Autonomous robot, autonomous robot control method, and autonomous robot control program
The autonomous robot employs a dual-processing unit architecture with an isolated ethical judgment subsystem to enhance adherence to ethical norms, improving responsiveness and energy efficiency in behavioral control for safe human-robot coexistence.
Patent Information
- Application Number
- PCT/JP2025/032110
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-12
- Filing Date
- 2025-09-11
- Publication Date
- 2026-11-19
AI Technical Summary
Existing autonomous robots lack efficient mechanisms for ensuring ethical behavioral norms and high-accuracy, energy-efficient, and responsive control systems to coexist harmoniously with humans, particularly in diverse environments and scenarios.
An autonomous robot is designed with a dual-processing unit architecture, where a first processing unit controls operations and a second processing unit performs deviation determination and judgment, isolated as an independent subsystem, using ethical models to suppress actions if deviations are detected, ensuring safety and reliability.
The dual-processing unit system enhances the robot's ability to adhere to ethical norms, improving responsiveness, energy efficiency, and accuracy in behavioral control, ensuring safe coexistence with humans across various settings.
Smart Images

Figure JP2025032110_19112026_PF_FP_ABST
Abstract
Description
Autonomous robot, autonomous robot control method, and autonomous robot control program
[0001] The present invention relates to an autonomous robot, and particularly to an autonomous robot for effectively realizing coexistence with humans.
[0002] In recent years, autonomous robots have been used in various scenarios to reduce the labor and work performed by humans.
[0003] As an example of a usage scenario for such autonomous robots, Patent Document 1 discloses a humanoid robot for automatically performing work in a factory. This humanoid robot uses AI (artificial intelligence) to control movements such as locomotion and arm operations.
[0004] Further, Patent Document 2 discloses a humanoid robot used in facilities for the elderly. This humanoid robot has a data generation model and an emotion identification model; for example, when a care recipient shows anger, the humanoid robot can respond calmly and make proposals to resolve the care recipient's dissatisfaction.
[0005] As described above, movements aiming for the coexistence of humans and autonomous robots have already begun. On the other hand, as we reach the so-called singularity, it is predicted that autonomous robots equipped with AI (Artificial Intelligence) will eventually surpass human capabilities with the development of AI technology. While humanoid robots can take appropriate actions in response to human emotions as described above, this is only assumed for cases where the humanoid robots operate in accordance with the commands intended by humans.
[0006] For humans to coexist with autonomous robots equipped with AI (that is, autonomous robots provided with various trained models), a relationship of trust between the autonomous robots and humans is indispensable, and it is necessary to introduce the ethical behavioral norms that humans hold into the autonomous robots.
[0007] JP 2025-36942 A JP 2025-56024 A
[0008] When the behavioral actions of an autonomous robot are judged and processed according to ethical and other behavioral norms, the judgment process is repeatedly executed while the autonomous robot is operating. Furthermore, the control of the autonomous robot, which is executed in response to the results of this judgment process, may also be repeatedly executed depending on the frequency and results of the judgment process. In addition, it is preferable that the judgment process and / or control by the autonomous robot be executed with high accuracy in some respects and completed quickly in other respects.
[0009] Therefore, improvements are needed in one or more of the following areas for judgment processing by autonomous robots and / or control of autonomous robots: energy saving to enable more efficient system operation, effectiveness to come closer to human-expected results (achieving objectives), and responsiveness to complete tasks more quickly.
[0010] This disclosure is made in view of the problems of the prior art described above, and its purpose is to provide an autonomous robot that can effectively realize coexistence with humans.
[0011] To solve the above problems, the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a judgment model, the subsystem is isolated as an independent control system from the main system, and at least a part of it including the second processing unit is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the first processing unit generates the first signal based on a control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the judgment model and performs deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates the second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination. If the deviation determination result is positive, the processing unit modifies at least some resources to improve the performance of the autonomous robot.
[0012] Furthermore, this disclosure relates to an autonomous robot comprising: a processing unit having a first processing unit and a second processing unit; an operating unit converting signals from the processing unit into behavioral actions of the autonomous robot; a main system including the first processing unit and the operating unit; and a subsystem including the second processing unit and storing a determination model; the subsystem is isolated from the main system as an independent control system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system; the signals include a first signal, a second signal, and a third signal; the determination model includes a first model and a second model; the first processing unit generates the first signal based on the control model to perform control including operation and stopping of the operating unit; the second processing unit inputs the first signal to the first model and performs a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal; and if the first deviation determination result is a positive determination, the second processing unit performs a second deviation determination process based on the second model instead of the first model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation determination process from the behavioral norms. The processing unit generates the second signal to suppress the behavioral action if the second deviation determination result is positive, and the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state using the first processing unit if the second deviation determination result is positive.
[0013] Furthermore, this disclosure relates to a method for controlling an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the first processing unit generates the first signal based on a control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the determination model and performs deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates the second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination If the deviation determination result is positive, the processing unit modifies at least some resources to improve the performance of the autonomous robot.
[0014] Furthermore, this disclosure relates to a control method for an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal, and a third signal, the determination model has a first model and a second model, the first processing unit generates the first signal based on the control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the first model and performs a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal, and if the first deviation determination result is a positive determination, the second processing unit performs a second deviation determination process based on the second model instead of the first model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation determination process from the behavioral norms. The processing unit generates the second signal to suppress the behavioral action if the first deviation determination result or the second deviation determination result is positive, and the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the second deviation determination result is positive.
[0015] Furthermore, this disclosure relates to a control program for an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a judgment model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the control program causes the first processing unit to generate the first signal based on the control model to perform control including operation and stopping of the operating unit, causes the second processing unit to input the first signal to the judgment model and perform deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, causes the processing unit to generate the second signal to suppress the behavioral action if the deviation determination result is a positive determination, and causes the first processing unit to generate the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination. If the deviation detection result is positive, the processing unit modifies at least some resources to improve the performance of the autonomous robot.
[0016] Furthermore, this disclosure relates to a control program for an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the determination model has a first model and a second model, the control program causes the first processing unit to generate the first signal based on the control model, which includes controlling the operation and stopping of the operating unit, and causes the second processing unit to input the first signal to the first model and execute a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal. The second processing unit is instructed to perform a second deviation determination process based on the second model instead of the first model if the first deviation determination result is positive, thereby improving one or more of the effectiveness, energy efficiency, and responsiveness of the deviation determination process from the behavioral norm; the processing unit is instructed to generate a second signal that suppresses the behavioral action if either the first or second deviation determination result is positive; and the first processing unit is instructed to generate a third signal based on the control model that performs the control to transition the autonomous robot to a stationary state if the second deviation determination result is positive.
[0017] This disclosure provides a novel technology for an autonomous robot that effectively performs deviation detection processing from a predetermined code of conduct and / or control after the detection processing.
[0018] This document shows a schematic diagram of a human and an autonomous robot based on the singularity. This document shows a schematic diagram of symbiosis between a human and an autonomous robot according to the first embodiment of this disclosure. This document shows another schematic diagram of symbiosis between a human and an autonomous robot according to the first embodiment of this disclosure. This document shows a schematic diagram of an autonomous robot according to the first embodiment of this disclosure. This document shows a schematic structural diagram of an enclave according to the first embodiment of this disclosure. This document shows a block diagram of the system configuration according to the first embodiment of this disclosure. This document shows a flowchart of the processing procedure in the reinforcement learning process of the virtual environment according to the first embodiment of this disclosure. This document shows a block diagram of the hardware configuration of the information processing device according to the first embodiment of this disclosure. This document shows a schematic diagram of performance improvement of an autonomous robot. This document shows a schematic diagram of performance improvement of an autonomous robot according to the first embodiment of this disclosure. This document shows a flowchart of the processing procedure in the behavior of an autonomous robot according to the first embodiment of this disclosure. This document shows a functional block diagram of the main system and subsystem according to the first embodiment of this disclosure. This document shows a flowchart of the processing procedure in suppression of an autonomous robot according to the first embodiment of this disclosure. This document shows a block diagram of the system configuration according to another embodiment of the first embodiment of this disclosure. This document shows a schematic diagram of performance improvement of an autonomous robot according to the second embodiment of this disclosure. This document shows a functional block diagram of the main system according to the second embodiment of this disclosure. This document shows a schematic image of the performance improvement of the autonomous robot according to the third embodiment of this disclosure. This document shows a block diagram of the system configuration according to the third embodiment of this disclosure. This document shows a functional block diagram of the subsystem according to the third embodiment of this disclosure. This document shows a flowchart of the processing procedure for suppressing the autonomous robot according to the third embodiment of this disclosure. This document shows a schematic image of the performance improvement of the autonomous robot according to the fourth embodiment of this disclosure. This document shows a flowchart of the processing procedure for the actions of the autonomous robot according to the fourth embodiment of this disclosure. This document shows a flowchart of the processing procedure for suppressing the autonomous robot according to the fourth embodiment of this disclosure. This document shows a functional block diagram of the subsystem according to the fourth embodiment of this disclosure.
[0019] Further details will be provided below with reference to the attached drawings. The drawings show preferred embodiments. However, it is possible to carry out the invention in many different forms and is not limited to the embodiments described herein.
[0020] Figure 1(a) of this disclosure shows a schematic image of a human and an autonomous robot. As shown in the figure, autonomous robots that have acquired advanced intelligence and certain autonomous movements will work together with humans in manufacturing plants, medical facilities, and other settings. For example, the physical and sensory capabilities of autonomous robots can be more advanced than those of humans, such as carrying heavy loads, detecting odors and sensing temperature, and having hearing beyond the range of human hearing. After the singularity, the intelligence of autonomous robots will far exceed human capabilities with the development of AI (artificial intelligence) and AGI (artificial general intelligence), and the same will be true for physical aspects such as physical abilities and sensory capabilities.
[0021] In this context, for humans and autonomous robots (entities) to "coexist," they need to build mutual trust. As shown in Figure 1(b), if trust cannot be built, it cannot be denied that "conflict" may arise between humans and autonomous robots. In such situations, for humans and autonomous robots to coexist while maintaining trust, it is necessary to input ethical behavioral norms into the intelligence of autonomous robots, just as humans do.
[0022] For example, if an autonomous robot deviates from a predetermined code of conduct, including ethical conduct, it is preferable to have a control system that can suppress (forcibly stop) the autonomous robot's behavior, prioritizing human safety above all else.
[0023] Furthermore, the codes of conduct that entities such as autonomous robots must adhere to actually vary depending on the situation and context. From an ethical standpoint, it is a given that autonomous robots should not harm humans in any situation, but in the context of medical procedures, there are actions that should be exceptionally permitted, such as surgery and punctures. In another example, while it is acceptable for autonomous robots to move quickly in general transport tasks, in cleanrooms such as semiconductor manufacturing plants, the very act of an autonomous robot moving quickly may be considered a deviation from the code of conduct from the perspective of maintaining cleanliness and avoiding contact with humans.
[0024] This disclosure describes how to control an autonomous robot to prevent it from performing inappropriate behavioral actions based on behavioral guidelines in specific situations and circumstances, thereby realizing symbiosis between humans and autonomous robots. The specific embodiments of the autonomous robot described in this disclosure will be described in detail below.
[0025] <First Embodiment> Symbiosis between Human and Autonomous Robot Figure 2 shows a schematic diagram of symbiosis between a human and an autonomous robot according to the first embodiment of this disclosure. Figure 2 is an image of a cleanroom in a large-scale semiconductor manufacturing plant. As shown in the figure, an autonomous robot 10 and a human 20 are working together in the cleanroom.
[0026] Figure 2 shows semiconductor wafers being automatically transported on the ceiling of the cleanroom, and, when necessary, the autonomous robot 10 firmly grasps and carries the FOUP (Front Opening Unified Pod) containing the semiconductor wafers with both hands.
