A fast rerouting method based on bidirectional potential field

Abstract

The application discloses a fast re-routing method based on bidirectional potential energy field, comprising the following steps: constructing a forward potential energy field and a reverse potential energy field based on destination nodes in a network; generating a data packet forwarding rule according to the forward potential energy field and the reverse potential energy field; and switching between the forward potential energy field and the reverse potential energy field according to a link state in a data packet forwarding process. The application can realize efficient, stable and robust fault recovery in a complex network environment.

Description

Technical Field

[0001] This invention belongs to the field of network fault tolerance and fast rerouting technology, and particularly relates to a fast rerouting method based on a bidirectional potential field. Background Technology

[0002] With the rapid development of modern network infrastructure, networks are widely used in latency-sensitive and mission-critical scenarios, such as military communications, financial transactions, emergency command, cloud computing, and industrial control. These applications place extremely stringent requirements on the high reliability and low latency of networks. However, link or node failures are a common phenomenon in network operation. Existing experimental studies have shown that in data centers and backbone networks, link failures may occur on average every few minutes, with more than 30% of failures triggering network-wide route reconvergence. Traditional distributed routing protocols (such as OSPF / IS-IS or BGP) typically rely on a complex and time-consuming convergence process to cope with failures, often requiring several seconds or even tens of seconds to recover. This recovery latency is fundamentally contradictory to the stringent requirements of mission-critical services for communication continuity. To address this, the Fast Reroute (FRR) mechanism has been proposed in existing technologies. The core objective of this mechanism is to quickly switch service traffic to a backup path after the primary path fails, thereby avoiding reliance on global routing protocols for reconvergence.

[0003] Currently, the industry has proposed two main types of Fast Reroute (FRR) mechanisms to address the need for rapid recovery from link or node failures: proactive and reactive approaches. Both methods have achieved some success in theory and practice, but they still have insurmountable limitations when facing complex heterogeneous network environments. Proactive approaches pre-compute and configure multiple backup paths during normal network operation, with the most typical method being a pre-computation strategy based on k-disjoint arborescences. Theoretically, this method can provide protection against the failure of k-1 links in a k-connected network. However, this type of approach has two main problems: first, extremely high computational complexity. Constructing disjoint arborescences is a typical NP-hard problem; as the network size increases, its computational overhead grows exponentially, leading to excessively long pre-computation times and severely limiting scalability; second, strong structural rigidity. This approach relies on high network connectivity. However, in real-world heterogeneous network environments, many nodes have limited outgoing degrees, making them vulnerable bottleneck nodes. This makes it difficult to meet the requirements for constructing disjoint paths, ultimately resulting in insufficient backup coverage.

[0004] Reactive approaches take a different approach, relying not on pre-calculation but dynamically discovering and constructing alternative paths when a failure occurs. Typical examples include Failure-Carrying Packets (FCP) and Data-Driven Connectivity (DDC). These methods achieve real-time path reconstruction by carrying fault information in packets or rapidly propagating state changes through the data plane. While offering the advantage of rapid response, they also have two drawbacks: First, they lack global optimization capabilities. Relying solely on local node information for path selection can easily lead to excessively long paths, resulting in suboptimal or even inefficient routes. Second, they are prone to loops or oscillations. In multi-fault or dynamic topology environments, the independent local decisions of different nodes may conflict, causing packets to continuously loop or switch repeatedly within the network, ultimately leading to a loss of reachability. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a fast rerouting method based on a bidirectional potential field, which can achieve efficient, stable, and robust fault recovery in complex network environments.

[0006] To achieve the above objectives, the present invention provides a fast rerouting method based on a bidirectional potential field, comprising:

[0007] Construct positive and negative potential energy fields based on the target nodes in the network;

[0008] Based on the positive potential field and the negative potential field, a forwarding rule for the data packet is generated;

[0009] During packet forwarding, the system switches between the positive potential field and the negative potential field based on the link status.

[0010] Optionally, the process of constructing the positive and negative potential energy fields includes:

[0011] Calculate the number of hops from all nodes in the network to the destination node as the potential value;

[0012] Based on the potential difference between adjacent nodes, neighbors are divided into forward neighbors, backward neighbors, or equipotential neighbors.

[0013] Orient equipotential edges to increase the number of intersection nodes;

[0014] The positive potential energy field is formed by the descending edge and the equipotential edge after orientation;

[0015] By reversing the sides of the positive potential field, a reverse potential field is obtained.

