A multi-modal perception, adaptive evolution method and system based on a space-time field medium

Abstract

The application discloses a kind of multi-modal perception, adaptive evolution method and system based on space-time field medium, it is related to artificial intelligence and robot control technical field.The method includes: converting multi-modal perception signal into continuous physical wave structure in space-time field medium, realize the ontology isomorphism mapping of vision, hearing and action;Based on the potential energy topographic map of external environment, realize the natural emergence of action by conforming potential energy gradient;Convert negative feedback of environment into high tension energy, and record energy gully in space-time field medium to realize adaptive evolution.The application discards the traditional discrete digital calculation and serial processing architecture, and completes perception, calculation and execution through the natural evolution of physical medium, realizes zero-delay instinct level response and calculation, and the endogenous intelligence exists, significantly reduces the computing power consumption and improves the robustness of system in complex dynamic environment.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and robot control technology, and in particular to a low-level architecture that abandons traditional discrete digital computing and realizes multimodal perception, action emergence and autonomous evolution based on continuous spatiotemporal field medium. Background Technology

[0002] Currently, mainstream artificial intelligence and robotics systems generally adopt the "von Neumann architecture" or traditional neural network models. Their core logic is a serial processing mode of "perception-computation-execution." This mode has the following significant drawbacks: First, perception and action are separated. Data from sensors such as vision and hearing must undergo complex encoding, feature extraction, and logical operations before being converted into control commands, resulting in high system latency and an inability to achieve instantaneous reactions similar to biological instincts. Second, computational power consumption is enormous. Traditional AI relies on massive matrix multiplications for probability prediction or path planning. With the increase of multimodal information, computational power increases exponentially, and the system is prone to overload or crashes in complex environments. Finally, existing systems lack physical-level adaptive evolutionary capabilities. Their "learning" relies on manually written code and external database updates, unlike biological systems that can directly reshape cognitive models at the underlying physical structure through "pain feedback," resulting in a lack of true "body intuition" and intrinsic intelligence. Summary of the Invention

[0003] The purpose of this invention is to overcome the above-mentioned defects in the prior art and provide a multimodal sensing and adaptive evolution method and system based on spatiotemporal field medium to solve problems such as the separation of perception and action, huge computational consumption, and lack of physical-level adaptive evolution capabilities.

[0004] To achieve the above objectives, the technical solution adopted by this invention is to transform the information carrier from a "discrete digital signal" to a "continuous physical field structure," realizing the physical integration of computation, storage, and transmission. Specifically, this includes: transforming multimodal sensing signals into a continuous physical wave structure in a spatiotemporal field medium, achieving ontological isomorphic mapping of vision, hearing, and action; based on the potential energy topography of the external environment, achieving natural emergence of action by conforming to the potential energy gradient; and transforming negative environmental feedback into high-tension energy, etching energy grooves in the spatiotemporal field medium to achieve adaptive evolution.

[0005] The beneficial effects of this invention are as follows: The computational power consumption drops dramatically: the computation process is completed by the natural evolution of the physical medium, without the need for complex logic gate circuits, realizing "computation is existence".

[0006] Breaking through the physical delay limit: perception and action occur synchronously within the same field structure, achieving an instinctive response with zero delay.

[0007] Extremely strong system robustness and resilience: When subjected to extreme information overload, the system will not crash, but will automatically perform "fuzzification and dimensionality reduction", sacrificing secondary information to preserve the core task. Attached Figure Description