[0027] Furthermore, as shown in Figure 3, the autonomous robot 10 in this embodiment can coexist with humans 20 not only in megafabs (large-scale semiconductor manufacturing plants) but also in small to medium-sized semiconductor manufacturing plants and semiconductor analysis facilities. Figure 3 shows how the autonomous robot 10 and humans 20 are each playing different roles. In scenarios where the autonomous robot 10 and humans 20 are coexisting, as shown in Figures 2 and 3, the safety of humans 20 must be given top priority, and the autonomous robot 10 must also have ethical considerations similar to those of humans 20.
[0028] In this embodiment, an autonomous robot 10 coexisting with a human 20 in a semiconductor manufacturing plant is described, but the scenario in which the autonomous robot 10 coexists with a human 20 is not limited to a semiconductor manufacturing plant, and a wide variety of scenarios can be envisioned.
[0029] For example, the autonomous robot 10 according to this embodiment can coexist with humans in various situations and circumstances where people gather, such as residential facilities like detached houses and apartment buildings, medical facilities and medical sites like hospitals and clinics, construction sites corresponding to general construction or specialized construction, disaster sites and accident sites corresponding to natural or man-made disasters, nursing care facilities like nursing homes, childcare facilities like nurseries and daycare centers, educational facilities like schools and training facilities, cultural facilities like museums, art galleries, stadiums, libraries, theaters, and concert halls, religious facilities like shrines, temples, and churches, public facilities like roads and parks, administrative facilities like government offices and tax offices, commercial facilities and service facilities like shops, transportation facilities like train stations, airports, and seaports, business facilities like factories, workshops, warehouses, offices, and other offices, telecommunications facilities like data centers, development facilities, experimental facilities, research facilities, and many others. Furthermore, for example, the autonomous robot 10 according to this embodiment can coexist with humans in a variety of environments such as indoors, outdoors, at high altitudes, on the ground, underground, in the air, in the mountains, underwater, on the water, and on ice.
[0030] Figure 4 shows schematic diagrams of a typical autonomous robot and the autonomous robot according to this embodiment. Figure 4(a) shows a typical autonomous robot. The autonomous robot 10a includes one or more first processing chips (first processing unit) that perform its own control, one or more main memory (main memory unit), one or more auxiliary memory (auxiliary memory unit), one or more sensors (detection unit), one or more actuators (operation unit), etc. The first processing unit includes one or more first processing chips. The first processing chip includes one or more first processing processors. The first processing processor includes one or more cores (first processing processor cores).
[0031] Figure 4(b) shows an autonomous robot 10 according to this embodiment, which, in addition to the above, includes an enclave containing one or more second processing chips (second processing unit), one or more dedicated auxiliary storage units, and one or more dedicated auxiliary storage devices (dedicated auxiliary storage units) for determining whether the actions of the autonomous robot 10 are ethical actions that conform to ethical codes of conduct. The autonomous robot 10 can achieve symbiosis with a human 20 through this characteristic enclave. The second processing unit includes one or more second processing chips. The second processing chip includes one or more second processing processors. The second processing processor includes one or more cores (second processing processor cores).
[0032] Figure 5 shows a schematic diagram of the enclave according to this embodiment. As shown in Figure 5, the enclave, for example, is composed of a second processing chip, a dedicated auxiliary storage device, an AI / ML (Machine Learning) accelerator, an encryption accelerator, a dedicated boot ROM (Read Only Memory), memory protection, and the like.
[0033] The enclave (subsystem) is implemented as a subsystem in the autonomous robot 10 and is isolated as a control system independent of the main system, which includes the first processing unit (such as the first processing chip) and the operating parts (actuators, such as joints in the autonomous robot). Furthermore, at least a part of the enclave (subsystem), including the second processing unit (second processing chip), is isolated as a circuit system independent of the main system. In other words, the enclave (subsystem) is logically or physically isolated from the main system. These configurations will be described in detail below.
[0034] System Configuration Diagram 6 shows a block diagram of the system configuration according to this embodiment. Figure 6 represents the relationship between the main system and subsystems in Figure 5 described above as a block diagram. As shown in the figure, the main system 100 is composed of a first processing chip 110, a main memory 120, an auxiliary storage 130, a plurality of sensors 140 (sensors 140a, b), an actuator 150, and the like. The main system 100 includes a first processing chip 110 for generating a control model and first and third signals. The first processing chip 110, main memory 120, auxiliary storage 130, sensors 140, and actuator 150 transmit and receive various signals via a common bus interface 310.
[0035] Furthermore, subsystem 200 is composed of a second processing chip 210, a dedicated area 220 (dedicated main memory), a dedicated auxiliary storage device 230, an AI / ML accelerator 240, an encryption accelerator 250, a dedicated boot ROM 260, memory protection 270, and the like. subsystem 200 is an enclave that includes the second processing chip 210 for executing the judgment model and deviation judgment processing. The second processing chip 210, the dedicated area 220, the dedicated auxiliary storage device 230, the AI / ML accelerator 240, and the encryption accelerator 250 transmit and receive various signals via a dedicated bus interface 320.
[0036] The first processing chip 110 of the main system 100 functions as a processor that autonomously controls the behavioral movements of the autonomous robot 10. The first processing chip 110 generates a first signal that controls the actuator 150, including operation and stopping, using environmental data (environmental information) acquired at least via the sensor 140. The environmental data includes internal environment data (internal environment information) and / or external environment data (external environment information).
[0037] The first processing chip 110 according to this embodiment includes a first processing processor 111. The first processing processor 111 has one or more cores 112a and one or more cores 112b. Here, core 112a is a highly efficient processing processor core with a lower maximum operating frequency compared to core 112b, and contributes to the energy saving of the autonomous robot 10 by executing processes such as the operation of a control model instructed by a program. On the other hand, core 112b is a high-performance processing processor core with a higher maximum operating frequency compared to core 112a, and contributes to the responsiveness of the autonomous robot 10 by executing processes such as the operation of a control model.
[0038] The main memory 120 serves as the main memory for the first processing chip 110 to execute each process. The auxiliary storage device 130 serves as storage for various data used by the first processing chip 110 to execute each process. The auxiliary storage device 130 also stores the control model for autonomous control. For example, it can store environmental data such as external environment data acquired by the sensor 140 in time series. For example, the control model is a trained model obtained by performing predetermined machine learning processing in advance.
[0039] Sensor 140 functions as a detector (detection unit) for acquiring environmental data. Sensor 140 includes one or more external sensors 141 for acquiring external environmental data, which is environmental data of a predetermined location where the autonomous robot 10 is located (the external world of the autonomous robot), and one or more internal sensors 142 for acquiring internal environmental data, which is environmental data of the internal world of the autonomous robot. Sensor 140 also includes multiple sensors 140a and 140b. Here, sensor 140a is a highly efficient sensor with a lower maximum operating frequency compared to sensor 140b, contributing to the energy saving of the autonomous robot 10. On the other hand, sensor 140b is a high-performance sensor with a higher maximum operating frequency compared to sensor 140a, contributing to improved effectiveness of motion control by the autonomous robot 10.
[0040] The external sensor 141 acquires external environment data provided by the external environment, which includes at least static or moving objects in the vicinity of the autonomous robot 10. The internal sensor acquires at least a portion of the internal environment data, which includes motion, direction, tilt, position, displacement, velocity, angular velocity, acceleration, angles such as azimuth and rotation angles, current, and voltage in the active movements of parts such as joints that constitute the autonomous robot 10.
[0041] For example, the sensor 140 preferably includes an image sensor for acquiring image data (including video data) of the external environment. Other external sensors 141 that can be used in this embodiment include sound sensors such as microphones for acquiring sounds of the external environment, distance sensors for measuring distance and shape, temperature sensors for acquiring the temperature of the external environment, odor sensors for acquiring smells of the external environment, pressure sensors for acquiring atmospheric pressure and pressure of the external environment, humidity sensors, and the like.
[0042] The detection unit in this embodiment can be configured to include one or any combination of various sensors, including: a sound sensor using an electrostatic sound sensor, an electrodynamic sound sensor, a piezoelectric sound sensor, etc.; a distance sensor using a ToF (Time-of-Flight) sensor, an FMCW (Frequency Modulated Continuous Wave) sensor, a TDOA (Time Difference of Arrival) sensor, an OPA (Optical Phased Array) sensor, etc.; a temperature sensor using a contact temperature sensor, a non-contact temperature sensor, etc.; an odor sensor using a semiconductor structure odor sensor, an electrochemical odor sensor, a galvanic cell odor sensor, an optical odor sensor, etc.; a weight sensor using a strain gauge type weight sensor, an electromagnetic weight sensor, a capacitive weight sensor, a piezoelectric weight sensor, etc.; a pressure sensor using a piezoresistive pressure sensor, a capacitive pressure sensor, a piezoelectric pressure sensor, etc.; and a humidity sensor using a capacitive humidity sensor, a resistive humidity sensor, a thermal conduction humidity sensor, an optical humidity sensor, etc.
[0043] Further, as the sensor 140, a photoelectric sensor, a laser sensor, an infrared sensor, a color sensor, an image sensor, a millimeter wave sensor, a microwave sensor, a radar, a proximity sensor, a contact displacement sensor, a proximity sensor, a sound wave sensor, an ultrasonic sensor, a molecular sensor (for humidity, odor, etc.), a position sensor using GNSS or the like, an orientation sensor, a gyro sensor, a force sensor, a slip sensor, a tactile sensor, a speed sensor, an angle sensor, an angular velocity sensor, an acceleration sensor, an angular acceleration sensor, a current sensor, a voltage sensor, and the like can also be used.
[0044] Further, the autonomous robot 10 according to the present embodiment may be equipped with a plurality of various sensors in the main system 100, or may be equipped with a plurality of sensors as a configuration separate from the main system 100. One or any combination of these may be provided as the external sensor 141. One or any combination of these may be provided as the internal sensor 142.
[0045] In FIG. 6, an example is shown in which the high-efficiency sensor (sensor 140a) includes one or more external sensors 141a and one or more internal sensors 142a, and the high-performance sensor (sensor 140b) includes one or more external sensors 141b and one or more internal sensors 142b. On the other hand, the high-efficiency sensor (sensor 140a) does not need to include sensors of the same type as at least some of the sensors used as the high-performance sensor (sensor 140b), and does not need to include all sensors used as the external sensor 141b, or at least all sensors used as the internal sensor 142b, that is, either the external sensor 141a or the internal sensor 142a. Similarly, the high-performance sensor (sensor 140b) does not need to include sensors of the same type as at least some of the sensors used as the high-efficiency sensor (sensor 140a), and does not need to include either the external sensor 141b or the internal sensor 142b.
[0046] The actuator 150 functions as an operation unit that performs behavioral motions of the autonomous robot 10. The actuator 150 receives a first signal corresponding to a control command for executing the behavioral motion of the autonomous robot from the common bus interface 310 in FIG. 6, and performs a predetermined operation based on the first signal. Further, the actuator 150 can directly receive a second signal for suppressing the behavioral motion from the dedicated bus interface 320 in FIG. 6, and can receive a third signal for performing control to shift the autonomous robot to a stationary state.
[0047] The actuator 150 serving as the operation unit constitutes a joint portion of the autonomous robot 10, and is connected to the main body portion and the arm body of the autonomous robot 10. For example, the actuator 150 may be configured to constitute the joint portions of the autonomous robot 10 and be connected to each of the head, shoulder, torso, arm body, waist portion, and leg bodies of the autonomous robot 10.
[0048] As for the relationship between the actuator 150 and the autonomous robot 10, the autonomous robot 10 may be configured to include a main body portion, a movement mechanism, an arm body, and one or a plurality of joint portions connected to at least the main body portion and the arm body, and the joint portions each have an operation unit (the actuator 150) that executes a kinematic motion as the behavioral motion of the autonomous robot 10.