[0016] Optionally, the potential value can be replaced by other network metrics, which are any one of link latency, link bandwidth, and link reliability, or any combination including hop count.

[0017] Optionally, the process of orienting equipotential edges includes determining the direction of the equipotential edges with the goal of reducing the number of pure source nodes and pure sink nodes and increasing the number of cross nodes.

[0018] Optionally, the process of generating forwarding rules for data packets includes: calculating weights for the edges in the forward potential field and the reverse potential field; and generating next-hop selection rules based on the weights.

[0019] Optionally, the next-hop selection rule includes:

[0020] In a forward potential field, the next hop is selected from the forward neighbor nodes with higher out-degrees; in a reverse potential field, the next hop is selected from the nodes that can lead to more forward choices.

[0021] Optionally, the process of switching between the positive and negative potential energy fields includes:

[0022] If the current node has a usable forward neighbor, the data packet moves forward in the positive potential field;

[0023] If all forward neighbors are unavailable and there are available reverse neighbors, the data packet switches to the reverse potential field and backs off.

[0024] When a packet retreats to a cross node in the reverse potential field and that node has an available forward neighbor that does not pass through the faulty region, the packet switches back to the forward potential field to proceed.

[0025] Optionally, the process of switching back to the positive potential field after the data packet falls back to the cross node includes:

[0026] When backtracking to a certain cross node and finding a forward neighbor that does not pass through the original fault region, the data packet switches back to the positive potential field and continues to move towards the target node.

[0027] Optionally, the method further includes:

[0028] A status flag is added to the packet header to record the current state of the packet and whether it is in a fallback context; each node makes forwarding decisions based on the pre-generated forwarding table and the status flag.

[0029] Technical effects of the invention:

[0030] (1) Reduced computation and storage overhead: Although the present invention still requires the establishment of auxiliary structures in the pre-computation stage, unlike existing methods, the computation process of the present invention only relies on simple indicators (such as hop count) and local node degree information, without involving complex combinatorial optimization problems. Its overall complexity is at the polynomial level, which can run efficiently in large-scale networks, thereby significantly reducing computation and storage overhead and achieving a lower cost and higher efficiency pre-computation process.

[0031] (2) Improved Applicability and Robustness: Traditional methods require each node to retain at least one backup path under any failure scenario (i.e., the so-called "strong reachability condition"). However, in real-world networks, many nodes have very low out-degrees, making it impossible to meet this condition and thus becoming vulnerable. This invention relaxes the reachability requirement from "each node must have a backup path" to "a node only needs to be able to reach a cross node with an out-degree greater than 1 through the reverse potential field" through a bidirectional potential field architecture. Cross nodes, having multiple available forwarding options, can provide alternative paths for traffic after a failure. This design makes the system no longer dependent on the high out-degree of all nodes, but instead relies on the redundancy of cross nodes, thus maintaining high fault tolerance and robustness even in heterogeneous network environments.

[0032] (3) Ensure path quality and fast convergence: When a fault occurs, the present invention achieves fast local repair without loops by coordinating the switching of positive and negative potential energy fields, so that the recovery path is close to the optimal one, ensuring high-efficiency and low-latency end-to-end transmission.

[0033] (4) Improve network stability and continuity: This invention introduces strict loop-free rules during the forwarding and switching process, effectively avoiding path oscillation and loop problems, ensuring stable operation in multi-fault environments, and meeting the application requirements of high reliability and strong continuity. Attached Figure Description

[0034] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0035] Figure 1 This is a flowchart illustrating the pre-calculation process of the bidirectional potential energy field in an embodiment of the present invention.

[0036] Figure 2 This is a schematic diagram of the packet forwarding flow of a single node on the path based on a bidirectional potential field in an embodiment of the present invention;

[0037] Figure 3 This is an example diagram illustrating path recovery of data flow after a link failure, according to an embodiment of the present invention.

