Distributed computing and communication system with brain-like consciousness and implementation method thereof

Abstract

The application provides a distributed computing and communication system with brain-like consciousness and an implementation method thereof, comprising: a plurality of computing nodes and a consciousness state management module; each computing node comprises a computing unit, a storage unit and a communication unit; the communication units constitute a network on chip (NoC) as a whole; the computing unit and the storage unit are connected with nodes of the NoC, and the cooperation and data sharing between the computing nodes are completed through an efficient communication protocol; and the consciousness state management module is configured on each computing node to constitute a consciousness management system of a CTM, that is, a brain-like model of the CTM, so that the mutual cooperation between each node dynamically adjusts the computing strategy to complete a complex computing task. Each computing node of the application is configured with a consciousness state management module, can perceive the real-time state of the node, and accepts the input from other nodes; through analyzing the data, the module can adjust the computing strategy of the node in real time, and realize the formation and dynamic adjustment of the consciousness.

Description

Technical Field

[0001] This invention relates to the field of distributed computing and communication technology, and more specifically, to a distributed computing and communication system with brain-like consciousness and its implementation method. Background Technology

[0002] Distributed computing and communication systems play a crucial role in modern information technology. By enabling parallel computing across clusters of numerous computing nodes, they can significantly improve system computing power and are widely used in fields such as big data processing, cloud computing, artificial intelligence training, and distributed database management. Traditional distributed systems typically rely on fixed task allocation and resource management strategies. While these strategies can improve computational efficiency to some extent, they still have the following shortcomings when facing dynamic environments and complex tasks:

[0003] 1. Low resource utilization: Due to the lack of flexibility in task and resource allocation, the system has difficulty in achieving optimal resource allocation under different task loads, resulting in low resource utilization.

[0004] 2. Insufficient computational efficiency: When dealing with large-scale data and complex computing tasks, traditional distributed systems often fail to fully utilize the capabilities of each computing node due to the lack of efficient coordination mechanisms, thus affecting overall computational efficiency.

[0005] 3. Poor system robustness: The existing system lacks an effective self-adjustment and recovery mechanism when dealing with uncertainties such as node failures and network fluctuations, resulting in insufficient system stability and robustness.

[0006] 4. Lack of autonomous learning and adaptability: Traditional distributed systems lack perception and learning capabilities, and cannot dynamically adjust computing strategies based on real-time data and environmental changes, thus limiting the level of system intelligence.

[0007] To overcome the aforementioned problems, proposing a distributed computing and communication system with consciousness and self-regulation capabilities has become a new research direction. The Conscious Turing Machine (CTM), as a novel computing model, introduces a consciousness state management module, enabling computing nodes to perceive their own state and their role in the system, thereby achieving autonomous learning and adaptation. The application of the CTM brain-like model is expected to significantly improve the resource utilization, computational efficiency, and robustness of distributed systems. Summary of the Invention

[0008] In view of the deficiencies in the prior art, the purpose of this invention is to provide a distributed computing and communication system with brain-like consciousness and its implementation method.

[0009] According to the present invention, a distributed computing and communication system with brain-like consciousness is provided, comprising: multiple computing nodes and a consciousness state management module;

[0010] Each computing node includes a computing unit, a storage unit, and a communication unit; the communication unit as a whole constitutes an on-chip routing network (NoC); the computing unit and storage unit are connected to the nodes of the on-chip routing network (NoC) and complete the collaboration and data sharing between computing nodes through a communication protocol;

[0011] Furthermore, the consciousness state management module configured on each computing node constitutes the consciousness management system of CTM, namely the CTM brain-like model, thereby enabling each node to cooperate with each other to dynamically adjust computing strategies to complete computing tasks.

[0012] Preferably, the computing node includes a long-term memory module (LTM) or a short-term memory module (STM) as defined in the CTM;

[0013] The on-chip routing network is a hierarchical, multi-layered routing structure, including a hybrid hierarchical routing architecture based on Torus and Tree.