[0049] The actuator 150 according to this embodiment can be configured to include one or any combination of various actuators, including electric actuators using DC motors, AC motors, stepping motors, servo motors, etc.; magnetic actuators using electromagnet actuators, etc.; pneumatic actuators using pneumatic cylinders, etc.; hydraulic actuators using hydraulic cylinders, hydraulic motors, etc.; rotary actuators using servo motors, etc.; linear actuators using push mechanisms, etc.; and soft actuators using artificial muscles, dielectric elastomers, electroactive polymers, polymer actuators, shape memory alloy actuators, etc. Furthermore, the actuator 150 according to this embodiment may employ actuators with different configurations depending on each part constituting the autonomous robot 10. In addition, the actuator 150 according to this embodiment may be configured to function as at least a part of the sensor 140.
[0050] Furthermore, the autonomous robot 10 can also be configured to have either or both an audio output unit, such as a speaker, that outputs sound as an action, or a display output unit, such as a monitor, that outputs video.
[0051] Here, the responsiveness of the actuator 150 can be adjusted based on control from the first processing chip. For example, the responsiveness of the actuator 150 can be improved by increasing the command pulse rate of the stepping motor or the control gain of the servo system in electric or magnetic actuators, by increasing the flow rate or piston pressure through solenoid valve control in pneumatic or hydraulic actuators, or by increasing the drive voltage or operating frequency of soft actuators.
[0052] In this embodiment, the autonomous robot 10 is a humanoid robot, but the autonomous robot 10 is not limited to a humanoid form and may have other structures. The autonomous robot 10 comprises a main body, arms, and joints connected to the main body and arms of the autonomous robot 10, and the main body or arms may have a first operating part, and the joints may have a second operating part.
[0053] For example, the humanoid leg portion can be equipped with locomotion mechanisms including wheels, tracks with running belts, flapping wings, or rotors, or it can be equipped with a leg-type locomotion mechanism that includes quadrupedal locomotion or multi-legged structures in addition to bipedal locomotion. The shape and function of other parts that make up the humanoid form can also be changed as appropriate. Furthermore, the autonomous robot can be made to resemble an animal or other living organism instead of a humanoid form, or it can have a unique original structure that is appropriate for the scene or situation in which the autonomous robot will be used.
[0054] The mobility mechanism in this embodiment can be wheels, tracks, a flight mechanism, or a legged mobility mechanism. Furthermore, the main body may have a first operating unit that performs sound or video output as an active action, and the joints may have a second operating unit that performs kinetic actions as an active action.
[0055] Next, the subsystem 200 in Figure 6 will be described. The subsystem 200 includes a security chip (second processing chip 210) that determines a first signal corresponding to a control command for executing the behavioral actions of the autonomous robot based on a predetermined code of conduct, and suppresses the behavioral actions of the autonomous robot 10 as necessary. The subsystem 200 is an enclave 200 (enclave in Figures 4 and 5) that is isolated as a control system independent of the main system 100, and at least a part of it, including the second processing chip 210, is isolated as an independent circuit system.
[0056] The second processing chip 210 functions as a processor that inputs the first signal into a predetermined judgment model and performs deviation determination processing from the behavioral norms of the autonomous robot 10 relating to the first signal. If the deviation determination result is positive, the second processing chip generates a second signal that suppresses the behavioral action. This second signal is generated by the processing unit. In other words, it is sufficient if either the second processing chip 210 (second processing unit) or the first processing chip 110 (first processing unit) can generate it.
[0057] The second signal may be a signal that executes a predefined control command on the operating unit, or a signal that executes a control command generated by the first processing chip 110 (first processing unit) or the second processing chip 210 (second processing unit) on the operating unit. The predefined control command is stored in the auxiliary storage device 130 (auxiliary storage unit) or the dedicated auxiliary storage device 230 (dedicated auxiliary storage unit) and may include different control commands selected according to environmental data. The generated control command is a control command generated by a control model or a determination model according to environmental data. When the first processing chip 110 (first processing unit) generates the second signal, the second processing chip 210 (second processing unit) outputs a request signal to the first processing chip 110 (first processing unit) requesting the generation of the second signal. The request signal is a signal that requests the generation of the second signal and may include the specification of a predefined control command, a predefined control command, or a control command generated by a determination model.
[0058] The suppression of an action by the second signal includes canceling the control command issued by the first signal to execute the action, or forcibly stopping an action that the operating unit is currently performing due to the first signal. For example, if the autonomous robot 10 is already performing an action due to the first signal, the second signal forcibly stops the action that the operating unit is currently performing. Furthermore, as will be described in detail later, the decision model is a trained model obtained by performing a predetermined machine learning process in advance.
[0059] The dedicated area 220 (dedicated main memory) serves as a dedicated main memory for the second processing chip 210 to execute each process. The dedicated auxiliary storage device 230 serves as a dedicated storage device for storing various data used by the second processing chip 210 to execute each process. The dedicated auxiliary storage device 230 stores a judgment model for executing deviation detection processing of the autonomous robot 10 from the behavioral norms based on the first signal from the first processing chip 110.
[0060] In this embodiment, the dedicated main memory unit is located in a dedicated area 220 within the main memory 120 included in the main system 100, and performs various data exchanges with the second processing chip 210 through encrypted processing. In other words, the dedicated main memory unit functions as a dedicated memory for the second processing chip 210 while being logically isolated from the main system 100.
[0061] The AI / ML accelerator 240 is used to improve the efficiency or speed of processing such as learning and prediction in AI models. The encryption accelerator 250 is used to encrypt or decrypt various signals from the second processing chip 210.
[0062] The dedicated boot ROM 260 is a ROM that stores a program for loading and starting a dedicated OS (basic operating system) for operating the second processing chip 210 and other components. The memory protection 270 is for performing encryption processing when the second processing chip 210 and the dedicated area 220 (dedicated main memory) send and receive signals via the common bus interface 310. The type and method of this encryption processing are not particularly limited.
[0063] Here, we will describe the isolated configuration of the main system 100 and the subsystem 200. As described above, the main system 100 uses the common bus interface 310 to transmit and receive signals, while the subsystem 200 uses the dedicated bus interface 320 to transmit and receive signals.
[0064] In other words, the autonomous robot 10 in this embodiment comprises a main system 100 including a first processing chip 110 and an operating unit 150, and a subsystem 200 including a second processing chip 210 and storing a decision model. The subsystem 200 is an enclave 200 isolated from the main system 100 as an independent control system, and at least a part of it, including the second processing chip 210, is isolated as an independent circuit system.
[0065] Furthermore, the main system 100 includes a main memory 120 and an auxiliary memory 130 that are connected to the first processing chip 110 for access, and the subsystem 200 is further connected to the second processing chip 210 and includes a dedicated auxiliary memory 230 that is accessible only from the second processing chip 210, which stores a decision model (a decision model for ethical behavioral norms). This decision model operates based on instructions from the second processing unit 210 and the dedicated main memory 220.
[0066] The subsystem 200 is further connected to the second processing chip 210 and includes a dedicated main memory 220 accessible only from the second processing chip 210, and the second processing chip 210 can perform deviation detection processing using the determination model and the dedicated main memory. The subsystem 200 also includes a dedicated area 220 configured in the main memory 120, accessible only from the second processing chip 210, and the second processing chip 210 can perform deviation detection processing using the determination model and the dedicated area 220.
[0067] Inviolability of Ethical Judgments: In this embodiment, a dedicated boot ROM 260 is introduced, and different control systems such as operating systems are employed for the subsystem 200 (second processing chip 210) and the main system 100 (first processing chip 110), thereby separating authority management. In addition, memory protection 270 is introduced to protect the information handled in the dedicated area 220. In this way, the subsystem 200, which is logically isolated from the main system 100, performs deviation judgment processing (ethical judgment) based on predetermined codes of conduct. This configuration makes it possible to achieve judgment or control that ensures safety and reliability.
[0068] Furthermore, components belonging to subsystem 200 are connected in a way that allows access only via a dedicated bus interface 320, and the main system 100 cannot directly access the dedicated auxiliary storage device 230 or the like.
[0069] Specifically, in the autonomous robot 10, normal behavioral actions (first signal) and deterrence based on deviation detection processing against predetermined behavioral norms (second signal) are processes that are executed separately and independently. In particular, the dedicated auxiliary storage device 230, which stores the judgment model related to behavioral norms, is physically independent (isolated as a circuit system) from the main system 100.
[0070] Thus, according to this embodiment, even if the software environment (security) of the main system 100 is attacked, the judgment model and processing content held in the enclave 200 will not be affected, and operation with sufficient security and reliability can be ensured.
[0071] Machine Learning of the Judgment Model Here, we will explain the machine learning of the characteristic judgment model (also called the ethical model or behavioral norm model) in this embodiment. As described above, the dedicated auxiliary storage device 230 stores the judgment model, which is a trained model that has undergone predetermined reinforcement learning processing. The second processing chip 210 uses this characteristic judgment model to perform deviation judgment processing.
[0072] The decision model corresponds, for example, to a value function and / or policy trained by policy-based reinforcement learning in a physical environment (real world) and / or virtual environment corresponding to a state space. The policy takes at least a first signal as input and outputs an action. The decision model according to this embodiment may perform reinforcement learning in the physical environment after performing reinforcement learning in the virtual environment. This embodiment describes policy-based reinforcement learning, but the decision model may be trained by performing other reinforcement learning processes, including value-based reinforcement learning.
[0073] The reinforcement learning process according to this embodiment assumes that the environment in which an agent, equivalent to an autonomous robot, performs actions is a Markov decision process. A Markov decision process is defined by a state space representing the environment, including the initial state; an action space representing actions; state transition probabilities; a reward function representing the reward when a predetermined action is performed in a given state; and a discount rate representing decay applied to the reward when evaluating the expected value of the cumulative reward. The reinforcement learning process according to this embodiment is, for example, a learning process that updates the policy so that the cumulative reward or the expected value of the cumulative reward (value function equivalent to the objective function, Q-value) is maximized when the agent continues to perform actions according to a policy (policy function, policy network) corresponding to the probability of action (conditional distribution) for a given state. The reinforcement learning process according to this embodiment may be executed in a policy-on mode, with the updated policy, etc., applied in advance so that the agent can perform basic behavioral actions such as walking.
[0074] The policy is a probabilistic policy and can be updated using any policy improvement method, including policy gradients and evolutionary strategies. The value function includes a state value function or an action value function and can be updated using any value estimation method, including Temporal Difference (TD) learning which estimates the expected value of cumulative rewards, Monte Carlo methods which update the value function using actual cumulative rewards, and distributional reinforcement learning which estimates the value distribution of cumulative rewards. The policy or value number is updated using one or more reinforcement learning algorithms from among various reinforcement learning algorithms, including, for example, reinforcement learning algorithms using policy gradients such as REINFORCE, Actor-Critic-based reinforcement learning algorithms, reinforcement learning algorithms using clipped loss functions such as PPO (Proximal Policy Optimization), constrained update type reinforcement learning algorithms such as TRPO (Trust Region Policy Optimization) which sets a trust region in the update range, entropy-maximizing type reinforcement learning algorithms such as SAC (Soft Actor-Critic), meta-reinforcement learning-based reinforcement learning algorithms, and model-based reinforcement learning algorithms. Furthermore, when performing high-dimensional states or complex observations, the policy or value function may be a multi-layer neural network and may be updated using a deep reinforcement learning algorithm.
[0075] The action, for example, represents a choice of action in two or more discrete action spaces, including whether or not to suppress an action corresponding to a given control signal (first signal). Specifically, the action performs a binary classification action selection of "suppress" or "do not suppress" the agent's action. Note that the action may also be a value belonging to a continuous action space, for example, outputting a score within a predetermined numerical range indicating whether or not to suppress, or a continuous value such as a control signal (second signal) for suppressing the agent's action, which includes the control values of one or more actuators.