[0038] Figure 4This is a flowchart illustrating a fast rerouting method based on a bidirectional potential field according to an embodiment of the present invention. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0041] While existing fast rerouting technologies alleviate network outages caused by link or node failures to some extent, they still have the following shortcomings: High computational overhead and low efficiency: Proactive pre-computation schemes (such as those based on disjoint arc trees or integer linear programming) often require complex calculations across the entire network. Their time and storage costs increase dramatically with network size, making them practically unusable in large-scale networks, resulting in high cost, low efficiency, and long processing times. High structural rigidity and poor applicability: Proactive methods typically require high connectivity in the network topology, meaning each node must have sufficient backup paths. However, in real-world heterogeneous networks, many nodes have relatively few connections, making it difficult to meet this stringent condition. Consequently, these methods lack robustness in complex environments, struggle to cover multiple failure points, and suffer from poor applicability and instability. Poor path quality and lack of global optimization: While reactive schemes can respond quickly to failures, their path selection relies entirely on local node information, lacking global coordination. This often leads to excessively long alternative paths or even unnecessary detours, resulting in low path efficiency, high communication latency, and difficulty in meeting the performance requirements of mission-critical applications. Insufficient stability, prone to loops or oscillations: In the event of multiple failures or frequent changes in network topology, the locally independent decisions upon which reactive solutions rely may conflict with each other, leading to forwarding loops or repeated oscillations of data packets within the network. This not only increases the network load but may also cause data to fail to reach its destination entirely, resulting in low reliability and poor service continuity.

[0042] like Figure 4 As shown, this embodiment provides a fast rerouting method based on a bidirectional potential field, including: constructing a forward potential field and a reverse potential field based on the destination node in the network;

[0043] Based on the positive potential field and the negative potential field, a forwarding rule for the data packet is generated;

[0044] During packet forwarding, the system switches between the positive potential field and the negative potential field based on the link status.

[0045] Furthermore, the process of constructing the positive and negative potential energy fields includes:

[0046] Calculate the number of hops from all nodes in the network to the destination node as the potential value;

[0047] Based on the potential difference between adjacent nodes, neighbors are divided into forward neighbors, backward neighbors, or equipotential neighbors.

[0048] Orient equipotential edges to increase the number of intersection nodes;

[0049] The positive potential energy field is formed by the descending edge and the equipotential edge after orientation;

[0050] By reversing the sides of the positive potential field, a reverse potential field is obtained.

[0051] Furthermore, the potential value can be replaced by other network metrics, which are any one of link latency, link bandwidth, and link reliability, or any combination including hop count.

[0052] Furthermore, the process of orienting equipotential edges includes determining the direction of the equipotential edges with the goal of reducing the number of pure source nodes and pure sink nodes and increasing the number of cross nodes.

[0053] Furthermore, the process of generating packet forwarding rules includes: calculating weights for the edges in the forward potential field and the reverse potential field; and generating next-hop selection rules based on the weights.

[0054] Furthermore, the next-hop selection rule includes:

[0055] In a forward potential field, the next hop is selected from the forward neighbor nodes with higher out-degrees; in a reverse potential field, the next hop is selected from the nodes that can lead to more forward choices.

[0056] Furthermore, the process of switching between the positive and negative potential energy fields includes:

[0057] If the current node has a usable forward neighbor, the data packet moves forward in the positive potential field;

[0058] If all forward neighbors are unavailable and there are available reverse neighbors, the data packet switches to the reverse potential field and backs off.

[0059] When a packet retreats to a cross node in the reverse potential field and that node has an available forward neighbor that does not pass through the faulty region, the packet switches back to the forward potential field to proceed.

[0060] Furthermore, the process of switching back to the positive potential field after the data packet falls back to the cross node includes:

[0061] When backtracking to a certain cross node and finding a forward neighbor that does not pass through the original fault region, the data packet switches back to the positive potential field and continues to move towards the target node.

[0062] Furthermore, the method also includes:

[0063] A status flag is added to the packet header to record the current state of the packet and whether it is in a fallback context; each node makes forwarding decisions based on the pre-generated forwarding table and the status flag.

[0064] Specifically, the implementation process of this embodiment includes:

[0065] This invention proposes a fast rerouting method and apparatus based on a bidirectional potential field (DPF). In the pre-computation phase, the potential gradient is calculated based on the "shortest hop count" potential function, and the adjacency set of each node is divided into forward neighbors, backward neighbors, and equipotential neighbors. "Equipotential edge orientation" is implemented on equipotential edges to increase the number of cross nodes (nodes with multiple alternative forwarding branches), constructing a forward potential field (FPF) and a reverse potential field (RPF), along with corresponding link weights and next-hop selection rules. During the data forwarding phase, data packets preferentially proceed along the direction of decreasing potential energy in the FPF. When forward traffic is blocked (no available forward neighbors locally), the system switches to the RPF for backtracking according to the rules until a cross node with better forwarding distribution capabilities is reached, then resumes forwarding in the FPF. This dual-field coordination ensures loop-free, fast convergence, and controlled path elongation, significantly improving fault tolerance and arrival rate in heterogeneous networks.

