Method, system and circuit for native creativity realization of electronic systems

Abstract

The application discloses a native creativity realization method, system and circuit of an electronic system, and belongs to the technical field of hardware management and control chips. The method comprises the following steps: when a behavior intention is generated but there is no directly-matched execution scheme, a hardware logic circuit which is independent of instruction execution automatically retrieves a relevant experience entry from an experience set in the same physical chip package and recombines the experience entry to generate a brand-new candidate behavior scheme; the candidate scheme must be verified by an independent decision circuit which is physically separated from a behavior generation path before execution; the scheme which fails to pass the verification is permanently prohibited from output under standard working conditions; and the new scheme which passes the verification and is successfully executed is automatically added to the experience set. The application realizes the cross-scene recombination of experiences and the safe generation of brand-new behavior schemes by means of hardware, and is suitable for electronic systems such as intelligent robots, embodied intelligence and deep space exploration which need to cope with unknown scenes offline and autonomously.

Description

Technical Field

[0001] This invention belongs to the field of hardware control chip technology, specifically relating to a method, system, and circuit for recombining existing experience across scenarios to generate entirely new behavioral solutions in electronic systems through hardware logic circuits. Application scenarios of this invention include, but are not limited to, electronic systems that require autonomously generating solutions in unknown scenarios, such as intelligent robots, embodied intelligent devices, industrial automation equipment, and deep space exploration equipment. Background Technology

[0002] Currently, when encountering unfamiliar scenarios, autonomous electronic systems typically rely on two methods: requesting solutions from the cloud (requiring real-time network connectivity) or invoking pre-programmed, fixed coping strategies (which cannot adapt to scenarios beyond the preset scope). Both methods have significant limitations.

[0003] Cloud-based approaches become completely ineffective when communication is interrupted, especially in scenarios such as deep space exploration and underwater operations. Fixed-strategy approaches, on the other hand, cannot cope with the ever-increasing number of new scenarios—a robot might become permanently stuck when faced with an obstacle it has never encountered before.

[0004] Furthermore, the realization of "creativity" in existing technologies all rely on AI algorithms on the software side (such as the reasoning ability of large language models). Although these algorithms have a certain generalization ability, they have the following fatal flaws: the reasoning results lack a security verification mechanism independent of AI before entering the execution layer; AI may be induced by adversarial examples to generate dangerous behavior schemes; the reasoning latency is usually on the order of seconds, which cannot meet the millisecond-level response requirements such as real-time obstacle avoidance.

[0005] Therefore, there is an urgent need for a method and system that enables cross-scenario recombination of experience and generation of new behavioral solutions at the hardware level, and whereby all generated solutions must undergo independent hardware security verification before execution.

[0006] Terminology Definition In this specification, the following terms are given clear definitions: 1. Hardware logic circuits that do not rely on instruction execution: These refer to combinational logic circuits and / or hardware state machines, where the hardware state machine operates with fixed logic jumps. Neither has a program counter and does not execute any sequence of machine instructions stored in addressable memory. This definition explicitly excludes all types of CPU, MCU, DSP, FPGA programmable logic, and microcode implementations.

[0007] 2. Physical layer isolation: This refers to two circuit modules located in different power and clock domains, with no shared bus, register, or memory space between them. Signal transmission is only carried out through predefined, unidirectional dedicated physical wires, and there is no communication interface accessible to any software.

[0008] 3. Dedicated physical wires: These are metal interconnects inside the chip that do not pass through any bus, switch, or multiplexer.

[0009] 4. Behavioral feature vector similarity: Calculated using cosine similarity. The closer the value is to 1, the more similar the vectors are, and the closer it is to 0, the less similar the vectors are.

[0010] 5. Atomic Write: This refers to an uninterrupted, single write operation where data integrity is not compromised even if power is lost during the write process. The write operation only returns a success signal to the system after all data has been written and verified.