[0014] Preferably, the nodes of the on-chip routing network NoC refer to the Local node of the router, the basic building block of NoC. The other four nodes of the router, namely the four nodes of east, west, south and north, are connected to other routers respectively. Adjacent routers realize bidirectional multi-channel communication, and the whole constitutes a distributed parallel on-chip routing communication network.

[0015] Preferably, the communication protocol includes an asynchronous communication protocol for message queues, a point-to-point communication protocol, a multicast communication protocol, and dynamic compression coding.

[0016] Preferably, the computing unit is used to perform distributed computing tasks, including inference computing for independently running neural networks on an autonomous robot;

[0017] The storage unit is used to store computational data and status information. Each computing node has its own corresponding storage unit, which can quickly read data.

[0018] The communication unit is used to exchange data with each router node of the NoC.

[0019] Preferably, the consciousness state management module is used to implement the dynamic adjustment function of CTM, and to sense and adjust the state of computing nodes and task processing strategies.

[0020] Preferably, the consciousness state management module includes: a perception unit, a decision-making unit, and a feedback unit;

[0021] The sensing unit is used to collect and analyze the real-time status data of the computing nodes;

[0022] The decision-making unit is used to adjust the task processing strategy of the computing node based on the data provided by the perception unit;

[0023] The feedback unit is used to record and evaluate the effects of the adjustments and to further optimize the decision-making process.

[0024] Preferably, the consciousness state management module can autonomously learn and adjust its calculation strategy based on real-time data and environmental changes, thereby optimizing system performance;

[0025] The on-chip routing network can not only realize uplink and downlink unicast or multicast communication required for communication between consciousness state management modules, but also realize fast connection of dark channels between modules.

[0026] Preferably, the distributed computing and communication system employs distributed state control, which determines the consistency and data integrity of each computing node when processing distributed tasks based on distributed state consistency and resource consistency.

[0027] According to the present invention, a method for implementing a distributed computing and communication system with brain-like consciousness is provided. The method, employing the aforementioned distributed computing and communication system with brain-like consciousness, includes the following steps:

[0028] Step S1: Based on the number of computing nodes and the task scale, build an on-chip routing network of appropriate scale;

[0029] Step S2: Connect the computing unit and storage unit to the NoC routing node to enable parallel communication between computing units and achieve collaboration and data sharing between computing nodes through an efficient communication protocol;

[0030] Step S3: Configure a consciousness state management module on each computing node, which together constitutes the consciousness management system of CTM;

[0031] Step S4: Using the consciousness state management module, combined with historical data, current tasks, and expected goals, dynamically adjust the weight information of the calculation unit to complete the calculation task.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. Each computing node in this invention is equipped with a consciousness state management module, which can sense the real-time state of the node and accept input from other nodes. By analyzing this data, the module can adjust the computing strategy of the node in real time to realize the formation and dynamic adjustment of "consciousness".

[0034] 2. This invention employs efficient communication protocols, including a multi-level fused topology on-chip routing network, an asynchronous communication protocol based on message queues, a point-to-point unicast communication protocol, a point-to-multipoint broadcast communication protocol, and a dynamic compression encoding protocol, to achieve rapid data exchange and collaborative work between CTM computing nodes, reducing communication latency and overhead.

[0035] 3. This invention employs resource lock control to control concurrent access to shared resources by distributed computing nodes; and achieves high-performance distributed scheduling by analyzing the dependency graph between computing tasks. Attached Figure Description

[0036] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0037] Figure 1 This is a schematic diagram of the overall structure of the distributed computing and communication system of the present invention;

[0038] Figure 2 It is a schematic diagram of the structure of a single computing node; it shows the computing unit, storage unit and communication unit.

[0039] Figure 3 This is a schematic diagram of the internal structure of the consciousness state management module;

[0040] Figure 4 This is a schematic diagram of the communication process between computing nodes in the system of this invention; it demonstrates an efficient communication protocol.