[0008] Figure 1 The following is an overall architecture diagram of a multimodal sensing and adaptive evolution system based on a spatiotemporal field medium provided in an embodiment of the present invention: (① is the main sensing path from the sensor group to the field programmer; ② is the fast reflection path from the sensor group directly to the spatiotemporal field module; ③ is the closed-loop link from the robot execution back to the sensor; ④ is the drive command path from the spatiotemporal field module output to the robot). Figure 2 This is a schematic diagram of multimodal signal isomorphism fieldization and ontological isomorphism mapping in an embodiment of the present invention (where 1 is the visual reference spherical field matrix, 2 is the audio modulation surface wave texture, and 3 is the high-risk thorn-like field structure). Figure 3 This is a schematic diagram of the principle of spontaneous emergence of action based on potential energy topography in an embodiment of the present invention (1 is the flexible intelligent body, 2 is the landing execution foot, and 3 is the high-risk thorn-like field structure). Figure 4 This is a schematic diagram of the field structure changes in pain recording and adaptive evolution in an embodiment of the present invention (1 is the reference field sphere, 2 is the high-energy imprinting energy shear head, 3 is the imprinting energy groove, and 4 is the resonance early warning diffusion ripple). Figure 5 The flowchart shows the multimodal sensing and adaptive evolution method provided in the embodiment of the present invention, and the steps in the figure correspond to the complete process of the method of the present invention in sequence. Detailed Implementation

[0009] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Example 1

[0010] like Figure 1 , Figure 2 and Figure 5 As shown, this embodiment discloses the multimodal sensing and adaptive evolution implementation method based on spatiotemporal field media of the present invention. The overall hardware architecture of the system is as follows: Figure 1 As shown, it consists of a multimodal sensor, an array of field programmers, a spatiotemporal field medium, and an actuator. The overall workflow logic is as follows: Figure 5 The flowchart is shown.

[0011] The system acquires multimodal sensory signals, including visual, auditory, and tactile, in real time through an array of external field programmers. When the system's visual end identifies a spherical target object with high-temperature hazard attributes and simultaneously picks up low-frequency audio signals resembling boiling in the environment, the system maps the hazard semantic features into a high-tension, high-sharpness thorny field structure and coats it onto the surface of the visual reference field structure; at the same time, it modulates the picked-up low-frequency sound wave signals into a fine vibration texture on the surface of the field structure.

[0012] This invention eliminates the need for traditional image recognition and audio decoding processes. Instead, it directly relies on the differentiated characteristics of the composite field structure within the spatiotemporal field to achieve integrated synesthetic understanding of hazard attributes, sound source information, and target morphology, thus completing the isomorphic mapping of multimodal information ontology. Example 2

[0013] like Figure 3 As shown, this embodiment demonstrates the adaptive control process of spontaneously emerging actions based on potential energy topographic maps according to the present invention.

[0014] When the agent traverses muddy, uneven slopes, the visual perception module collects real-time terrain information and constructs a potential energy topography map matching the terrain in the spatiotemporal field medium at the bottom of the agent. The agent itself exhibits flexible and deformable medium characteristics, and can spontaneously complete flexible deformation according to the potential energy gradient direction of the potential energy topography map, automatically adjust its foot contact posture, avoid protruding rocks on the ground, and autonomously lower its center of gravity.

[0015] This embodiment requires no discrete command control or step-by-step calculations throughout the entire process. It relies on the natural gradient characteristics of the physical field to generate a continuous, smooth, and terrain-fitting hydrodynamic biomimetic gait. Example 3

[0016] like Figure 4 As shown in the figure, this embodiment illustrates the mechanism of pain recording and autonomous evolution based on negative feedback in this invention.

[0017] When an intelligent agent first comes into contact with a container that is stable on the outside but boiling at high temperature inside and is stimulated by high-temperature steam, generating negative overload energy exceeding the normal threshold, the high-intensity overload energy acts as an imprinting source, etching permanent concave energy trenches on the spatiotemporal field structure surface corresponding to the target, thus completing the initial experience solidification.

[0018] When the system detects a target object with the same appearance again, the solidified energy channels inside the field structure will resonate with the target field signal, automatically triggering a long-distance early warning mechanism and initiating evasion, protection and other defensive actions in advance, achieving instinctive adaptive evolution without database retrieval or logical judgment. Example 4

[0019] This embodiment discloses the overload adaptive control mechanism of the present invention under extreme complex working conditions.