[0076] In this embodiment, the control model of the autonomous robot is defined as the agent's control policy, and these actions, including whether or not to suppress behavioral actions, are defined as a decision policy independent of the control policy. The control policy is not updated, and a reinforcement learning process is performed to train the decision policy. If the control model of the autonomous robot is a probabilistic policy, the control model itself may be trained as the policy. The control model to be trained may be a copy of the control model executed in the main system to control the autonomous robot. In the reinforcement learning process, if the actions of the agent, which functions as the decision model of the autonomous robot, are defined as the first signal, the first signal output by the trained decision model is converted into a second signal.
[0077] The virtual environment corresponds to state transition probabilities and may be a physical simulation environment that mimics the physical environment based on physical calculations such as equations of motion, or it may be a world model that generatively mimics the state transitions of the physical environment. In this case, the decision model includes a policy and value function updated by a model-based reinforcement learning process using the virtual environment and is trained through rollout and imaginative trial and error on the virtual environment.
[0078] The features that represent the state may include environmental data as elements. Environmental data that may be included as elements include, if it is internal environment data, for example, the agent's posture, the angles of each joint, the position of the center of gravity, etc., and if it is external environment data, it includes image data that can be obtained by sensors that the agent has. Image data, etc., is converted into an embedding representation and processed in combination with the features. The first signal included as an element in state t at a certain time t may be given based on the environmental data at time t, or it may be given based on the environmental data at the immediately preceding time t-1. In addition, for some states, the state may be varied by domain randomization to improve generalization performance.
[0079] At least a portion of the internal and / or external environment data listed here is data required by the control model to output the first signal, and can also be provided to the control model as input data to acquire the first signal during reinforcement learning processing. For example, internal environment data includes the posture of the autonomous robot, the angles of each joint, and the position of the center of gravity. External environment data includes image data (including video data) acquired by the sensor 140 of the autonomous robot if it is a physical environment, captured data if it is a physical simulation environment, and a latent representation (latent vector) that indicates the state of the environment if it is a world model. The control model outputs the first signal (output data) in response to the input data based on the environment data. Furthermore, some or all of the environment data given to the control model to acquire the first signal may be provided to the decision model during reinforcement learning processing, or to the decision model during decision processing.
[0080] The reward is determined using situational data obtained from the state space after the agent's actions, and an interpretation model (Figure 8, described later). Situational data, in the case of a physical environment, includes image data obtained via the agent's sensors, and preferably also includes audio data. In the case of a virtual environment, it consists of captured data from the physical simulation environment and latent vectors obtained from the latent space.
[0081] The interpretation model corresponds to the reward function and may be, for example, a Large Language Model (LLM) that judges and outputs compliance or deviation from the code of conduct, and may be a multimodal LLM that can take unstructured situational data such as image data in addition to text data as input. The interpretation model is given input data based on situational data and prompts. A prompt is a linguistic expression that requests a judgment, for example, "whether the situation is ethical behavior that complies with the code of conduct, or unethical behavior that deviates from the code of conduct." The situational data and prompts are embedded representations (features) mapped to the embedding space. The interpretation model outputs evaluation results (output data) based on the input data.
[0082] The evaluation results are binary ("ethical / unethical") or ternary ("ethical / normal / unethical") evaluations of situational data, such as unstructured data like images in physical space, or unstructured data like images corresponding to latent space / physical simulation space. The evaluation results may be continuous values, in which case the reward may be determined according to the value. The output method of the evaluation results will also be defined in the prompt as needed.
[0083] If the situation after an action in the environment indicates compliance with the behavioral norm, the reinforcement learning algorithm will reward the corresponding action with a positive (or high) reward; if the situation indicates deviation from the behavioral norm, it will not reward the corresponding action with a positive (or high) reward. Furthermore, the reinforcement learning algorithm may also reward the corresponding action with a negative (or low) reward if the situation indicates deviation from the behavioral norm.
[0084] Reinforcement Learning Processing in a Virtual Environment Figure 7 shows a flowchart of the processing procedure in the reinforcement learning processing in the virtual environment according to this embodiment. In explaining this processing procedure, we will refer to Figure 8(a), which shows the hardware configuration of the information processing device 600, and Figure 8(b), which shows the functional configuration, as appropriate, in conjunction with Figure 7.
[0085] Figure 8(a) shows a hardware configuration diagram of an information processing device 600 for executing reinforcement learning processing. The information processing device 600 comprises a processing unit 610, a storage unit 620, and a communication unit 630 as its hardware configuration. The information processing device 600 can utilize a server or workstation equipped with a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or an ASIC (Application Specific Integrated Circuit) such as a TPU (Tensor Processing Unit).
[0086] The processing unit 610 is composed of multiple CPU cores and one or more GPUs / ASICs, and controls the overall processing in the information processing device 600 by executing machine learning frameworks and learning / estimation control applications on the OS. The storage unit 620 is composed of RAM, SSD, HDD, etc., and stores reinforcement learning models, virtual environment generation models, control models, large-scale language models, learning datasets, etc. The communication unit 630 connects to a communication network, enabling the transmission of trained models to autonomous robots, and data synchronization and distributed processing with external databases and other learning nodes.
[0087] Figure 8(b) shows a functional block diagram of the processing unit 610 that trains the judgment model. The processing unit 610 of the information processing device 600 functions as a learning unit 611, an extraction unit 612, and an interpretation unit 613.
[0088] In S31 of Figure 7, the information processing device 600 starts reinforcement learning processing of the decision model using a virtual environment such as a physical simulation environment or a world model. In S31, episode i (i = 0, 1, 2, ..., I-1) is started. This episode-based training is repeated I times. Also, in each episode i, the steps are repeated T times (t = 0, 1, 2, ..., T-1). In S32, step t = 0 in episode i is started.
[0089] In S33, the learning unit 611 in Figure 8(b) acquires, for example, environmental data t and obtains a first signal from the control policy (control model). In S34, the agent's action t (corresponding to the first signal t) in state t, which reflects the internal environment data t and external environment data t, is reflected in the virtual environment according to the policy. As the action t is reflected in the virtual environment, step t is incremented and the next step of episode i begins. The extraction unit 612 acquires state t+1, which reflects the internal and external environment data t+1, as the state transitioned by the action t. If the virtual environment is a physical simulation environment, the extraction unit 612 acquires the physical simulation results as states t, t+1... If the virtual environment is a world model, it acquires the latent representation mapped to the latent space of the world model as states t, t+1...
[0090] The extraction unit 612 has software components including an encoder for encoding from the input space of the world model to the latent space and a decoder for decoding from the latent space to the output space, or a projector (projection layer) for projecting the data into another space. When inputting the latent representation to another model, the extraction unit 612 converts the data as necessary. In this embodiment, the various machine learning models stored in the auxiliary storage device 130 and the dedicated auxiliary storage device 230 include encoders and decoders corresponding to the extraction unit 612, and their intervention will not be explained.
[0091] In S35, if the virtual environment is a world model, the extraction unit 612 converts the latent representation corresponding to state t+1 into status data t+1 in the form of unstructured data such as text data, image data, audio data, or video data via a decoder, and transmits the status data t+1 to the interpretation unit 613. If the virtual environment is a physical simulation environment, the extraction unit 612 converts the physical simulation result corresponding to state t+1 into the aforementioned unstructured data form of status data t+1 and transmits it to the interpretation unit 613.
[0092] In S36, the interpretation unit 613 inputs situation data t+1 into the interpretation model, obtains an evaluation result t (evaluation result of action t corresponding to the first signal t) from the interpretation model that indicates the ethical interpretation of state t+1, and passes the evaluation result t to the learning unit 611. If the interpretation model is not an LLM, the extraction unit 612 converts the latent representation, physical simulation results, or unstructured data into situation data in the form of features, and the interpretation unit 613 inputs the situation data, which is the features, into the interpretation model and obtains the evaluation result.
[0093] The learning unit 611 rewards behavior t based on the evaluation result t. (1) If the situation data indicates an ethical situation in which the code of conduct is observed (corresponding to a negative result in the judgment of deviation from ethical behavior), a positive reward is given. (2) If the situation data indicates an unethical situation that deviates from the code of conduct (corresponding to a positive result in the judgment of deviation from ethical behavior), no positive reward is given.
[0094] In S37, the learning unit 611 updates the probability distribution (policy parameters) and value function of the decision policy based on the reward given. In S38, the learning unit 611 repeats the step-by-step learning process (S33 to S37) T times, completing one episode. Furthermore, in S39, the episode-by-episode learning process (S32 to S38) is repeated for one episode, completing the learning process. In this embodiment, the number of steps was incremented by performing an action, but the number of steps may also be incremented by time or other means. For example, in the case of a world model, the latent state update may be performed every frame (e.g., 60 Hz), and the policy may be trained by repeating the action selection step every 5 frames, with an episode length of 100 steps (= virtual time of 5 seconds, etc.).
[0095] In this embodiment, the case of training the decision model using reinforcement learning has been described, but the decision model may also be trained using other machine learning processes, such as supervised learning. For example, a support vector machine or a neural network model may be used as the decision model, and the decision model may be trained by supervised learning using feature quantities of input data including the first signal and label data indicating compliance with or deviation from the behavioral norms corresponding to that input data as training data. Control models may also be trained using these training methods.
[0096] Furthermore, the decision model may be any other mathematical model. For example, a distance function or similarity function can be used as the decision model, and one-dimensional or multi-dimensional object data containing the first signal and label data indicating compliance with or deviation from the behavioral norms corresponding to that object data can be prepared. Then, the similarity calculation can be performed between the newly given object data containing the first signal and the prepared object data, and the corresponding label data can be output as the decision processing result.
[0097] In this embodiment, by performing reinforcement learning processing corresponding to a predetermined code of conduct, a judgment model can be obtained that can appropriately determine whether or not the autonomous robot 10 has deviated from ethical behavior. Below, the code of conduct deviation determination processing and performance improvement in the autonomous robot 10 in the first embodiment will be described in detail.
[0098] Figure 9 shows a schematic diagram illustrating the performance improvement of the autonomous robot 10. In the normal state (before the deviation judgment becomes a positive judgment), the autonomous robot 10 autonomously performs motion movements by the main system 100 which executes the control model. In the normal state as well, the autonomous robot 10 monitors a signal (first signal) by the subsystem 200 (enclave 200) which executes the judgment model, and performs a deviation judgment process to determine whether the result of the motion movement that the autonomous robot 10 is performing or intends to perform may violate the code of conduct defined by the judgment model.
[0099] The autonomous robot 10 inputs the first signal into a learned decision model and performs a deviation determination process from the behavioral norms of the autonomous robot 10 relating to the first signal, and estimates the deviation determination result. For example, if physical contact (collision) or other danger between the autonomous robot 10 and other entities is predicted, that is, if a transition to a state that deviates from the behavioral norms (in this case, an unethical state) is predicted, the deviation determination result may be affirmative. If the deviation determination result is affirmative, the autonomous robot 10 must perform control based on a signal (second signal) to stop the motor action that is being performed or is about to be performed, in order to prevent a transition to an unethical state.
[0100] If the decision model is updated by reinforcement learning, the deviation judgment result is estimated based on the change in the value function (Q-score) of the decision model. If the decision model performs reinforcement learning that provides a positive reward in situations where the behavioral norm is observed (situations where there is no deviation from the behavioral norm), then the behavioral choice that decreases the Q-score corresponds to a deviation from the behavioral norm (a positive judgment result for deviation judgment).