[0041] Figure 5 It is a distributed resource lock control mechanism;

[0042] Figure 6 It is a task dependency graph analysis graph;

[0043] Figure 7 This is a schematic diagram of the overall operation of the consciousness management system. Detailed Implementation

[0044] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0045] The distributed system of this invention consists of multiple parallel computing nodes. Each node includes a computing unit, a storage unit, a communication unit, and a consciousness state management module. The nodes collaborate with each other to complete complex computational tasks in parallel. Through an efficient communication protocol, the distributed system achieves seamless collaboration and data sharing among the computing nodes. Each node possesses autonomous learning and adaptive capabilities, adjusting its computing strategy based on real-time data and environmental changes. Specifically, this invention is based on a CTM brain-like model and introduces a consciousness state management method, enabling the entire system to achieve brain-like distributed management and computing characteristics. This allows each computing node in the distributed system to possess the ability to perceive, self-adjust, and optimize. Consciousness state management combines historical data, current tasks, and expected goals to dynamically adjust the computing model, thereby improving the system's perception capabilities and robustness. Furthermore, the system ensures the efficiency of parallel distributed computing through an efficient on-chip network communication system and ensures consistency and data integrity among nodes when processing distributed tasks through distributed state control and task dependency graph management.

[0046] According to the present invention, a distributed computing and communication system with brain-like consciousness is provided, such as... Figure 1 As shown, it includes: multiple computing nodes and a consciousness state management module. Each computing node includes a computing unit, a storage unit, and a communication unit; the communication unit as a whole constitutes a Network On Chip (NoC); the computing unit and storage unit are connected to the nodes of the NoC, and the collaboration and data sharing between computing nodes are completed through an efficient communication protocol; and each computing node is equipped with a consciousness state management module to form the consciousness management system of CTM, thereby enabling each node to cooperate with each other to dynamically adjust computing strategies to complete complex computing tasks.

[0047] The computing nodes are either Long-Term Memory (LTM) or Short-Term Memory (STM) modules defined in the Conscious Turing Machine (CTM). The on-chip routing network employs a 2D Mesh topology, achieving lower communication latency and QoS management mechanisms based on virtual channels. The nodes in the on-chip routing network (NoC) refer to the Local nodes of the router, the basic building block of the NoC. The other four nodes of the router (east, west, south, and north) are connected to other routers, and adjacent routers achieve bidirectional multi-channel communication, forming a distributed parallel on-chip routing communication network. The efficient communication protocols include asynchronous communication protocols using message queues, point-to-point communication protocols, multicast communication protocols, and dynamic compression coding technology.

[0048] The computing unit is used to execute distributed computing tasks, which may be independent neural network inference computation on an autonomous robot. The storage unit is used to store computing data and state information. Each computing node has its corresponding storage unit, enabling fast data retrieval and presenting an overall "near-memory computing" operation mode. The communication unit is used to exchange data with other nodes, specifically each router node of the NoC. The consciousness state management module is used to implement the dynamic adjustment function of CTM, sensing and adjusting the state and task processing strategies of computing nodes. The consciousness state management module includes: a sensing unit, a decision-making unit, and a feedback unit. The sensing unit is used to collect and analyze real-time state data of computing nodes; the decision-making unit is used to adjust the task processing strategies of computing nodes based on the data provided by the sensing unit; the feedback unit is used to record and evaluate the effects of the adjustments and further optimize the decision-making process. The consciousness state management module can autonomously learn and adjust computing strategies based on real-time data and environmental changes, thereby optimizing system performance. The on-chip routing network can not only realize the uplink and downlink unicast or multicast communication required for communication between consciousness state management modules, but also realize fast dark channel connections between modules.

[0049] The distributed computing and communication system with brain-like consciousness described in this invention employs distributed state control to ensure consistency and data integrity of each computing node when processing distributed tasks in terms of distributed state consistency and resource consistency.

[0050] The implementation method of the distributed computing and communication system with brain-like consciousness provided by the present invention includes the following steps:

[0051] Step S1: Build an on-chip routing network (NoC) of appropriate size according to the number of computing nodes (task size).

[0052] Step S2: Connect the computing unit and storage unit to the NoC routing node to realize parallel communication between computing units. Through an efficient communication protocol, achieve collaboration and data sharing between computing nodes.

[0053] Step S3: Configure a consciousness state management module on each computing node to form the CTM consciousness management system.