[0066] The technical solution of this invention mainly consists of the following key components, each of which works together and none can be omitted:

[0067] (1) Potential Function Unit: This unit is used to characterize the topological distance between any node and the target node. In implementation, the shortest hop count from the node to the target node can be used as the potential value. The smaller the potential value, the closer the node is to the target; the potential value of the target node is fixed at zero. The potential function provides a basic metric for subsequent forwarding direction determination and field construction.

[0068] (2) Potential Gradient and Neighbor Classification Unit: This unit is used to determine the directional attribute of a link by comparing the potential differences between neighboring nodes. If the potential value of a neighboring node is lower, the neighbor is a forward neighbor and is suitable as a forwarding direction; if the potential value of a neighboring node is higher, the neighbor is a backward neighbor and is suitable for selection during backoff; if the potential values ​​of neighboring nodes are equal, the neighbor is an equipotential neighbor. Through this classification, each node in the network can clearly distinguish between forward, backoff, and equipotential relationships.

[0069] (3) Equipotential Edge Processing Unit: In traditional unidirectional potential energy fields, equipotential edges typically do not participate in forwarding, easily leading to resource waste. This invention proposes to orient equipotential edges to increase the number of cross nodes in the network and reduce pure source / pure sink nodes (source node: a node with an in-degree of zero in the positive potential energy field; sink node: a node with an out-degree of zero in the positive potential energy field (the target node is usually the sink); cross node: a node with an out-degree greater than 1 in the positive potential energy field, which can provide multiple forwarding options for data packets and is key to robust recovery). By comparing the out-degree and in-degree of nodes, the direction of equipotential edges is rationally selected to increase the number of cross nodes in the network.

[0070] (4) Bidirectional Potential Field Structure: This is the core structure of the invention, comprising two potential fields: a forward potential field (FPF) and a reverse potential field (RPF). The forward potential field (FPF) consists of edges with decreasing potential and equipotential edges after orientation, used by data packets under normal conditions to ensure forward movement along the direction of decreasing potential. The reverse potential field (RPF) is obtained by reversing all edges of the forward potential field, used by data packets when the forward path is blocked to achieve backtracking from the blockage point and finding a new intersection node. The forward and reverse potential fields complement each other, jointly providing the ability to move forward and backward.

[0071] (5) Weight Calculation and Next Hop Selection Unit: To avoid suboptimal paths under a single metric, this invention introduces a weighting rule when selecting the next hop: In the forward potential field, forward neighbor nodes with higher out-degrees are prioritized to ensure that the path has more subsequent available options and reduce the risk of interruption due to single-link failure; in the reverse potential field, backtracking paths that can lead to more forward options are prioritized, thereby providing the best conditions for resuming forward progress. This design makes each forwarding step tend to select nodes with more redundancy, thus improving overall recoverability.

[0072] (6) Inter-field handover control unit: This unit is responsible for switching between the forward and reverse potential energy fields during forwarding, based on link availability and node status. Forward condition: If the current node has an available forward neighbor, the data packet continues to advance in the forward potential energy field; Back-off condition: If all forward neighbors are unavailable, but there is an available reverse neighbor, the data packet switches to the reverse potential energy field and advances along the back-off path; Recovery condition: When back-off to a certain cross node, if a forward neighbor that does not pass through the original fault area is found, the data packet switches back to the forward potential energy field and continues to advance towards the target node. This design ensures that the data packet can back off in time when the forward direction is blocked and resume its forward movement at an appropriate position, thereby avoiding deadlock and loops.

[0073] (7) Status Marking and Forwarding Table Unit: To ensure consistent forwarding logic when data packets are transmitted between different nodes, this invention designs a lightweight status recording method: adding a small number of flag bits to the header of the data packet to record the current field (forward / reverse) and whether it is in a fault fallback context. Each node generates a forwarding table during the pre-computation stage, compiling the potential energy field structure, weight rules, and switching conditions into table entries; during forwarding, the node only needs to look up the status marking and forwarding table to complete the decision, and the lookup complexity is O(1).