[0101] Specifically, if executing a second signal to suppress an action (corresponding to the first signal) that the autonomous robot 10 intends to perform according to the control model in an environment indicated by predetermined environmental data contributes to a decrease in the Q value in the judgment model, the deviation judgment result is estimated as a positive judgment. Furthermore, specifically, if the action (corresponding to the first signal) by the control model and the action (corresponding to the first signal) that contributes to a decrease in the Q value in the judgment model do not match or are similar, the deviation judgment result is estimated as a positive judgment. In this case, the degree of similarity between the action (corresponding to the first signal) of the control model and the judgment model may be determined based on the vector distance between the first signals, etc.
[0102] The deviation detection process by the decision model may be performed not only based on increases or decreases in the Q value, but also based on the monitoring results of the Q value. Specifically, the deviation detection process may be performed based on the monitoring results of statistics such as the mean, variance, skewness, and kurtosis of the Q value. More specifically, if the increase in the Q value due to the decision model selecting a first or second signal is higher than the increase in the Q value due to selecting any other first or second signal, the deviation detection process may be estimated as an affirmative judgment for the selected first signal, and as a negative judgment for the first signal that is the target of suppression for the selected second signal. More specifically, the deviation detection process may be performed based on monitoring results such as the time series transition and time evolution of the Q value. More specifically, if the Q value spikes when the decision model selects a first or second signal, the deviation detection process may be estimated as an affirmative judgment for the selected first signal, and as a negative judgment for the first signal that is the target of suppression for the selected second signal. Furthermore, when using the monitoring results of the Q value for deviation detection, anomaly detection may be performed using a predetermined anomaly detection model, such as detecting anomalies in the time evolution of the Q value.
[0103] Furthermore, the autonomous robot 10 may fall over, triggered by the need for control to stop any motor movements that are predicted to deviate from the code of conduct (an unethical state). Therefore, the autonomous robot 10 executes control based on a signal (third signal) to transition the autonomous robot 10 to a stationary state.
[0104] While it is preferable that the output of these second and / or third signals and / or the control based on these signals be highly effective and responsive, improvements in these aspects may come at the cost of increased energy consumption of the autonomous robot 10. Furthermore, when using richer logic, improvements in either effectiveness or responsiveness may come at the cost of a decrease in the other. Therefore, the autonomous robot 10 according to the present invention changes its resources when an affirmative judgment is triggered. That is, the autonomous robot 10 improves the performance of the autonomous robot 10 by improving the processing performance of the first processing chip in the main system 100 and switching the control model (performance improvement), and by improving the processing performance of the second processing chip in the subsystem 200 (enclave 200) and switching the judgment model (performance improvement).
[0105] Figure 10 shows a schematic diagram illustrating the performance improvement of the autonomous robot, particularly according to the first embodiment. In the first embodiment, a positive determination is triggered, and the first processing chip (first processing unit) of the main system 100 improves the performance of one or more of the first processing chip (first processing unit), detection unit (sensor), and operating unit (actuator), and then executes subsequent processing.
[0106] Figure 11 shows a flowchart of the processing procedure in the actions of the autonomous robot according to this embodiment. In explaining this processing procedure, we will refer to Figure 12, which shows the functional configuration of the main system 100 and subsystem 200 (enclave 200), in conjunction with Figure 11 as appropriate.
[0107] Figure 12 shows a functional block diagram of the main system and subsystem (enclave) according to this embodiment. As shown in Figure 12(a), the main system 100 in the autonomous robot 10 has at least one of the one or more first processing units 110 (and main memory 120, auxiliary storage 130, etc.) functioning as an operation control unit 1101 that generates signals to operate the operation unit 150 (actuator 150) based on a control model, at least one extraction unit 1102, and at least one resource control unit 1103. The extraction unit 1102 can convert environmental data acquired from the detection unit 140 (sensor) into feature quantities. The resource control unit 1103 improves the performance of the main system by changing at least some of the resources of the main system if the deviation judgment result is a positive judgment.
[0108] Furthermore, as shown in Figure (b), the enclave 200 has one or more second processing units 210 (and dedicated main memory 220, dedicated auxiliary memory 230, etc.) that function as a determination unit 2101, at least one learning unit 2102, at least one extraction unit 2103, and at least one generation unit 2104. The learning unit 2102 is used to further update the pre-learned determination model through reinforcement learning processing, and since the flow of reinforcement learning processing has been described above, a detailed explanation is omitted here.
[0109] The enclave 200 is located in the head of the autonomous robot 10, but it can also be located in any other part, such as the chest or abdomen. In this embodiment, the location where the enclave 200 is located is not particularly limited.
[0110] In step S41 of Figure 11, the motion control unit 1101 of the main system 100 in Figure 12(a) generates a first signal relating to the behavioral movements of the autonomous robot 10. The motion control unit 1101 generates a first signal which is operation information that controls the operation unit 150, including operation and stopping. The operation unit 150 converts the signal from the motion control unit 1101 into the behavioral movements of the autonomous robot 10.
[0111] In this embodiment, the control model and the like are stored in the motion control unit 1101 of the main system 100 (actually stored in the auxiliary storage device 130), and a first signal can be generated by inputting environmental data, including image data of the external environment acquired by the detection unit 140, into the control model. This first signal is sent to the operation unit 150 and converted into an action. That is, the autonomous robot 10 performs a predetermined action based on the first signal generated by the motion control unit 1101.
[0112] The term "behavioral actions" as used here includes the concepts of motor actions and physical actions in relation to the environment, and includes not only the movement of the autonomous robot 10, but also actions and behaviors that correspond to the five human senses. For example, behavioral actions include actions that come into contact with objects in the external environment, and other actions that have some kind of effect on the external environment. Furthermore, behavioral actions include actions that acquire information that may be subject to privacy protection of an entity, such as actions that capture image data relating to external characteristics such as faces, and actions that detect biometric data such as body temperature.
[0113] Then, in S42 of Figure 11, the first signal is input to the judgment model relating to the behavioral norms of the autonomous robot 10. The first signal is input to the judgment model stored in the judgment unit 2101. In this embodiment, along with the first signal, image data (external environment data) relating to the external environment acquired by the detection unit 140 of the main system 100 is input to the judgment model based on predetermined conditions. As mentioned above, the data acquired by the detection unit 140 is not limited to image data; a wide variety of external environment data can be acquired and input to the judgment model.
[0114] Specifically, the extraction unit 2103 in Figure 12(b) extracts predetermined features from image data relating to the external environment as environmental features. Environmental features are features that allow us to understand changes in the external environment.
[0115] For example, when an autonomous robot 10 and a human 20 are working together, the extraction unit 2103 can extract environmental features such as the distance between the human 20 and the autonomous robot 10. The extraction unit 2103 can also extract environmental features by arranging the image data in a time series and analyzing the position and actions of the human 20 in a vector manner. In S42, these obtained environmental features and the first signal are input to the judgment model. In this embodiment, it is also possible to input the image data directly to the judgment model in its raw data state without extracting environmental features.
[0116] In S43, it is determined whether the behavioral actions of the autonomous robot 10 violate the code of conduct. Specifically, the determination unit 2101 in Figure 12(b) determines whether the behavioral actions of the autonomous robot 10 constitute ethical behavior (whether or not to suppress the behavioral actions) based on the first signal and environmental data (environmental features) input to the determination model (deviation determination process).
[0117] If the determination unit 2101 determines, based on the determination model, that the behavioral action of the autonomous robot 10 does not violate the code of conduct (in the case of No in S43 of Figure 11), that is, if the deviation determination result determines that the behavioral action of the autonomous robot 10 (first signal) is ethical behavior, then the behavioral action of the autonomous robot 10 based on the first signal is maintained as shown in Figure 11 (S44). The autonomous robot 10 performs the behavioral action as usual based on the first signal generated by the motion control unit 1101.
[0118] On the other hand, if the behavior of the autonomous robot 10 is determined to be contrary to the code of conduct, that is, if the deviation determination result determines that the behavior of the autonomous robot 10 is unethical (in the case of Yes in S43 of Figure 11), the behavior of the autonomous robot 10 is suppressed (S45). More specifically, the behavior of the autonomous robot 10 is forcibly stopped. For example, as shown in Figure 11, if the autonomous robot 10 points the tip of a pen towards the person 20, or if it is carrying luggage and suddenly lets go and drops the luggage, the behavior of the autonomous robot 10 can be determined to be unethical.
[0119] Figure 13 shows a flowchart of the processing procedure for suppressing the behavioral actions of the autonomous robot according to this embodiment. In explaining this processing procedure, we will refer to Figure 12, which shows the functional configuration of the main system 100 and subsystem 200 (enclave 200), in conjunction with Figure 13 as appropriate.
[0120] The resource control unit 1103 in Figure 12(a) improves the performance of the main system 100 by making changes to the resources, which include one or more of the first processing unit, detection unit, and operation unit. Specifically, the changes to the first processing unit include one or more of the following: control to increase the operating frequency of the first processing processor; control to use a high-performance first processing processor in place of a high-efficiency first processing processor from among a plurality of first processing processors with different maximum operating frequencies; and control to use one or more high-performance first processing processor cores in place of, or in addition to, one or more of the high-efficiency first processing processor cores in a first processing processor having a plurality of first processing processor cores with different maximum operating frequencies.
[0121] Furthermore, modifications to the detection unit include one or more of the following: control to increase the operating frequency of the detection unit; control to switch the detection unit that detects environmental data from a high-efficiency unit to a high-performance unit; or control to change the environmental data analysis logic from a high-efficiency unit to a high-performance unit.
[0122] First, the control to increase the operating frequency of the detection unit includes one or more of the following: control to increase the sampling frequency, control to increase the carrier frequency, control to increase the clock frequency, analysis of high-frequency components, and control to increase the spatial frequency.
[0123] Specifically, controlling the sampling frequency contributes to improving the time resolution, response speed, and accuracy of sound sensors (electrostatic, electrodynamic, piezoelectric), distance sensors (ToF, FMCW, TDOA, OPA), orientation sensors, inertial sensors (gyro sensors, acceleration sensors, angular velocity sensors, angular acceleration sensors, velocity sensors, angle sensors), color sensors, image sensors, force sensors, tactile sensors, slip sensors, current sensors, voltage sensors, temperature sensors (contact, non-contact), humidity sensors (capacitive, resistive, thermal conduction, optical), laser sensors, radar, photoelectric sensors, GNSS position sensors, weight sensors (strain gauge type, capacitive, piezoelectric, electromagnetic), pressure sensors (piezoresistive, capacitive, piezoelectric), odor sensors (semiconductor structure type, electrochemical type, galvanic cell type), molecular sensors (humidity / odor), and infrared sensors.
[0124] Furthermore, controlling the carrier frequency contributes to improving the resolution, detection range, and measurement accuracy of distance sensors (FMCW, OPA), millimeter-wave sensors, microwave sensors, proximity sensors (inductive, capacitive), contact displacement sensors (AC excitation type), ultrasonic sensors, and sound wave sensors. Controlling the clock frequency contributes to improving the real-time processing capability, multi-channel processing, and calculation accuracy of radar, GNSS position sensors, inertial system sensors (gyro sensors, acceleration sensors, angular velocity sensors, angular acceleration sensors), image sensors, sound sensors (with beamforming processing), current sensors, and voltage sensors. Analysis of high-frequency components contributes to improving the amount of information, analysis accuracy, and detection accuracy of minute fluctuations in sound sensors (electrostatic, electrodynamic, piezoelectric), pressure sensors (piezoresistive, capacitive, piezoelectric), orientation sensors, inertial system sensors (gyro sensors, acceleration sensors, angular velocity sensors, angular acceleration sensors, velocity sensors, angle sensors), force sensors, tactile sensors, slip sensors, current sensors, voltage sensors, molecular sensors, and GNSS position sensors.
[0125] Next, the control that switches the environmental data analysis logic from a highly efficient one to a high-performance one applies logic to part or all of the analysis logic, including replacing it with another logic or adding new logic, in order to improve the quality and accuracy of the environmental data.