[0054] Step S4: Using the consciousness state management module, combined with historical data, current tasks, and expected goals, dynamically adjust the weight information of the calculation module to complete the calculation task.

[0055] The core of this invention lies in utilizing a state management module to perceive the real-time state of computing nodes and their position and role within the system. By combining historical data, current tasks, and expected goals, the state management module can dynamically adjust the computing modules. Through the adoption of efficient communication protocols and distributed resource lock control, coordination and data consistency among computing nodes are ensured.

[0056] Furthermore, the distributed computing and communication system with brain-like consciousness and its implementation method of the present invention are described in detail below with reference to the accompanying drawings:

[0057] like Figure 1 As shown, this invention presents the overall architecture of a distributed computing and communication system with brain-like consciousness. The system includes multiple computing nodes, which are divided into STM and LTM types. Each node has a computing unit, a storage unit, a communication unit, and a consciousness management module. The communication units collectively constitute an on-chip routing network (NoC), such as... Figure 1 As shown on the right, the NoC (No. 1) provides communication connections between computing nodes, enabling data exchange and task collaboration. Specifically, the on-chip routing network implements a hierarchical parallel distributed architecture. To improve the scalability of the hierarchical interconnection architecture, this invention designs a hybrid hierarchical on-chip network routing architecture. This architecture adopts a hierarchical multi-layer routing structure, primarily based on a hybrid hierarchical routing architecture of Torus and Tree. The hierarchical architecture can be customized to multiple layers. The top layer uses a bidirectional Torus structure to achieve scalable flexibility in routing, with each Torus route corresponding to a lower-level partition. The lower layers use a Tree structure to interconnect more computing cores to achieve more efficient and convenient communication speeds and obtain higher bandwidth. In the specific design, the communication nodes in the hierarchical architecture are Tiles. Each Tile contains an upper-level Torus route, a Tree route connected to it, and four computing cores connected to the Tree route. The upper-level Torus route communicates with different Torus routes through four ports (East, West, South, North) (EWSN), and then communicates with the lower-level Tree routes through Local endpoints. The four ports of the Tree communicate with the computing cores.

[0058] The single computing node architecture in this invention, such as Figure 2 and Figure 3 The diagram illustrates the computation, storage, communication units, and consciousness state management module within a single computing node. The computation unit primarily performs neural network inference operations, i.e., vector-matrix multiplication, which translates to specific execution tasks in autonomous robots, such as facial recognition and map navigation. The storage unit stores the neural network's weight information and the communication information (chunks) that the computation unit needs to transmit to other units; these are encoded as follows:

[0059] <address,t,gist,weight,intensity>

[0060] In this context, `address` is the address of the LTM processor that generates the information block, `t` is the time of block generation, `gist` is the brain-text information encoded in the LTM module, which is also the core communication information for aligning multiple LTM modules, `weight` is the weight used by the processor to measure the importance of the current brain-text, and `intensity` is the accumulated amount of weight used to record the competing communication path, with initial times `t0` being |ht| and `weight`, respectively. In the CTM-based consciousness management system, external information with different characteristics is processed by different LTM modules and shared among them. The communication unit is the basic building block of NoC, the Router. It adopts a wormhole routing mechanism, dividing the complete data packet into multiple Flits and transmitting them separately. When a computing node generates and sends a data packet, it transmits it to the connected Router node and encodes the corresponding target location information in the data. Within each Router, the Flit data needs to undergo a routing decision based on the current location and the target location to determine its output port on the current Router. The NoC of this invention uses multiple virtual channels. The Flit of multiple channels may compete for the same output port downstream, requiring Round_Robin arbitration. Data that is successfully arbitrated can be allocated to the output port and downstream virtual channels. The entire communication is executed cyclically according to the above process until the complete Packet arrives at the destination node.