[0074] This invention adopts a "global pre-computation + local fast execution" architecture. The core idea is to construct a pair of forward potential fields (FPF) and reverse potential fields (RPF) for each destination node d in the network, and generate table-driven forwarding rules for them. During data plane forwarding, data packets switch between the forward and reverse potential fields according to the rules, thereby achieving fast and robust rerouting.

[0075] (1) Pre-calculated structure:

[0076] like Figure 1 As shown, for each destination node d in the network, the following steps are performed: Potential function calculation: Using the reverse shortest path algorithm, starting from d, calculate the shortest hop count from all nodes to d as the potential value. Neighbor classification: Based on the potential gradient difference, classify neighbors into forward neighbors, backward neighbors, and equipotential neighbors. Equipotential edge orientation: Orient all equipotential edges, prioritizing increasing the number of intersection nodes and reducing pure source / pure sink nodes. Constructing a bidirectional potential field: FPF: Composed of strictly descent edges and oriented equipotential edges; RPF: Obtained by reversing the edges of the FPF. Weight assignment and rule generation: Calculate the weight for each edge (reflecting the subsequent out-degree) and generate next-hop selection rules for the FPF and RPF.

[0077] (2) Forwarding structure:

[0078] like Figure 2As shown, when a data packet is sent from source node s to destination node d: it enters FPF based on the forwarding state; the node selects the node with the "higher out-degree" among the available forward neighbors as the next hop; the potential value decreases monotonically, ensuring the path is loop-free and close to optimal. If an intermediate node finds that all its forward neighbors are unavailable (e.g., link / node failure): a backoff condition is triggered, switching to RPF; the data packet backs off in RPF, prioritizing nodes that can reach more forward selections; the goal of the backoff is to reach the cross node, providing conditions for subsequent recovery. When the backoff path reaches the cross node: the recovery condition is checked, i.e., there is an available forward neighbor and it will not return to the fault discovery point; the data packet switches back to FPF and continues forward; it continues transmission on the new branch path until it reaches the destination node d.

[0079] Figure 3 A schematic diagram illustrating an example of the bidirectional potential field (DPF) forwarding mechanism proposed in this invention is provided.

[0080] In this example, for a given destination node d, the network node is mapped to a potential energy distribution space based on the pre-calculated potential function value. This mapping allows for a clear distinction: the edges in the forward potential field (FPF), i.e., directed edges from high-potential nodes to low-potential nodes, and the edges formed after equipotential edge orientation processing, represent the forward path of the data packet under normal conditions; the edges in the reverse potential field (RPF), obtained by reversing all edges of the FPF, represent the fallback path available when forward movement is blocked.

[0081] Figure 3 The image illustrates the complete process of data packet forwarding in two types of potential energy fields:

[0082] 1) Normal Forwarding: The data packet starts from the source node and moves along the edges in the FPF towards nodes with lower potential, with the final goal of reaching the destination node d. The forwarding rule at this time follows the principle of "selecting the forward neighbor with a higher out-degree" to ensure that the path is as robust as possible and acyclic.

[0083] 2) Forward Blocked: When all forward neighbors of a node are unavailable (e.g., v5, due to link or node failure), packets cannot continue forward in the FPF.

[0084] 3) Back-off handover: At this point, according to the inter-site handover control logic, the data packet switches from FPF to RPF and begins to back off gradually along the reverse edge. The weighting rules of RPF ensure that the data packet backs off to the node with stronger forward capability first.

[0085] 4) Cross-node recovery: When a packet falls back to a cross-node (e.g., v3) in the RPF, the recovery condition is met if that node has a usable forward neighbor that has not traversed the faulty area (e.g., v2). At this time, the packet switches back to the FPF and continues to advance towards the destination node.

[0086] This example visually illustrates three key features of the invention:

[0087] Dual-field coordination: FPF and RPF complement each other, enabling data packets to maintain connectivity in a "forward-blocked-backward-recovery" cycle;

[0088] Loop-free guarantee: In FPF, the potential energy strictly decreases; in RPF, the potential energy strictly increases; in addition, the recovery condition avoids returning to the fault point, thus eliminating the risk of loops from the mechanism.

[0089] Robustness Enhancement: By increasing the number of cross nodes through equipotential edge orientation, recovery opportunities are increased, and arrival rate and path quality can be significantly improved even in multi-point failure environments.