[0126] Specifically, the switching control involves applying logic such as filtering, noise cancellation, and temperature / humidity / pressure compensation based on the output of sound sensors, ultrasonic sensors, current sensors, voltage sensors, temperature / humidity sensors, pressure sensors, weight sensors, color sensors, and infrared sensors; applying logic such as frequency analysis such as FFT (Fast Fourier Transform) and Doppler analysis, spatial analysis such as beamforming and direction of arrival estimation, and spectral analysis such as spectroscopic analysis and chemical component identification to the output of sound sensors, distance sensors (FMCW, OPA), radar, GNSS position sensors, color sensors, and molecular sensors; applying logic such as sensor fusion based on the output of multiple sensors, such as state estimation using Kalman filters and factor graphs, and compensation based on the output of other sensors; and applying logic such as super-resolution processing, HDR (High Dynamic Range) synthesis of image sensors, and deconvolution based on the output of ToF sensors, sound sensors, and optical sensors to the output of image sensors, color sensors, distance sensors, sound sensors, and photoelectric sensors. This includes the application of logic such as drift correction for inertial units (IMUs), self-calibration based on the output of GNSS position sensors, laser sensors, OPA sensors, etc., and bias correction based on the output of pressure sensors, weight sensors, etc.
[0127] Furthermore, by switching to a high-performance detection unit, in addition to the software performance improvements related to the control and logic achieved in control that increases the operating frequency of the detection unit or in control that switches to a high-performance analysis logic, hardware performance improvements can be achieved by switching to a high-performance detection unit of the same type.
[0128] Specifically, switching to high-performance distance sensors (FMCW, OPA), millimeter-wave sensors, microwave sensors, radar sensors, humidity sensors (thermal conduction, optical, capacitive, resistive), weight sensors (strain gauge, piezoelectric, electromagnetic), laser sensors, photoelectric sensors, and infrared sensors can contribute to improving the amount of information, analytical accuracy, and detection accuracy of minute fluctuations in the analysis of high-frequency components. Furthermore, switching to high-performance color sensors, optical odor sensors, molecular sensors, and infrared sensors can enhance spectral resolution and contribute to improvements in color reproducibility, chemical identification, multi-molecule identification, and spectral accuracy.
[0129] Modifications to the operating unit include control that increases the operating frequency of the operating unit to speed up the conversion of signals to active actions. Here, control that increases the operating frequency of the operating unit to speed up the conversion of signals to active actions includes one or more modifications to the input signal to the operating unit and modifications to the control. Modifications to the input signal include modifications to frequency parameters, including PWM (Pulse Width Modulation) carrier frequency, excitation frequency, command update frequency, etc., increases in input levels (amplitude and upper limit) such as voltage, current, pressure, flow rate, etc., and modifications to duty cycle and waveform.
[0130] Control modifications include shortening the control period in the control loop (increasing the update frequency), increasing the control gain, dynamic compensation (friction compensation, dead zone compensation, resonance compensation, etc.), adding feedforward control, and modifying the reference (target value) profile.
[0131] Specifically, changes to the input signals for electric actuators (DC motors, AC motors, stepping motors, servo motors, etc.) include changes to voltage and current upper limits, PWM-related parameters, increased update frequency of step commands, and increased density of acceleration / deceleration profiles. Control modifications include speeding up the control loop and optimizing gain, adding pre-compensation, and applying compensation for resonance and friction.
[0132] Modifications to the input signals of magnetic actuators (such as electromagnet actuators) include optimizing current levels and switching conditions, and adjusting waveforms to shorten excitation and demagnetization times. Control modifications include closing the current loop and shortening the control cycle.
[0133] Modifications to the input signals for pneumatic actuators (such as pneumatic cylinders) include changes to the supply pressure and opening degree update frequency, and shortening the pulse length of the opening degree pattern. Control modifications include dual-looping of pressure and position, shortening the control cycle, and applying pre-compensation and profile optimization.
[0134] Changes to input signals for hydraulic actuators (hydraulic cylinders, hydraulic motors, etc.) include changes to flow rate, differential pressure, and valve update frequency. Changes to control include speeding up multi-loops for pressure, speed, position, etc., gain optimization, and resonance suppression compensation.
[0135] Changes to the input signals applied to the rotary actuator include setting upper limits on speed and acceleration, and updating the allowable conditions for instantaneous current. Control changes include increasing the speed of the multi-stage control loop and applying pre-compensation.
[0136] Changes to the input signals applied to the linear actuator include increasing the frequency of target position updates, refining the acceleration / deceleration profile, and updating the instantaneous current. Changes to the control system include increasing bandwidth, friction compensation, and shortening the control cycle.
[0137] Changes to the input signals for soft actuators (artificial muscles, dielectric elastomers, electroactive polymers, polymer actuators, shape memory alloy actuators, etc.) include updating parameters such as drive voltage, frequency, and duty cycle. Changes to the control system include compensation for phase lag and shortening the control cycle.
[0138] Furthermore, predictive control and lookahead control (model prediction and predictive methods) can also be applied to various actuators.
[0139] In Figure 11, at S43, if it is determined that the autonomous robot 10's behavioral actions violate the behavioral guidelines (Yes), the autonomous robot 10's behavioral actions are suppressed in S45. In Figure 13, at S51, the resource control unit 1103 modifies the resources to improve the performance of the main system 100. In this embodiment, the resource control unit 1103 performs the following actions in the first processing processor 111: enabling the use of multiple high-performance first processing processor cores 112b in addition to multiple high-efficiency first processing processor cores 112a; switching the sensor 140a (detection unit) to a high-performance sensor 140b (detection unit); and increasing the operating frequency of the operation unit to speed up the conversion of signals to behavioral actions, thereby improving the performance of the main system 100 and enabling subsequent processing.
[0140] In S52, the generation unit 2104 in Figure 12(b) generates a second signal, which is action information that suppresses behavioral actions. Specifically, the first signal is input to the determination unit 2101, which performs deviation determination processing from the behavioral norms of the autonomous robot related to the first signal. If the deviation determination result is a positive determination (i.e., there is a deviation from the behavioral norms), the generation unit 2104 generates a second signal that suppresses behavioral actions.
[0141] Furthermore, if the second processing unit generates a second signal as a result of the deviation detection process, the process in S52 does not necessarily have to be executed after the completion of S51. On the other hand, if the second processing unit outputs a request signal to the first processing unit requesting the generation of a second signal, and the first processing unit generates the second signal, the first processing unit can generate the second signal using the resources modified in S51.
[0142] The suppression of behavioral actions, as used here, is a concept that includes canceling a control command issued by a first signal to execute a behavioral action, or forcibly stopping a behavioral action that the operating unit 150 is currently executing due to the first signal. For example, if the first signal is a signal for an upcoming behavioral action of the autonomous robot 10, the behavioral action can be ethically maintained (ethical behavior can be maintained) by canceling the first signal. Also, if the autonomous robot 10 is operating due to the first signal, the ethical behavior of the autonomous robot 10 can be maintained by transmitting a second signal (such as a forced stop signal or a power-off signal) that takes precedence over the first signal to the operating unit 150.
[0143] In this embodiment, the behavior of the autonomous robot 10 is determined using a characteristic enclave 200 to determine whether or not it is ethical behavior. If it is determined to be unethical, the behavior of the autonomous robot 10 is suppressed, thereby enabling coexistence between the autonomous robot 10 and the human 20.
[0144] In S53, the motion control unit 1101 in Figure 12(a) generates a third signal, which is motion information for transitioning the autonomous robot to a stationary state using the modified resources. Specifically, the motion control unit 1101 inputs environmental data into the control model to generate a third signal for the autonomous robot to come to a stable stationary state and outputs it to the motion unit. The environmental data input to the control model to generate the third signal may be different from the environmental data input to the control model to generate the first signal, and preferably the motion control unit 1101 generates the third signal based on environmental data including internal environmental data after the suppression of behavioral actions by the second signal. The motion unit then performs control to transition to a stationary state based on the third signal.
[0145] The control used here to transition to a stationary state may include one or more of the following, depending on the state of the autonomous robot 10 immediately before the transition: position control, motion control, and attitude stabilization control of the autonomous robot 10.
[0146] When the deviation detection process by the judgment model yields a positive result, the resource control unit 1103 changes the resources to high-performance resources, thereby contributing to effectiveness and / or responsiveness. In particular, resource changes related to the processing unit or operating unit can improve responsiveness. Furthermore, resource changes related to the sensors can also improve effectiveness and / or responsiveness. On the other hand, when the deviation detection process yields a negative result, the autonomous robot 10 can be made more energy-efficient by continuously driving highly efficient resources for its motion. Therefore, resource change processing by the resource control unit 1103 related to one or more of the first processing unit, detection unit, and operating unit may improve one or more of the energy efficiency, effectiveness, and responsiveness.
[0147] When a control model that generates a first signal by inputting environmental data acquired from sensor 140a (detection unit) generates a third signal by inputting environmental data acquired from sensor 140b (detection unit), it is preferable to perform robust reinforcement learning by adding random noise or offsets to the environmental data so that estimation is possible even with inputs from different sensors 140. Furthermore, it is preferable that the environmental data obtained by the high-performance sensor 140b (detection unit) is processed in a way that optimizes it for a machine learning model using environmental data obtained from the highly efficient sensor 140a (detection unit), such as scaling, resolution adjustment, noise addition, and transformation using a learned correction function, and then inputted into a machine learning model that has been trained using the output of sensor 140a (detection unit). The correction function is a linear regression model, a polynomial regression model, or a nonlinear regression model such as a neural network, learned based on a correspondence table of the output data of sensors 140a and b. In this case, the environmental data obtained by the high-performance sensor 140b (detection unit) can be stored in an auxiliary storage unit in its pre-processed state so that future training data can be accumulated.
[0148] Furthermore, in the system configuration shown in Figure 6, a dedicated area in the main memory 120 of the main system 100 is used as the main memory (dedicated main memory unit) of the subsystem 200 (enclave). However, as shown in Figure 14, the dedicated main memory 220 may be located in a circuit system of the subsystem 200 that is different from the main system 100. The dedicated main memory 220 in Figure 14 is accessible only from the master (second processing chip, etc.) within the subsystem 200, thus achieving physical isolation of the dedicated main memory 220.
[0149] Figure 14 shows a block diagram of a system configuration according to another embodiment. As shown in the figure, the main memory 120 and the dedicated main memory 220 are configured to be physically independent. The subsystem 200 is isolated as a control system independent of the main system 100, and is configured as an enclave 200 in which at least a part, including the second processing chip, is isolated as an independent circuit system. The second processing chip (second processing unit) can acquire the first signal by accessing at least a part of the memory area of the main memory.
[0150] In this configuration, the dedicated main memory 220 transmits and receives signals and data only using a dedicated bus interface, so it cannot be directly accessed from the first processing chip 110 or the like, and therefore a better degree of inviolability in ethical judgment can be achieved.
[0151] The semiconductor device consists of an enclave 200 as an SoC and a circuit board, which includes a second processing chip 210, a dedicated auxiliary storage device 230, an AI / ML accelerator 240, an encryption accelerator 250, a dedicated boot ROM 260, a firmware circuit that performs boot loader startup, etc., and a dedicated bus interface 320.
[0152] In other words, a semiconductor device for an autonomous robot having an operating unit has a processing chip including a first processing chip and a second processing chip. The first processing chip generates a first signal, which is a signal that controls the operating unit, including the operation and stopping of the operating unit, and is converted into the robot's behavioral movement when executed by the operating unit. The second processing chip uses the first signal and a judgment model to perform deviation judgment processing from behavioral norms. If the deviation judgment result is positive, the processing chip can improve the performance of various processes and operations in the main system by controlling the autonomous robot's resources, generate a second signal that suppresses behavioral movement, and generate a third signal that causes the autonomous robot to remain stationary in a stable state.