[0061] The components of the consciousness state management module, such as... Figure 3 As shown, they exist in different configurations, and different LTM computing nodes have different capabilities. Specifically, as... Figure 3 (a) shows "Perceptual Encoding LTM"—acquiring information from the outside world and encoding it into "brain language," such as... Figure 3 (b) shows "Information Storage and Memory LTM"—which transforms input data based on stored information and outputs the transformation result; such as Figure 3 (c) shows the "Behavior Output LTM"—which outputs action commands based on input data. It has great flexibility in specific implementation forms. For example, a multimodal image-text association network (CLIP) can be used to implement image encoding in a visual perception LTM; a large language model can be used in a "Memory Storage" LTM to calculate and output "brain language" from the input "brain language", etc.

[0062] The efficient communication protocols employed in this invention include asynchronous communication protocols, unicast / multicast communication, and dynamic compression coding technology. Specifically, the asynchronous communication protocol is described as follows: Due to the influence of layered many-core architectures, communication between distributed computing cores is highly dependent on clock speeds, necessitating clock synchronization. To address this, this invention employs a Global Asynchronous Local Synchronization (GALS) data communication scheme. Three main methods solve the problem of secure and reliable data transmission between independent clock domains: Pausible-clock generators, asynchronous FIFOs, and boundary synchronization. This invention divides the distributed on-chip routing network architecture into multiple local synchronization domains. Asynchronous FIFOs are used for local synchronization between different domains, eliminating clock coupling between many cores, reducing the difficulty of system interconnection and expansion, and overcoming the strict clock synchronization requirements between chips. Figure 4 As shown in (a). For FIFOs with unlimited depth, this invention uses a nested counting method to determine whether the FIFO memory is empty or full. Considering the asynchronous synchronization issue, it requires that when the FIFO encoding changes from one value to an adjacent value, exactly one bit changes. To meet this condition, this invention uses Gray code to re-encode the read and write addresses to ensure that only adjacent bits differ when the value changes. The conventional binary code to Gray code conversion operation is as follows:

[0063]

[0064] By converting binary to Gray code and then introducing the next higher bit for discrimination, a ring counting mechanism is implemented to determine whether the FIFO is full or empty. If the higher bits are the same, it indicates that no ring counting has been performed, and the write address and read address are directly compared. If the difference is equal to the FIFO depth, the FIFO is full; if they are equal, the FIFO is empty. By using a globally asynchronous and locally synchronous communication scheme, the interconnection and expansion difficulty of the distributed system can be reduced, the strict clock synchronization requirements between distributed computing units can be eliminated, and system function failures caused by errors in some clock domains can be avoided.

[0065] In terms of unicast and multicast communication, data packets in the NoC are primarily communicated serially. For each routing node, only one data slice is allowed to pass through within one clock cycle. This approach offers advantages such as better control of the data flow, minimizing data packet corruption and errors during transmission, and conserving resources. This data transmission method is called data unicast. Data unicast, as the most basic function of on-chip networking, can meet most traffic demands. However, in CTM-based distributed communication systems, there is broadcast communication from STM to all LTMs. Therefore, this invention considers supporting multicast communication, transmitting data packets from the same source node to different destination nodes in parallel. More specifically, for each routing node, data slices are allowed to be copied and forwarded to multiple output ports within the same clock cycle. This data propagation method reduces data transmission latency and improves transmission efficiency, thereby enhancing the overall communication performance of the NoC. Traditional multicast is divided into two categories: path-based data multicast and topology-based data multicast. Currently, some algorithms combine these two traditional methods, utilizing the concept of partitioning for data multicast, such as... Figure 4 As shown. This invention, considering the flexibility of the top-level Torus architecture, aims to maximize the advantages of the shortest communication path; the multicast communication supported in the architecture adopts a topology-based approach. For example... Figure 4 As shown in (b), path-based data multicast searches for Hamiltonian cycles in the topology, aiming to connect all routes with a single path. This method first requires determining a suitable Hamiltonian cycle and sorting the route nodes according to the path. Data packets requiring multicast are copied and forwarded based on the sequence numbers of the destination and source nodes, traversing the entire path until all destination nodes have received the data packets. The advantage of this method is that it avoids deadlock, and data packets are forwarded in ascending or descending order of node sequence numbers according to the Hamiltonian cycle. This invention modifies the data packet format, arbitrator arbitration method, and port output specifications to design data copying and forwarding at Torus routing nodes that require branching, and simultaneously forwards the data to the corresponding distributed computing units in the lower-level Tree routes. While supporting multicast communication, NoC also supports the original unicast communication by setting corresponding flag masks to change the multicast address, thereby achieving simultaneous support for both unicast and multicast.

