characteristics exceed the preset stability threshold or the directional stability characteristics are lower than the preset consistency threshold, it is determined that the reference container has undergone non-rigid coupling motion. The propagation weight of the reference container is adjusted to the attenuation weight value to obtain a weighted reference container set. Based on the stacking dependency between the reference container and the target half-height container in the weighted reference container set, the vertical position vector change is projected or masked to obtain a layered projection vector set. The hierarchical projection vector set is weighted and fused, and the uncertainty metric is calculated. When the uncertainty metric exceeds the preset convergence threshold, the extended reference container with second-order adjacency relationship is extracted, the second-order propagation weight is calculated and projected to supplement the hierarchical projection vector set, and the process is iterated until convergence. The spatial location distribution domain is obtained by intersecting the converged set of layered projection vectors with the geometric boundary of the bearing tool.

3. The method according to claim 2, characterized in that, The hierarchical projection vector set is weighted and fused, and an uncertainty metric is calculated. When the uncertainty metric exceeds a preset convergence threshold, an extended reference container with extracted second-order adjacency relationships is added, second-order propagation weights are calculated, and the container is projected back into the hierarchical projection vector set. This process is iterated until convergence, including: For each projection vector in the layered projection vector set, a weighted sum is performed according to the propagation weight of the corresponding reference container to obtain the fused position vector; Calculate the spatial deviation between each projection vector in the layered projection vector set and the fused position vector, and obtain an uncertainty measure based on the weighted variance of the spatial deviation; When the uncertainty metric exceeds the preset convergence threshold, the node adjacent to any reference container in the reference container set is found from the topology constraint graph. After excluding the nodes already in the reference container set, the extended reference container set is obtained. For each extended reference container in the extended reference container set, obtain the first propagation weight of the adjacent reference containers in the reference container set, and multiply the first propagation weight by the propagation weight of the adjacent reference containers to obtain the second-order propagation weight of the extended reference container. The position vector change of the extended reference container is projected onto the constraint space of the target half-height container with second-order propagation weights, and the projection result is supplemented into the hierarchical projection vector set. Repeat the process until the uncertainty metric converges below a preset convergence threshold.

4. The method according to claim 1, characterized in that, For each node with a changed position, a kinematically reachable cone is constructed with the current position of the transport equipment as its vertex. When the node's spatial position is outside the reachable cone, the shortest path is extracted, and occluded nodes on the path are identified. Predecessor dependencies from occluded nodes to the node with a changed position are established, including: Using the current position of the handling equipment as the cone apex, the cone opening angle is determined based on the joint angle range of the handling equipment, and the cone height is determined based on the extension radius of the end effector. The cone geometry range of the kinematically reachable cone is constructed. The distance of the position change node from the cone apex to the distance and the deflection angle from the cone central axis are calculated in the initial topology constraint diagram. When the distance exceeds the cone height or the deflection angle exceeds the cone opening angle, it is determined that the position change node is not within the kinematically reachable cone range. The initial topology constraint graph is converted into a spatial connectivity graph. In the spatial connectivity graph, the shortest path search is performed with the current position of the transport equipment as the starting point and the spatial position of the position change node as the ending point to obtain the path node sequence. The path node sequence is traversed, and the height difference between each path node and the position change node in the vertical direction is calculated. When the height difference is greater than the preset occlusion judgment threshold, the path node is marked as an occlusion node. Create a predecessor dependency edge in a temporally dependent directed graph, pointing from an occluded node to a node whose position changes.

5. The method according to claim 1, characterized in that, For candidate locations in the spatial distribution domain, stacking hierarchy features and neighborhood height features are extracted to obtain spatial stability indices. Physical compatibility verification is performed according to priority. Locations are selected and determined based on remaining bearing capacity and overlapping area ratio, including: For each candidate location in the spatial location distribution domain, the stacking level number of the candidate location in the topological constraint graph is extracted to determine the stacking level feature, the vertical height range of the nodes in the neighborhood of the candidate location is extracted to determine the neighborhood height feature, the stacking level feature and the neighborhood height feature are weighted and combined to obtain the spatial stability index, and the candidate locations are prioritized according to the spatial stability index. Physical compatibility verification is performed on candidate locations in order of priority. The difference between the remaining load capacity of the load-bearing node below the candidate location and the mass of the half-height container is calculated. The overlap area ratio between the load-bearing node below the candidate location and the bottom of the half-height container is calculated. When the difference in remaining load capacity is positive and the overlap area ratio exceeds the preset area ratio threshold, the candidate location is marked as a confirmed location and the verification process for subsequent candidate locations is terminated.

6. The method according to claim 1, characterized in that, Construct a spatiotemporal alignment matrix by combining the changing operation sequence with the determined position, perform dynamic programming search to obtain the spatiotemporally optimal alignment sequence, and generate a continuous trajectory including: Extract the timestamp of each costume change operation node from the costume change operation sequence, and extract the spatial coordinates of each specific location from the set of specific locations; Construct a spatiotemporal alignment matrix, using timestamps as row indices and spatial coordinates as column indices. For each combination of a changeover operation node and a determined location, calculate the shortest path length in the topology constraint graph as the topological distance, and calculate the Euclidean distance between the spatial coordinates of the determined location and the corresponding spatial location of the changeover operation node as the spatial distance. Perform a weighted sum of the topological distance and the spatial distance to obtain the spatiotemporal matching cost, and fill the spatiotemporal alignment matrix with the spatiotemporal matching cost. Dynamic programming search is performed on the spatiotemporal alignment matrix. Starting from the start timestamp of the costume change operation sequence, the determined positions that minimize the cumulative spatiotemporal matching cost are selected sequentially in time order. At the same time, the shortest path length between the currently selected determined position and the previously selected determined position in the initial topology constraint graph is constrained to be less than a preset reachable threshold, so as to obtain the spatiotemporally optimal alignment sequence. By sequentially connecting the timestamps and spatial coordinates in the optimally aligned spatiotemporal sequence, a continuous trajectory is generated.

7. A semi-high container multimodal transport end-to-end monitoring and tracking system, used to implement the method of any one of claims 1-6, characterized in that, include: The first unit is used to collect spatial layout data of containers and construct a topological constraint graph. The topological constraint graph represents containers through nodes and represents spatial adjacency and stacking dependency relationships through edges. The second unit is used to extract reference containers adjacent to the target half-height container from the topology constraint graph when the tracking signal of the target half-height container is interrupted. Based on the stacking dependency, the position vector change of the reference containers is projected onto the constraint space of the target half-height container, and the weighted fusion is iterated until convergence. Combined with the geometric boundary of the carrying tool, a spatial position distribution domain is formed. The third unit is used to perform node matching on the topological constraint graphs before and after the replacement, extract the non-invariant set of position changes, construct a time-dependent directed graph based on the kinematic reachability cone and the bearing limit of vertical stacking of the handling equipment, and perform topological sorting to generate the replacement operation sequence. The fourth unit is used to extract stacking hierarchy features and neighborhood height features from candidate locations in the spatial location distribution domain to obtain spatial stability indicators, perform physical adaptability verification according to priority, and screen and determine locations based on remaining bearing capacity and overlapping area ratio. The fifth unit is used to construct a spatiotemporal alignment matrix between the changing operation sequence and the determined position, perform dynamic programming search to obtain the spatiotemporally optimal alignment sequence, and generate a continuous trajectory.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images