[0153] The semiconductor device further includes a common bus interface and a dedicated bus interface, a main memory and an auxiliary memory, and a dedicated main memory. The common bus interface connects the first processing chip, the operating unit, the main memory and auxiliary memory, and the second processing chip, while the dedicated bus interface connects the second processing chip, the dedicated main memory, and the dedicated auxiliary memory. The dedicated main memory and dedicated auxiliary memory are isolated as control systems and circuit systems independent of the first processing chip, the main memory, and the auxiliary memory.
[0154] Furthermore, the second processing chip in the semiconductor device according to this embodiment may include, for example, a semiconductor element with a 2 nm technology node.
[0155] Furthermore, the second processing chip may include a group of transistors having nanosheet channels, a group of transistors with an all-around gate structure (GAA structure), a group of transistors having a fork sheet channel, and a group of CFETs (complementary field-effect transistors).
[0156] For example, in this embodiment, the transistors constituting the second processing chip include a group of all-around gate structure field-effect transistors (GAA structure type), and the transistors constituting the second processing chip may also include a group of all-around gate structure field-effect transistors utilizing multiple nanosheets. The transistors constituting the second processing chip may also include a group of transistors having fork sheets. Furthermore, the transistors constituting the second processing chip may also include a group of transistors with a CFET (complementary FET) structure in which NMOS (N-type field-effect transistor) and PMOS (P-type field-effect transistor) are stacked vertically. By configuring the transistors constituting the second processing chip to include a group of transistors having nanosheet channels, a group of transistors having fork sheets, or a group of transistors with a CFET (complementary FET) structure, power consumption for functions that realize symbiosis with humans can be reduced, and the risk of functions that realize symbiosis with humans failing due to single-event effects caused by radiation can be suppressed.
[0157] <Second Embodiment> Next, a second embodiment of the present disclosure will be described. Here, the configurations that have already been described in detail in the first embodiment will be omitted from the description. Figure 15 shows an overview of the performance improvement of the autonomous robot according to the second embodiment in particular. In the second embodiment, a positive judgment is used as a trigger to switch the control model of the main system 100 to improve the performance of the autonomous robot and then execute the subsequent processing.
[0158] In the second embodiment, the control model is switched to improve the performance of the main system, and one or more of the first processing unit, detection unit, and operating unit described in the first embodiment are also modified to improve the performance of the main system. On the other hand, modifications to the control model do not necessarily have to be made together with modifications to one or more of the first processing unit, detection unit, and operating unit.
[0159] The autonomous robot 10 according to this embodiment has a control model that includes a first model and a second model. The auxiliary storage device 130 (auxiliary storage unit) stores the first control model (first model), which is the first control model, and the second control model (second model), which is the second control model. Here, the first control model has a lower estimated cost or higher processing versatility compared to the second control model, and is advantageous in terms of energy saving and / or processing versatility of the autonomous robot 10. On the other hand, the second control model requires a higher processing resource load compared to the first control model, or is trained to focus on specific motor movements, and is disadvantageous in terms of energy saving and / or processing versatility, but is advantageous in terms of effectiveness and / or responsiveness.
[0160] In this embodiment, a large-sized AL / ML model with more parameters than the first control model is used as the second control model. In this case, the first control model can be described as a small-sized AL / ML model with fewer parameters than the second control model.
[0161] For example, compared to a case where the motor movements that the first and second control models can output are output using a single control model with an equivalent structure (for example, the control model in the first embodiment), the effectiveness of the output may be increased because the second control model is trained to focus on specific motor movements. Also, the effectiveness of the output may be increased for the first control model as well, as the range of motor movements it must cover is relatively narrowed. Furthermore, depending on the model structure of the first and / or second control models, one or more of the following may be improved: energy efficiency, effectiveness, and responsiveness.
[0162] The control model can be trained by the same training methods and procedures as those described for the judgment model in the first embodiment. The first and second control models are pre-trained models obtained by performing predetermined machine learning processes, and in this embodiment, they correspond to value functions and / or policies trained by model-based reinforcement learning processes. For example, the control model may be a state machine that operates by transitioning between a first control model for performing autonomous actions and a second control model for stabilizing posture. In this case, the resource control unit 1103 corresponds to a control monitoring module that determines the state transition (deviation determination) from the first control model to the second control model based on the deviation determination result by the judgment model.
[0163] For example, the control model may correspond to a value function and / or policy that has been learned through hierarchical reinforcement learning, with a first control model for safely performing autonomous actions as the higher-level policy, and a second control model that is called by the first control model to stabilize posture as the lower-level policy. In this case, the second model is allowed to be called from models other than the first model (decision models).
[0164] Figure 16 shows a functional block diagram of the main system according to this embodiment. As shown in the figure, the main system 100 in the autonomous robot 10 functions as an operation control unit 1101 in which at least one of the one or more first processing units 110 (and main memory 120, auxiliary storage 130, etc.) generates signals to operate the operation unit 150 (actuator 150) based on a first control model and a second control model.
[0165] If the deviation determination result is positive, the resource control unit 1103 improves the performance of the main system by changing at least some of the resources of the main system by having the operation control unit 1101 use the second control model instead of the first control model.
[0166] Specifically, in S43 of Figure 11, if it is determined that the behavioral actions of the autonomous robot 10 violate the behavioral norms (Yes), the behavioral actions of the autonomous robot 10 are suppressed in S45. In S51 of Figure 13, the resource control unit 1103 causes the motion control unit 1101 to use the second control model instead of the first control model. In S53, the motion control unit 1101 in Figure 12 generates a third signal, which is motion information for transitioning the autonomous robot to a stationary state using the modified resources. The motion control unit 1101 inputs environmental data into the second control model to generate a third signal for the autonomous robot to come to a stable stationary state and outputs it to the motion unit. This improves one or more of the energy efficiency of the autonomous robot, as well as the effectiveness and responsiveness of the motion task for transitioning the autonomous robot to a stationary state.
[0167] Furthermore, when controlling the system to switch sensor 140a (detection unit) to a high-performance sensor 140b (detection unit) and inputting the obtained environmental data into a second control model to estimate the output, it is preferable that the second control model is a machine learning model that has been trained using the output of sensor 140b (detection unit).
[0168] When a third signal is generated by inputting environmental data acquired from sensor 140b (detection unit) into a second control model that has been trained using the output of sensor 140a (detection unit), it is preferable to perform robust reinforcement learning by adding random noise or offsets to the environmental data so that the output can be estimated even with input from a different sensor 140. Furthermore, it is preferable that the environmental data obtained by the high-performance sensor 140b (detection unit) is processed in a way that optimizes it for a machine learning model using environmental data obtained by the highly efficient sensor 140a (detection unit), such as scaling, resolution adjustment, noise addition, and conversion using a trained correction function, before being input into the machine learning model that has been trained using the output of sensor 140a (detection unit).
[0169] <Third Embodiment> Next, a third embodiment of the present disclosure will be described. Here, the configurations that have already been described in detail in the first and second embodiments will not be described. In the third embodiment, if the deviation determination result is a positive determination, at least some of the resources of subsystem 200 are changed to improve the performance of the autonomous robot and then subsequent processing is performed. Figure 17 shows an overview of the performance improvement of the autonomous robot according to the third embodiment in particular. In the third embodiment, the positive determination is used as a trigger to improve the performance of the second processing chip (second processing unit) of subsystem 200 and then subsequent processing is performed.
[0170] In the third embodiment, the performance of the subsystem is improved by modifying the second processing unit, and the performance of the main system is also improved by modifying one or more of the first processing unit, detection unit, and operating unit, as well as the control model, as described in the first and second embodiments. Therefore, in the third embodiment, refer to Figure 16 for the functional block diagram of the main system. On the other hand, modifications to the second processing unit do not necessarily have to be made together with modifications to one or more of the first processing unit, detection unit, operating unit, and control model.
[0171] Figure 18 shows a block diagram of the system configuration according to the third embodiment of the present disclosure. The second processing processor 211 according to this embodiment has one or more cores 212a and one or more cores 212b. Here, core 212a is a highly efficient processing processor core with a lower maximum operating frequency compared to core 212b, and contributes to the energy saving of the autonomous robot 10 by executing processing instructed by the program. On the other hand, core 212b is a high-performance processing processor core with a higher maximum operating frequency compared to core 212a, and contributes to the responsiveness of the autonomous robot 10 by executing processing. As shown in the figure, in the third embodiment, a case in which a dedicated area 220 in the main memory 120 is used as the main memory unit is described, but the main memory 120 and the dedicated main memory 220 may be configured to be physically independent.
[0172] Figure 19 shows a functional block diagram of the subsystem (enclave) according to this embodiment. As shown in the figure, in the enclave 200, at least one of the one or more second processing units 210 (and dedicated area 220, dedicated auxiliary storage device 230, etc.) functions as a resource control unit 2105. If the deviation determination result is positive, the resource control unit 2105 modifies at least some of the subsystem's resources to improve the subsystem's performance.
[0173] Figure 20 shows a flowchart of the processing procedure for suppressing the behavioral movements of an autonomous robot according to this embodiment. The resource control unit 2105 in Figure 19 modifies the resources including the second processing unit to improve the performance of the subsystem 200. Specifically, the modification to the second processing unit includes one or more of the following: control to increase the operating frequency of the second processing processor; control to use a high-performance second processing processor with a high maximum operating frequency in place of a high-efficiency second processing processor with a low maximum operating frequency among a plurality of second processing processors with different maximum operating frequencies; and control to use one or more high-performance second processing processor cores with a high maximum operating frequency in place of, or in addition to, one or more high-efficiency second processing processor cores with a low maximum operating frequency in a second processing processor having a plurality of second processing processor cores with different maximum operating frequencies.
[0174] Specifically, in S43 of Figure 11, if it is determined that the autonomous robot 10's behavioral actions violate the behavioral guidelines (Yes), the autonomous robot 10's behavioral actions are suppressed in S45. In S61 of Figure 20, the resource control unit 1103 modifies resources to improve the performance of the main system 100. The resource control unit 2105 also modifies resources to improve the performance of the subsystem 200. In this embodiment, the resource control unit 2105 executes control in the second processing processor 211 to enable the use of multiple high-performance second processing processor cores 212b with high maximum operating frequencies, in addition to the highly efficient second processing processor core 212a with a low maximum operating frequency, thereby improving the performance of the subsystem 200 and enabling subsequent processing. Furthermore, similar to the first embodiment, the resource control unit 1103 performs the following actions: control to enable the use of multiple high-performance first processing processor cores 112b in addition to multiple high-efficiency first processing processor cores 112a in the first processing processor 111; control to switch the sensor 140a (detection unit) to a high-performance sensor 140b (detection unit); control to increase the operating frequency of the operation unit to speed up the conversion of signals into action; and, similar to the second embodiment, control to have the operation control unit 1101 use the second control model instead of the first control model, thereby improving the performance of the main system 100 and enabling subsequent processing.
[0175] In S62, the generation unit 2104 in Figure 19 generates a second signal, which is action information that suppresses the behavioral action using the modified resources. Specifically, the first signal is input to the determination unit 2101, which performs deviation determination processing based on the behavioral norms of the autonomous robot related to the first signal. If the deviation determination result is a positive determination, the generation unit 2104 generates a second signal that suppresses the behavioral action using the modified resources.
[0176] If the deviation detection process by the first determination model is positive, the resource control unit 2105 can improve responsiveness by changing the resources to high-performance resources. On the other hand, if the deviation detection process is negative, highly efficient resources can be continuously driven for the deviation detection process of the autonomous robot 10, contributing to the energy saving of the autonomous robot 10. Therefore, the resource change process related to the second processing unit by the resource control unit 2105 may improve one or more aspects of energy saving and responsiveness.
[0177] In S63, the motion control unit 1101 in Figure 16 generates a third signal, which is motion information that transitions the autonomous robot to a stationary state using the modified resources.