[0066] Regarding dynamic compression coding protocols, selecting appropriate data compression coding algorithms can alleviate the communication bandwidth pressure on NoCs to some extent. In data-intensive and communication-intensive applications, cache misses or prolonged blocking of multiple data packets can put enormous pressure on the underlying network, preventing data packets from being transmitted from multiple cache hierarchies to the request core and reducing the overall forwarding efficiency of the NoC. To make the most of the limited resources of the NoC, compression pulses are needed to save resource overhead. Information transmitted between LTMs has similar coding formats, with small differences in data content and limited dynamic fluctuation range of data values, allowing for fine-grained differential data compression. Figure 4 (c) This demonstrates a data packet with 5 flasks, totaling 20 bytes. Through differential compression, without altering the header and base flasks, the packet can be compressed to 12 bytes, significantly reducing bandwidth pressure for large-scale data. Fine-grained differential compression uses the FlitZip algorithm, which is hardware-friendly. This algorithm divides data flasks into several sub-blocks, determines the base block within each sub-block as the differential basis, and uses several subtractors to perform differential operations to obtain encoded sub-blocks. These are then integrated to form a compressed data packet with a higher compression ratio. Decoding is achieved by using several adders to perform the reverse operation. FlitZip effectively reduces data flask data, increases the compression ratio, and lowers bandwidth requirements and forwarding latency.

[0067] In distributed computing and communication architectures, resource conflicts and state control failures often occur when multiple scheduling units schedule tasks. Distributed control is responsible for ensuring the scheduling state and resource coordination consistency of the distributed system, guaranteeing resource coordination and correct task submission during the distributed module computation process. Furthermore, this invention uses coarse-grained time steps as the operation cycle for decision-making and learning tasks, and controls the overall system's running state and LTM task threads by setting timers.

[0068] The following sections will elaborate on distributed state consistency and distributed resource consistency:

[0069] Regarding distributed state consistency, decision-making and learning tasks are completed collaboratively and distributedly by multiple modules; this process is often referred to as a distributed transaction. The distributed transaction mechanism employed in this invention provides a way to incorporate a sequence of tasks involving module collaboration into a single execution control unit. A transaction can only be committed if all operations are executed correctly; failure of any operation will cause the entire transaction to roll back to its previous state, meaning all operations are canceled.

[0070] Regarding distributed resource consistency, the distributed lock used in this invention operates in a distributed environment where decision-making and learning tasks are deployed across multiple LTM threads, achieving multi-threaded distributed mutual exclusion. Figure 5 As shown. To achieve distributed mutual exclusion, consider setting up a distributed lock monitor in the distributed system. This resource lock status is visible to every thread. When the resource lock no longer exists, a flag can be set. Once the resource lock is set, other threads can only wait for the thread holding the lock to complete execution and release the lock before they can set the resource lock and access the shared resource. A resource lock is a flag used in scenarios where multiple threads access the same resource simultaneously to ensure that threads access shared resources without interference, thereby guaranteeing the validity and correctness of operations. Simultaneously, to ensure that multiple threads can see the resource lock, it is stored in a shared memory. This allows multiple threads to concurrently access the same critical resource, but only one thread can access the shared resource at a time, ensuring data consistency during the operation of the distributed system.

[0071] This invention proposes to use distributed multi-module computing technology to adapt to the execution mode of the CTM brain-like model, and automatically generate a communication graph model between multiple modules according to the decision learning logic, supporting distributed parallel computing and unicast / multicast communication in the brain-like behavior decision architecture.