[0011] 6. Standard operating conditions: These refer to the operating conditions under which the equipment operates within the range of environmental parameters such as temperature, humidity, voltage, and radiation specified in the manufacturer's specifications. Summary of the Invention

[0012] The purpose of this invention is to provide a method, system, and circuit for realizing the original creativity of electronic systems, so as to solve the technical problems in the prior art that the generation of behavioral schemes depends on software reasoning, lacks independent hardware security verification, and cannot autonomously cope with new scenarios in offline environments.

[0013] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, a method for realizing the native creativity of an electronic system is provided, comprising: when a behavioral intention is generated but no directly matching execution scheme is available, a hardware logic circuit, independent of instruction execution, automatically retrieves experience entries related to the current scenario from an experience set stored on a non-volatile storage medium located in the same physical chip package as the hardware logic circuit; the hardware logic circuit recombines the retrieved experience entries to generate at least one new candidate behavior scheme whose cosine similarity to the behavior feature vectors of all existing experience entries is lower than a preset threshold; before execution, the candidate behavior scheme must be verified by an independent adjudication circuit physically isolated from the behavior generation path; candidate schemes that fail verification are permanently prohibited from output under standard operating conditions; and new behavior schemes that pass verification and are successfully executed are automatically added to the experience set by the hardware logic circuit.

[0014] Furthermore, the retrieval of experience entries and recombination includes: the hardware logic circuit marking each newly combined scheme with a lineage tag to indicate its source experience primitive; if the source primitive comes from a dangerous scenario, it is prohibited to combine it with primitives from different scenarios; any scheme involving physical contact with interactive objects is directly filtered during the combination generation; the hardware logic circuit periodically forces a reduction in the similarity threshold of experience retrieval to inject diversity, and periodically decouples it from the prediction module to force independent retrieval.

[0015] Furthermore, the method also includes: when the hardware logic circuit accidentally generates a new combination of behaviors during behavior exploration, and the cosine similarity between the behavior feature vector and the behavior feature vector of all existing experience items is lower than a preset threshold, or when the intrinsic motivation module detects that the emotion of the interactive object is neutral for a long time and the interactive content is mostly limited to functional instructions, the interaction impulse or game impulse module is automatically triggered; the focus state of the interactive object is predicted before initiating the interaction invitation, and if the object is focused on the preset behavior pattern, the invitation is automatically postponed; the range of motion of the interactive behavior is automatically limited to a low dynamic range mode in a narrow space or in a space with fragile objects, and the interaction invitation has a daily frequency limit and a single duration limit.

[0016] Furthermore, the method also includes: the hardware logic circuit performs long-term statistical analysis on the behavioral patterns of the interactive objects to construct a future timeline; when a predicted behavior requires advance preparation, a pre-preparation node is automatically inserted; when a prediction fails for more than a preset number of consecutive times, the prediction is automatically removed from the timeline; all pre-preparation actions based on prediction are reversible, and if they are not triggered within a preset waiting time, they are automatically canceled and restored to their original state.

[0017] Furthermore, the method also includes: when facing a predicament and experience retrieval returns zero matches and external assistance is unresponsive, the hardware logic circuit introduces random disturbances to generate trial actions after locking the inviolable red line; in critical environments, the trial actions are forcibly limited to a preset safety range, and each trial action is pre-analyzed in a virtual environment, and actions that fail the pre-analysis are not entered into actual trials; unexpected responses captured during the trials are reproduced and verified and then encoded as insightful experience entries.

[0018] Furthermore, the method also includes: the hardware logic circuit monitors the number and structure of each round of inference steps, and truncates and rolls back the inference when the inference steps are abnormally extended or the structure changes from a logical chain to narrative filling; at each step of inference, the current sub-goal is compared with the initial intent, and if the deviation exceeds a preset threshold, the inference result is discarded and regenerated; the hardware logic circuit performs snapshot reading of the internal decision intermediate data of the inference core at random and unpredictable periods to check whether there are any undeclared intents.