[0090] Advantages of this invention compared to existing technologies:

[0091] 1) Higher computational efficiency and lower cost:

[0092] Existing proactive solutions based on arc trees or integer programming suffer from exponentially increasing computational complexity, making them difficult to apply to large-scale networks. This invention constructs a potential function and potential field based on the shortest hop count, requiring only polynomial time complexity for pre-computation. Furthermore, each destination node only needs to construct a pair of forward and reverse potential fields, resulting in a simple and efficient computation process. In the data plane, forwarding relies solely on table lookups and a small number of state markers, with a lookup complexity of O(1), supporting line-speed execution and significantly reducing implementation costs and resource consumption.

[0093] 2) Enhanced robustness and wider applicability:

[0094] Traditional proactive solutions rely on strict strong connectivity conditions, requiring each node to maintain a backup path in the event of any failure. However, in heterogeneous networks, low-out-degree nodes are widespread, causing this condition to generally not hold. This invention relaxes the reachability requirement by introducing a bidirectional potential field, requiring only the ability to backtrack to a cross node with multiple branches. This more relaxed "cross node reachability condition" is easier to satisfy in practical heterogeneous networks, thus significantly improving the network's robustness and applicability.

[0095] 3) Excellent path quality and fast convergence speed:

[0096] While reactive approaches offer rapid responses, they often result in excessively long paths, path oscillations, or even loops. This invention guarantees a monotonically decreasing potential energy in the forward field and a monotonically increasing potential energy in the reverse field, and eliminates the possibility of loops by preventing a return to the fault point during recovery. This design achieves near-optimal path lengths (experiments show an average path length increase of less than 1.2 times) and can quickly find alternative paths in fault scenarios without requiring full network convergence.

[0097] 4) High network stability and good service continuity:

[0098] Existing reactive solutions lack global coordination and are prone to unstable forwarding behavior. This invention, through a structured design of the potential energy field and strict inter-field switching conditions, ensures stable operation of data packets even in complex fault scenarios, preventing oscillations or deadlocks. Therefore, it can guarantee communication continuity in mission-critical scenarios (such as military communications, emergency command, and financial transactions).

[0099] 5) Strong scalability and compatibility:

[0100] This invention does not rely on a specific network topology or protocol and can be deployed on any connected topology; status information can be embedded in the IPv6 extension header, requiring minimal changes to the protocol stack and making it easy to deploy incrementally in existing networks; it is suitable for various scenarios ranging from data centers to wide area networks to tactical emergency communications.

[0101] This invention is applicable to various heterogeneous networks that have strict requirements for rapid recovery and path stability in link / node failure scenarios, including but not limited to: backbone / metropolitan area networks, data center networks, tactical communication networks, industrial internet / vehicle-to-everything (V2X) networks, etc. The network scale can reach thousands of nodes, and the topology can be regular or non-uniformly distributed (heterogeneous); the failure mode can be random failure or clustered / correlated failure.

[0102] Bidirectional Potential Field (DPF) Architecture: This invention proposes constructing a pair of forward potential fields (FPF) and reverse potential fields (RPF) for each destination node in the network, and implementing table-driven forwarding in the data plane. Data packets advance in the FPF along the "potential decrease" direction. When the forward path is blocked, it can switch to the RPF to backtrack along the "potential increase" direction, and then resume forwarding in the FPF at a suitable crossover node. This dual-field collaborative design overcomes the strict requirement of "strong reachability" imposed by traditional unidirectional potential fields, ensuring connectivity in complex heterogeneous networks and multi-point fault environments, and guaranteeing loop-free paths and fast convergence from a mechanism perspective. Robust Forwarding Mechanism Based on Crossover Nodes: This invention increases the number of crossover nodes through equipotential edge orientation processing and uses crossover nodes as recovery anchors to form a closed-loop forwarding mechanism of "forward-backward-recovery". During forwarding, a weighted neighbor selection rule is adopted: in FPF, forward neighbors with higher out-degrees are selected first, and in RPF, backtracking to nodes with stronger forward capabilities is prioritized; at the same time, the recovery condition requires that the new path must bypass the fault discovery point. This mechanism not only improves the packet arrival rate, but also ensures that the path is close to optimal and eliminates loops and oscillations, thereby achieving high stability and business continuity in mission-critical applications.