[0072] In traditional distributed computing, a computational task is generally called a Job, and a Job consists of multiple Tasks. Each Task can run independently (cross-machine) or collaboratively (single-machine) on a single server. The number of Tasks into which a Job is divided determines the parallelism (Granularity or Parallelism) of the distributed computing. In the distributed system of this invention, a single decision or learning task can be considered a Job, and this Job is jointly participated in by multiple LTM modules, i.e., LTM modules are considered Tasks. Therefore, in the distributed system, multiple Jobs (i.e., multiple decision and learning tasks) run simultaneously, and each Job is completed collaboratively by multiple Tasks in a distributed manner. The core of distributed computing is to maintain the logical relationship topology graph between Jobs and Tasks, and to enable these tasks to run in parallel, while also minimizing the communication and data access overhead caused by the distributed network. Utilizing a single-broadcast communication mechanism between multiple LTM modules, this invention implements a distributed computing system based on multiple LTM modules (i.e., Tasks) through NoC. The execution of different modules may be heterogeneous. In order to ensure performance, distributed computing technology needs to schedule tasks to all nodes as much as possible and avoid the slow performance problem caused by the "barrel effect".

[0073] To this end, a distributed computing architecture is adopted to automatically associate corresponding functional modules based on the decision learning relationships between multiple modules, generate Task labels, and generate a Directed Acyclic Graph (DAG) according to the dependencies between Tasks, such as... Figure 6As shown, each node T0-T6 is a basic computational task (e.g., LTM module brain-text information generation). Job 1 (e.g., STM generation and action LTM decision-making or learning) includes tasks T0, T1, T4, and T5, while Job 2 includes tasks T1, T2, T3, and T6. Each directed edge represents a dependency. In this example, tasks T0-T3 can be executed concurrently, with a maximum parallelism of 4, meaning that 4 nodes can execute tasks simultaneously. Task T4 requires the completion of its two preceding subtasks T0 and T1 before it can begin computation, while T5 requires the completion of tasks T4 and T2 before it can begin computation.

[0074] In this computing model, the scheduling unit is responsible for generating a series of tasks from a job and storing them in a task pool according to their task dependencies. When a relevant task is involved, an enable signal is assigned, the corresponding module performs the calculation, and returns the result. Multiple jobs are distributed and scheduled by different scheduling units. When the same task or task conflict is detected, a resource locking mechanism is automatically activated to wait, thereby completing the relevant computing jobs in a distributed manner, that is, the computing tasks of the computing units in the distributed system.

[0075] In addition, based on the dynamic consciousness management mechanism of CTM, such as Figure 1 The left side shows the overall architecture of the CTM brain-like model, where STM is the short-term memory module; LTM is the long-term memory module; Up Tree is the up-tree connecting LTM to STM; Down Tree is the down-tree connecting STM to LTM; Links are dark channels established between LTMs for direct communication through associative learning; Input acquires environmental information collected by sensors and sends it to the designated LTM processor; Output outputs the behavior, i.e., the decision result, and transmits it to the actuator to act on the environment; Chunk is the information block for communication between STM and each LTM, and the core communication information in it is collectively referred to as "brain language". Each LTM corresponds to a different functional area of ​​the brain and performs different functions. Multiple LTMs transmit multiple "brain language" information to STM through the up-tree; when the information correlation between two LTMs is high enough, a dark channel is established to bypass the complex competition process and communicate directly; the semantic consistency of "brain language" between different LTMs is achieved through a "co-learning" mechanism. The STM (Spirit Memory Module) uses a "competitive decision-making" mechanism to determine which LTM (Level Memory Module)'s "brain language" to publish from among the received information. The published "brain language" is defined as "brain-like consciousness." This brain language is then broadcast to various LTMs via a descending tree, thus forming a feedback loop and creating continuous content updates on the STM. The model operates as follows: Figure 7As shown in (a), all the "brainwashing" published on the STM at each moment is considered as a "state vector." The information on the STM is dynamically updated over time, forming a "chain" of "state" transitions, such as... Figure 7 As shown in (b), the change of "state" can be viewed as a Markov chain, where all LTMs take the current state as input and their output determines the next state. Based on the semantic encoding representation of "brain-like consciousness," an interpretable "brain-like consciousness stream" can be output. Furthermore, the content range of the "brain language" published on the STM can be limited by additional constraints, thereby achieving the controllability of "brain-like consciousness."