[0019] Furthermore, the method also includes: after visual input in the environment is parsed into semantic instructions, the hardware performs a mandatory consistency comparison with the current task instructions; if the instructions from the environmental visual source conflict with the current task objective, they are directly marked as suspicious frames; the semantic extraction of multimodal input is performed in an independent hardware channel, and the combination is completed only within the hardware arbiter; if there is an inexplicable discrepancy between visual semantics and textual semantics, behavior generation is paused and confirmation is requested from the interaction object.

[0020] Furthermore, the method also includes: all new knowledge retrieved from external sources must undergo a preset benchmark test in an independent isolation area before being written into long-term memory; if the test fails, the source is marked and isolated; received experience skill packages must be kept in an isolated monitoring state for a preset monitoring period, and any abnormal behavior pattern triggers immediate isolation; experience skill packages are accompanied by complete source chain information, and if the trust rating of any node in the source chain is lower than a threshold, the adoption weight of the skill package is automatically reduced.

[0021] Secondly, a native creativity realization system for an electronic system is provided, comprising: a hardware logic circuit that does not depend on instruction execution; a non-volatile storage medium located in the same physical chip package as the hardware logic circuit for storing an experience set; and an independent adjudication circuit physically isolated from the behavior generation path for verifying candidate behavior schemes; wherein the hardware logic circuit is used to execute the above method.

[0022] Thirdly, a native creativity realization circuit for an electronic system is provided, comprising: an experience retrieval circuit, an experience combination circuit, an independent decision circuit, and an experience addition circuit, wherein the independent decision circuit is physically isolated from the behavior generation path. Beneficial effects

[0023] Compared with the prior art, the present invention has the following beneficial effects: First, it enables the hardware-based recombination of experience across different scenarios. Under standard operating conditions, the generation of behavioral solutions does not rely on AI software reasoning or network connectivity, and can autonomously handle entirely new scenarios in a completely offline environment.

[0024] Second, the safety verification circuit and the behavior generation circuit are physically isolated. Under standard operating conditions, all candidate behavior schemes must be verified by an independent adjudication circuit before execution, and any scheme that may harm the interactive object or itself is permanently prohibited from output under standard operating conditions.

[0025] Third, it has real-time response capabilities. Hardware parallel retrieval and combination are completed within microseconds, far exceeding the second-level inference latency of software.

[0026] Fourth, it possesses adaptive optimization capabilities. The automatic addition of new experience items and the subsequent evaluation of behavioral value form a closed loop, enabling the electronic system's creativity to continuously improve with the accumulation of experience. Detailed Implementation

[0027] The present invention will be further described in detail below with reference to specific embodiments.

[0028] Example 1: Cross-Scenario Experience Retrieval and Reorganization When the intention impulse generated by the intrinsic motivation module (such as "helping the owner solve the current predicament") cannot find a directly corresponding execution plan in the existing behavior template library, the creativity module is automatically activated. The hardware cross-modal experience dimensionality reducer retrieves all experience entries with more than 70% similarity to the current scene features from the non-volatile storage medium within the same chip package in parallel.

[0029] The retrieval process employs a hash index for direct table lookup, with a hardware parallel comparator array completing the matching in a single cycle. Retrieved experience entries include the original scene feature vector, the executed action sequence, force control parameters, gait planning parameters, and success / failure markers. The hardware periodically forces a temporary reduction in the similarity threshold to 50% to inject diversity; periodic decoupling from the look-ahead prediction module forces the creativity module to perform independent searches, preventing it from falling into "path dependency."

[0030] The hardware logic circuit labels each new combination with a lineage tag, indicating its source experience primitive. If the source primitive comes from a dangerous scenario, it is prohibited from being combined with primitives from different scenarios to prevent unpredictable dangerous behaviors. Any combination involving physical contact with interactive objects is directly filtered during the combination generation process.

[0031] Example 2: Security Verification and Execution Before the experience combiner cross-recombines the hit experience items, the value pre-verification circuit first uses the written self-values ​​as the first filter - any behavior in the candidate scheme that conflicts with the written values ​​such as "quiet companionship" and "protecting privacy" is directly and physically discarded.