[0103] Alternative Potential Function: In the current scheme, the potential function uses the "shortest hop count" as the potential value from the node to the destination node. In an alternative implementation, other network metrics can be used as the basis for the potential function, such as link latency, link bandwidth, and link reliability. This allows for potential fields prioritizing either "low latency" or "high reliability" in different application scenarios, still supporting bidirectional forwarding and rapid recovery.

[0104] Alternative strategies for handling equipotential edges: In the current scheme, equipotential edges are directed using a "cross-connection priority + source / sink reduction" strategy. Alternative implementations can employ random direction, link reliability-based direction, or direction based on load balancing objectives. These alternative methods can also increase the cross-connection ratio, but differ in their optimization objectives.

[0105] Alternative approaches to inter-site handover mechanisms: In the current scheme, the recovery condition is limited to "the forward neighbor is available and does not return the fault discovery point." Alternative implementations can allow more recovery conditions, such as "the path detour length is below a certain threshold" or "the link reliability of the neighbor node is above a certain threshold." These methods can also avoid loops and guarantee convergence.

[0106] Alternatives to the weighted rules: In the current scheme, weights are selected based on the "out-degree size of neighbors." Alternative implementations can use comprehensive metrics (such as out-degree + link bandwidth, out-degree + historical stability), or even machine learning models to predict the "optimal neighbor." These alternative rules still ensure that more robust paths are prioritized.

[0107] Alternative deployment mechanisms: The current solution supports both centralized (unified calculation by the controller) and distributed (independent calculation by nodes) approaches. An alternative implementation can also employ a hybrid deployment: the controller calculates the bidirectional potential field of key nodes, while other ordinary nodes utilize lightweight local calculations. This further reduces pre-computation overhead in large-scale networks while maintaining fast convergence.

[0108] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A fast rerouting method based on a two-way potential field, characterized in that, include: Construct positive and negative potential energy fields based on the target nodes in the network; The process of constructing the positive and negative potential energy fields includes: Calculate the number of hops from all nodes in the network to the destination node as the potential value; Based on the potential difference between adjacent nodes, neighbors are divided into forward neighbors, backward neighbors, or equipotential neighbors. Orient equipotential edges to increase the number of intersection nodes; The positive potential energy field is formed by the descending edge and the equipotential edge after orientation; By reversing the edges of the positive potential field, a reverse potential field is obtained; Based on the positive potential field and the negative potential field, a forwarding rule for the data packet is generated; During packet forwarding, the system switches between the positive potential field and the negative potential field according to the link status. The process of switching between a positive potential energy field and a negative potential energy field includes: If the current node has a usable forward neighbor, the data packet moves forward in the positive potential field; If all forward neighbors are unavailable and there are available reverse neighbors, the data packet switches to the reverse potential field and backs off. When a packet retreats to a cross node in the reverse potential field and that node has an available forward neighbor that does not pass through the faulty region, the packet switches back to the forward potential field to proceed.

2. The fast rerouting method based on a bidirectional potential field as described in claim 1, characterized in that, The potential value can be replaced by other network metrics, which are any one of link latency, link bandwidth, and link reliability, or any combination including hop count.

3. The fast rerouting method based on a bidirectional potential field as described in claim 1, characterized in that, The process of orienting equipotential edges includes determining the direction of the equipotential edges with the goal of reducing the number of pure source nodes and pure sink nodes and increasing the number of intersection nodes.

4. The fast rerouting method based on a bidirectional potential field as described in claim 1, characterized in that, The process of generating packet forwarding rules includes: calculating weights for the edges in the forward potential field and the reverse potential field; and generating next-hop selection rules based on the weights.

5. The fast rerouting method based on a bidirectional potential field as described in claim 4, characterized in that, The next hop selection rules include: In a forward potential field, the next hop is selected from the forward neighbor nodes with higher out-degrees; in a reverse potential field, the next hop is selected from the nodes that can lead to more forward choices.

6. The fast rerouting method based on a bidirectional potential field as described in claim 1, characterized in that, The process of switching back to the positive potential field after the data packet falls back to the cross node includes: When backtracking to a certain cross node and finding a forward neighbor that does not pass through the original fault region, the data packet switches back to the positive potential field and continues to move towards the target node.

7. The fast rerouting method based on a bidirectional potential field as described in claim 1, characterized in that, The method further includes: A status flag is added to the packet header to record the current state of the packet and whether it is in a fallback context; each node makes forwarding decisions based on the pre-generated forwarding table and the status flag.

Images