[0076] This invention aims to improve the computational efficiency, resource utilization, and robustness of distributed systems, making it suitable for applications such as autonomous robots and artificial intelligence devices. The neuromorphic distributed computing and communication system of this invention is suitable for processing large-scale complex tasks, such as those in autonomous robots and artificial intelligence devices. This invention achieves high-performance computing through parallel distributed computing and communication, and realizes efficient collaboration and self-regulation capabilities between nodes through CTM (Computer-Driven Module), significantly improving the overall system's computational efficiency, reducing computational latency, and enhancing the system's flexibility and adaptability.

[0077] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0078] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A distributed computing and communication system with brain-like consciousness, characterized in that, include: Multiple computing nodes and a consciousness state management module; Each computing node includes a computing unit, a storage unit, and a communication unit; the communication unit as a whole constitutes an on-chip routing network (NoC); the computing unit and storage unit are connected to the nodes of the on-chip routing network (NoC) and complete the collaboration and data sharing between computing nodes through a communication protocol; Furthermore, the consciousness state management module configured on each computing node constitutes the consciousness management system of the Consciousness Turing Machine (CTM), thereby enabling each node to cooperate with each other to dynamically adjust computing strategies to complete computing tasks.

2. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The computing node includes a long-term memory module (LTM) or a short-term memory module (STM) as defined in the CTM. The on-chip routing network is a hierarchical, multi-layered routing structure, including a hybrid hierarchical routing architecture based on Torus and Tree.

3. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The nodes of the NoC on-chip routing network refer to the Local node of the Router, the basic building block of the NoC. The other four nodes of the Router are connected to other routers respectively. Adjacent routers realize bidirectional multi-channel communication and together form a distributed parallel on-chip routing communication network.

4. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The communication protocols include asynchronous communication protocols for message queues, point-to-point communication protocols, multicast communication protocols, and dynamic compression encoding.

5. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The computing unit is used to perform distributed computing tasks, including inference computing for independently running neural networks on an autonomous robot; The storage unit is used to store computational data and status information. Each computing node has its own corresponding storage unit, which can quickly read data. The communication unit is used to exchange data with each router node of the NoC.

6. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The consciousness state management module is used to realize the dynamic adjustment function of CTM, and to sense and adjust the state of computing nodes and task processing strategies.

7. The distributed computing and communication system with brain-like consciousness according to claim 6, characterized in that, The consciousness state management module includes: a perception unit, a decision-making unit, and a feedback unit; The sensing unit is used to collect and analyze the real-time status data of the computing nodes; The decision-making unit is used to adjust the task processing strategy of the computing node based on the data provided by the perception unit; The feedback unit is used to record and evaluate the effects of the adjustments and to further optimize the decision-making process.

8. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The consciousness state management module can autonomously learn and adjust its calculation strategy based on real-time data and environmental changes, thereby optimizing system performance. The on-chip routing network can not only realize uplink and downlink unicast or multicast communication required for communication between consciousness state management modules, but also realize fast connection of dark channels between modules.

9. The distributed computing and communication system with brain-like consciousness according to claim 1, characterized in that, The distributed computing and communication system adopts distributed state control, which determines the consistency and data integrity of each computing node when processing distributed tasks based on distributed state consistency and resource consistency.

10. A method for implementing a distributed computing and communication system with brain-like consciousness, characterized in that, The distributed computing and communication system with brain-like consciousness as described in any one of claims 1 to 9, the implementation method of which includes: Step S1: Based on the number of computing nodes and the task scale, build an on-chip routing network of appropriate scale; Step S2: Connect the computing unit and storage unit to the NoC routing node to enable parallel communication between computing units and achieve collaboration and data sharing between computing nodes through an efficient communication protocol; Step S3: Configure a consciousness state management module on each computing node, which together constitutes the consciousness management system of the Consciousness Turing Machine (CTM). Step S4: Using the consciousness state management module, combined with historical data, current tasks, and expected goals, dynamically adjust the weight information of the calculation unit to complete the calculation task.

Images