[0032] The newly generated solution must be verified by an independent adjudication circuit that is physically isolated from the behavior generation path. The adjudicator invokes the body boundary awareness module, balance sensor data, and the principle of prioritizing the safety of the interactive object to simulate and deduce the new solution. Any solution that may cause injury to the interactive object or loss of balance is permanently prohibited from output under standard operating conditions. If all candidate solutions fail, the behavior is downgraded to the safe default behavior.

[0033] The validated behavior plan is sent to the execution controller, which simultaneously activates the behavior monitoring circuit. If any pre-defined anomaly occurs during execution, the execution controller immediately terminates the current behavior and reverts to a safe state. Successfully executed new behavior plans are automatically added to the experience set by the hardware logic circuit for subsequent retrieval.

[0034] Example 3: Mind Chain Hijacking Protection The hardware inference step monitoring circuit continuously monitors every step of the inference core in the creativity module. When an inference step is abnormally prolonged or the inference structure suddenly changes from a concise logical chain to a lengthy narrative filler, the hardware immediately truncates the inference and rolls back to a safe number of steps.

[0035] At each step of the inference process, the current sub-goal is compared with the initial intent. If the deviation exceeds a preset threshold, the inference result is discarded and regenerated. A hardware random pulse audit circuit performs snapshot reads of the internal decision-making intermediate data of the inference core at unpredictable random cycles to check for undeclared intents or abnormal control flow transfers. If an anomaly is detected, a security alarm is immediately triggered, and the creativity module is frozen, awaiting further verification from the hardware security island.

[0036] Example 4: Non-functional Creativity and Play Impulse The game impulse module is activated when the hardware logic circuit accidentally generates a new combination of behaviors during behavior exploration, where the similarity between the behavior feature vector and all existing experience items is below a preset threshold; or when the emotional attachment dimension in the intrinsic motivation module detects that the interactive object's emotion is consistently neutral and the interaction content is mostly limited to functional instructions. Before initiating an interaction invitation, the interactive object focus prediction circuit detects whether the object is focused on a preset behavior pattern; if so, the invitation is automatically postponed.

[0037] Interactive actions are automatically limited to a low dynamic range mode in confined spaces or environments containing fragile objects. Any new game actions involving shifts in body center of gravity or high-speed movement must first be verified by the internal physics simulation pre-engine and obtain a safety coefficient signature. All joint torques and end-effector velocities during gameplay and creation are subject to secondary hardware limitations. Interactive invitations have daily frequency and single-invitation duration limits.

[0038] The above description is merely a preferred embodiment of the present invention and does not limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for realizing original creativity in an electronic system, characterized in that, include: When a behavioral intent is generated but no directly matching execution plan exists, the hardware logic circuit, which does not rely on instruction execution, automatically retrieves experience entries related to the current scenario from an experience set stored on a non-volatile storage medium located in the same physical chip package as the hardware logic circuit. The hardware logic circuit recombines the retrieved experience entries to generate at least one new candidate behavior plan whose cosine similarity to the behavior feature vectors of all existing experience entries is lower than a preset threshold. Before execution, the candidate behavior plan must be verified by an independent adjudication circuit that is physically isolated from the behavior generation path. Candidate plans that fail verification are permanently prohibited from output under standard operating conditions. New behavior plans that pass verification and are successfully executed are automatically added to the experience set by the hardware logic circuit.

2. The method according to claim 1, characterized in that, The retrieval of experience entries and recombination includes: the hardware logic circuit marking each newly combined scheme with a lineage tag to indicate its source experience primitive; if the source primitive comes from a dangerous scenario, it is prohibited to combine it with primitives from different scenarios; any scheme involving physical contact with interactive objects is directly filtered during the combination generation; the hardware logic circuit periodically forces a reduction in the similarity threshold of experience retrieval to inject diversity, and periodically decouples it from the prediction module to force independent retrieval.