[0020] When the system simultaneously receives tens of thousands of sets of high-speed moving and complex target perception data and enters a high-load extreme working condition, the system will not experience lag, crash, or data overflow. Instead, it will actively perform dimensionality reduction and collapse processing on the field structure of distant, low-threat, and secondary targets, aggregating secondary information into low-resolution energy clusters, and prioritizing the retention of high-precision field structure information of close-range, high-risk core targets.

[0021] In extreme scenarios such as supersonic high-speed hazard attacks, the system can capture the precursor ripple signals of environmental disturbances, construct a risk prediction field in advance, and complete body posture adjustments and evasive actions in advance, thereby achieving a highly dynamic risk avoidance response based on advanced perception of the spatiotemporal field.

Claims

1. A multimodal sensing and adaptive evolution method based on spatiotemporal field media, characterized in that, Includes the following steps: The system acquires multimodal sensing signals and transforms them into continuous physical wave structures in a spatiotemporal field medium. Based on these continuous physical wave structures, it establishes an ontological isomorphic mapping of the multimodal signals and maps the external environment into a potential energy topography map within the spatiotemporal field medium. By conforming to the potential energy gradient of the potential energy topography map, it drives the actuator to achieve the natural emergence of actions. It acquires negative environmental feedback signals, transforms them into high-tension energy, and etches energy grooves in the spatiotemporal field medium to complete the adaptive evolution of the system.

2. The method according to claim 1, characterized in that, The process of converting multimodal sensing signals into continuous physical wave structures in a spatiotemporal field medium specifically includes: converting visual signals into field structures with specific spatial topology; converting auditory signals into low-frequency fluctuations that modulate the surface tension of the visual field structures; and converting hazard semantic tags into thorny field structures with high-sharpness characteristics.

3. The method according to claim 1, characterized in that, The method of driving the actuator to achieve natural emergence of action by conforming to the potential energy gradient of the potential energy topography map specifically includes: converting the geometric features of the target terrain into the potential energy distribution at the bottom of the spatiotemporal field medium in real time; controlling the intelligent agent as a whole to act like a flexible medium, automatically undergoing physical deformation according to the gradient change of the potential energy distribution, and generating a continuous hydrodynamic gait.

4. The method according to claim 1, characterized in that, The etching of energy trenches in the spatiotemporal field medium specifically includes: When a negative environmental feedback signal is detected that exceeds a preset threshold, the overload energy is used as a chisel to form a permanent energy trench on the field structure of the corresponding sensing object. When the same sensing object is detected again, a resonance warning is generated through the energy trench, and defense or avoidance actions are automatically triggered.

5. The method according to claim 1, characterized in that, It also includes an adaptive step for limiting pressure: When the total amount of input multimodal sensing signals exceeds the carrying capacity threshold of the spatiotemporal field medium, the field structure of non-core regions is automatically fuzzified and dimensionality reduced, and secondary information is collapsed into low-resolution energy clusters to maintain the high precision of the core task field structure.

6. A multimodal sensing and adaptive evolution system based on spatiotemporal field media, characterized in that, include: An array of field programmers is used to acquire multimodal sensing signals and convert them into continuous physical wave structures in a spatiotemporal field medium; A spatiotemporal field medium is used to carry the continuous physical wave structure and generate a potential energy topography map; a scheduling controller is used to drive the actuator to achieve the natural emergence of action according to the potential energy gradient of the potential energy topography map. An adaptive recording module is used to convert negative environmental feedback into high-tension energy and record energy grooves in the spatiotemporal field medium.

7. The system according to claim 6, characterized in that, The spatiotemporal field medium is a flexible field structure medium with continuous physical properties, and its computing power consumption is borne by the natural evolution process of the medium.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 5.

9. An intelligent robot, characterized in that, It is equipped with the multimodal sensing and adaptive evolution system based on spatiotemporal field medium as described in claim 6 or 7.

Images