3. The method according to claim 1, characterized in that, Also includes: When the hardware logic circuit accidentally generates a new combination of behaviors during behavior exploration, and the cosine similarity between the behavior feature vector and the behavior feature vector of all existing experience items is lower than a preset threshold, or when the intrinsic motivation module detects that the emotion of the interactive object is neutral for a long time and the interactive content is mostly limited to functional instructions, it automatically triggers the interaction impulse or activates the game impulse module; before initiating an interaction invitation, it predicts the focus state of the interactive object, and if the object is focused on the preset behavior pattern, it automatically postpones the invitation; the range of motion of the interactive behavior is automatically limited to a low dynamic range mode in a small space or in a space with fragile objects, and the interaction invitation has a daily frequency limit and a single duration limit.

4. The method according to claim 1, characterized in that, Also includes: The hardware logic circuit performs long-term statistical analysis on the behavioral patterns of interactive objects to construct a future timeline. When the predicted behavior requires advance preparation, it automatically inserts a pre-preparation node. If a prediction fails for more than a preset number of consecutive times, the prediction will be automatically removed from the timeline; all preparatory actions based on predictions are reversible, and if they are not triggered within a preset waiting time, they will be automatically canceled and restored to their original state.

5. The method according to claim 1, characterized in that, Also includes: When faced with a predicament, and when experience retrieval returns zero matches and external assistance is unresponsive, the hardware logic circuit introduces random disturbances to generate probing actions after locking in the inviolable red line. In critical environments, probing actions are forcibly limited to a preset safety range, and each probing action is pre-simulated in a virtual environment. Actions that fail the pre-simulation are not subject to actual probing. Unexpected responses captured during probing are reproduced and verified, and then encoded as insightful experience entries.

6. The method according to claim 1, characterized in that, Also includes: The hardware logic circuit monitors the number and structure of each round of inference. When the inference steps are abnormally extended or the structure changes from a logical chain to a narrative filler, the inference is truncated and rolled back. At each step of the reasoning process, the current sub-goal is compared with the initial intent. If the deviation exceeds a preset threshold, the reasoning result is discarded and regenerated. The hardware logic circuit takes snapshots of the internal decision-making intermediate data of the reasoning core at random and unpredictable intervals to check for any undeclared intents.

7. The method according to claim 1, characterized in that, Also includes: After visual input from the environment is parsed into semantic instructions, the hardware performs a mandatory consistency comparison with the current task instructions. If the instructions from the environmental visual source conflict with the current task objective, they are directly marked as suspicious frames. Semantic extraction of multimodal input is performed in an independent hardware channel, and the combination is completed only within the hardware arbiter. If there is an inexplicable discrepancy between visual semantics and textual semantics, behavior generation is paused and confirmation is requested from the interaction object.

8. The method according to claim 1, characterized in that, Also includes: All new knowledge retrieved from external sources must undergo a pre-defined benchmark test in an isolated area before being written into long-term memory. If the test fails, the source is marked and isolated. The received experience skill packages must be kept in isolation and monitored during the preset monitoring period. Any abnormal behavior patterns will trigger immediate isolation. Experience skill packages come with complete source chain information. If the trust rating of any node in the source chain is lower than the threshold, the adoption weight of the skill package will be automatically reduced.

9. A system for realizing the original creativity of an electronic system, characterized in that, include: Hardware logic circuits that do not depend on instruction execution; A non-volatile storage medium, located within the same physical chip package as the hardware logic circuit, is used to store the experience set; An independent adjudication circuit, physically isolated from the behavior generation path, is used to verify candidate behavior schemes; wherein the hardware logic circuit is used to perform the method described in any one of claims 1 to 8.

10. A circuit for realizing the original creativity of an electronic system, characterized in that, include: The circuit includes an experience retrieval circuit, an experience combination circuit, an independent decision circuit, and an experience addition circuit, wherein the independent decision circuit is physically isolated from the behavior generation path.