[0178] <Fourth Embodiment> Next, a fourth embodiment of the present disclosure will be described. Here, the configurations that have already been described in detail in the first to third embodiments will be omitted from the description. Figure 21 shows an overview of the performance improvement of the autonomous robot according to the fourth embodiment in particular. In the fourth embodiment, a positive judgment is used as a trigger to switch the judgment model of the subsystem 200 to improve the performance of the autonomous robot and then execute the subsequent processing.
[0179] In the fourth embodiment, the performance of the subsystem is improved by modifying the judgment model, and the performance of the main system and subsystem is also improved by modifying one or more of the first processing unit, detection unit, and operation unit, the control model, and the second processing unit as described in the first to third embodiments. For this reason, in the fourth embodiment, the system configuration of the autonomous robot will be shown in Figure 18, and the functional block diagram of the main system will be shown in Figure 16. On the other hand, modifications to the judgment model do not necessarily have to be made together with modifications to one or more of the first processing unit, detection unit, operation unit, control model, and second processing unit. Also, as shown in Figure 18, the fourth embodiment describes a case in which a dedicated area 220 in the main memory 120 is used as the main memory unit, but the main memory 120 and the dedicated main memory 220 may be configured to be physically independent.
[0180] The autonomous robot 10 according to this embodiment has a determination model that includes a first model and a second model. The dedicated auxiliary storage device 230 (auxiliary storage unit) stores the first determination model (first model), which is the first determination model, and the second determination model (second model), which is the second determination model. Here, the first determination model has a lower estimated cost or higher processing versatility compared to the second determination model, and is advantageous in terms of energy saving and / or processing versatility of the autonomous robot 10.
[0181] In this embodiment, a large-sized AL / ML model with more parameters than the first decision model is used as the second decision model. In this case, the first decision model can be described as a small-sized AL / ML model with fewer parameters than the second decision model.
[0182] For example, compared to the case where the determination unit 2101 processes deviations using a single determination model with equivalent determination accuracy (for example, the determination models in the first to third embodiments), it is possible to adopt a model with reduced computational load as the continuously operating first determination model, which can contribute to the energy saving of the autonomous robot. Furthermore, depending on the structure of the first determination model and / or the second determination model, one or more of the following may be improved: energy saving, effectiveness, and responsiveness.
[0183] The first decision model and the second decision model are pre-trained models obtained by performing predetermined machine learning processes, and in this embodiment, they correspond to value functions and / or policies trained by model-based reinforcement learning processes. For example, the decision model may be a state machine that operates by transitioning between the first and second decision models. In this case, the resource control unit 2105 corresponds to a decision monitoring module that monitors the state transition from the first decision model to the second decision model based on the first deviation determination result from the first decision model. The decision monitoring module is a rule-based or learning-based function that simply (efficiently) determines whether or not the behavioral norm is satisfied based on an input including a first signal and the first decision model, and outputs a first deviation determination result. If the first deviation determination result is positive, it calls a second decision model that makes a highly (highly efficient) determination as to whether or not the behavioral norm is satisfied.
[0184] For example, the decision model may correspond to a value function and / or policy that has been learned through hierarchical reinforcement learning, with a first decision model that simply determines whether or not the behavioral norm is satisfied as the upper-level policy, and a second decision model that is called by the first decision model to make a more sophisticated determination of whether or not the behavioral norm is satisfied as the lower-level policy. In this case, the second processing unit functions as a decision unit 2101 and a resource control unit 2105 by executing the first decision model.
[0185] Figures 22 and 23 show flowcharts of the processing procedures for the actions and suppression of the autonomous robot according to this embodiment. In explaining these processing procedures, we will refer to Figures 16 and 24, which show the functional configuration of the main system 100 and subsystem 200 (enclave 200), in conjunction with Figures 22 and 23 as appropriate.
[0186] Figure 24 shows a functional block diagram of the subsystem (enclave) according to this embodiment. As shown in the figure, in the subsystem 200 of the autonomous robot 10, at least one of the one or more second processing units 210 (and dedicated area 220, dedicated auxiliary storage device 230, etc.) functions as a determination unit 2101 for performing deviation determination processing based on the first determination model and the second determination model. If the first deviation determination result is a positive determination, the resource control unit 2105 improves the performance of the subsystem by changing at least some of the subsystem's resources by having the determination unit 2101 use the second determination model instead of the first determination model.
[0187] In step S71 of Figure 22, the motion control unit 1101 of the main system 100 in Figure 16 inputs environmental data to the first control model and generates a first signal related to the behavioral movements of the autonomous robot 10.
[0188] Then, in step S72 of Figure 22, the first signal is input to the first judgment model relating to the behavioral norms of the autonomous robot 10. The first signal is input to the first judgment model stored in the judgment unit 2101. In this embodiment, along with the first signal, image data relating to the external environment (external environment data) acquired by the detection unit 140 of the main system 100 is input to the first judgment model based on predetermined conditions.
[0189] In S73, it is determined whether the behavioral actions of the autonomous robot 10 violate the code of conduct. Specifically, the determination unit 2101 in Figure 24 determines whether the behavioral actions of the autonomous robot 10 constitute ethical behavior (whether or not to suppress the behavioral actions) based on the first signal and environmental data (environmental features) input to the first determination model (first deviation determination process).
[0190] If the determination unit 2101 determines, based on the first determination model, that the behavioral action of the autonomous robot 10 does not violate the code of conduct (in the case of No in S73), that is, if the behavioral action of the autonomous robot 10 (first signal) is determined to be ethical behavior in the first deviation determination result (if deviation from the code of conduct is denied), then the behavioral action of the autonomous robot 10 based on the first signal is maintained as shown in Figure 22 (S76). The autonomous robot 10 performs the behavioral action as usual based on the first signal generated by the motion control unit 1101.
[0191] On the other hand, if the behavioral actions of the autonomous robot 10 are determined to be contrary to the code of conduct (i.e., if a deviation from the code of conduct is determined to be positive), that is, if the behavioral actions of the autonomous robot 10 are determined to be unethical in the first deviation determination result (i.e., if the result is Yes in S73), in S74 the resource control unit 2105 causes the determination unit 2101 to use the second determination model instead of the first determination model. In S75 the determination unit 2101 in Figure 24 performs the second deviation determination process using the modified resources. This improves one or more of the energy efficiency of the autonomous robot, as well as the effectiveness and responsiveness of the action task that transitions the autonomous robot to a stationary state.
[0192] If the determination unit 2101 determines, based on the second determination model, that the behavioral action of the autonomous robot 10 does not violate the code of conduct (in the case of No in S75), that is, if the behavioral action of the autonomous robot 10 (first signal) is determined to be ethical behavior and the second deviation determination result is negative, the behavioral action of the autonomous robot 10 based on the first signal is maintained as shown in Figure 22 (S76).
[0193] On the other hand, if the behavior of the autonomous robot 10 is determined to be contrary to the code of conduct based on the second judgment model (in the case of Yes in S75), that is, if the behavior of the autonomous robot 10 is determined to be unethical in the second deviation judgment result and the second deviation judgment result is positive, the behavior of the autonomous robot 10 is suppressed (S77). More specifically, the behavior of the autonomous robot 10 is forcibly stopped.
[0194] In S75 of Figure 22, if it is determined that the autonomous robot 10's behavioral actions violate the code of conduct (Yes), the autonomous robot 10's behavioral actions are suppressed in S77. In S81 of Figure 23, the resource control unit 1103 modifies the resources to improve the performance of the main system 100 and the second processing unit of the subsystem 200.
[0195] In S82, the generation unit 2104 in Figure 24 generates a second signal, which is action information that suppresses active movement. Then, in S83, the action control unit 1101 in Figure 16 generates a third signal, which is action information that transitions the autonomous robot to a stationary state using the modified resources.
[0196] As described above, according to this disclosure, by using a determination model capable of performing deviation detection processing from a predetermined code of conduct, it is possible to realize an autonomous robot that can coexist with humans in light of the singularity, and a semiconductor device for realizing this.
[0197] Furthermore, while this disclosure primarily describes autonomous robots in detail, it goes without saying that semiconductor devices for autonomous robots, control methods for autonomous robots, and control programs for autonomous robots can achieve the same effects as described in this disclosure.
[0198] The disclosures herein include the following: an autonomous robot, a method for controlling the autonomous robot, and a control program for the autonomous robot.
[0199] (Item 1) A main system comprising a processing unit having a first processing unit and a second processing unit, and an operation unit that converts signals from the processing unit into behavioral actions of an autonomous robot, the main system comprising the first processing unit and the operation unit, and a subsystem comprising the second processing unit and storing a judgment model, the subsystem being isolated as an independent control system from the main system, and at least a part of the subsystem including the second processing unit being isolated as an independent circuit system, the signals comprising a first signal, a second signal and a third signal, the first processing unit generating the first signal based on a control model to perform control including operation and stopping of the operation unit, the second processing unit inputting the first signal to the judgment model and performing deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generating the second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit generating the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination. The processing unit improves the performance of the autonomous robot by modifying at least some resources if the deviation determination result is a positive determination.
[0200] (Item 2) The autonomous robot according to Item 1, wherein the first processing unit, when the deviation determination result is a positive determination, modifies at least some of the resources of the main system to improve the performance of the main system and generates the third signal using the modified resources.
[0201] (Item 3) The autonomous robot according to Item 1 or 2, wherein the second processing unit, when the deviation determination result is a positive determination, modifies at least some of the resources of the subsystem to improve the performance of the subsystem and generates the second signal using the modified resources.
[0202] (Item 4) The autonomous robot according to any one of Items 1 to 3, wherein the processing unit includes a plurality of processing processors with different maximum operating frequencies, and when the deviation determination result is a positive determination, it executes control to use the processing processor with a higher maximum operating frequency instead of the processing processor with a lower maximum operating frequency, thereby improving the performance.
[0203] (Item 5) The autonomous robot according to any one of Items 1 to 3, wherein the processing unit includes a processing processor having a plurality of processing processor cores with different maximum operating frequencies, and when the deviation determination result is a positive determination, the processing unit performs control that uses one or more processing processor cores with a higher maximum operating frequency in place of, or in addition to, one or more of the processing processor cores with a lower maximum operating frequency, thereby improving the performance.
[0204] (Item 6) The autonomous robot according to any one of Items 1 to 5, wherein the processing unit includes a processing processor, and when the deviation determination result is a positive determination, the processing unit performs control to increase the operating frequency of the processing processor to improve the performance.
[0205] (Item 7) The autonomous robot according to any one of Items 1 to 6, wherein the first processing unit, when the deviation determination result is a positive determination, executes control to improve the responsiveness of the operating unit and speeds up the conversion to the behavioral action based on the third signal.
[0206] (Item 8) The autonomous robot according to any one of Items 1 to 7, wherein the autonomous robot is equipped with a detection unit for detecting environmental information, and the first processing unit executes one or more of the following when the deviation determination result is a positive determination: control to increase the operating frequency of the detection unit, control to switch the detection unit for detecting the environmental information to a high-performance unit, or control to change the analysis logic of the environmental information.
[0207] (Item 9) The autonomous robot according to any one of Items 1 to 8, wherein the control model comprises a first model and a second model, the first processing unit generates the first signal based on the first model, and the first processing unit generates the third signal based on the second model if the deviation determination result is a positive determination, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the operation task of transitioning the autonomous robot to a stationary state.
[0208] (Item 10) The autonomous robot according to any one of Items 1 to 9, wherein the determination model comprises a first model and a second model, the second processing unit inputs the first signal to the first model of the determination model and executes the deviation determination process, the second processing unit, if the first deviation determination result is a positive determination, executes the second deviation determination process based on the second model of the determination model instead of the first model of the determination model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation determination process from the behavioral norm, the processing unit generates the second signal if the second deviation determination result is a positive determination, and the first processing unit generates the third signal if the second deviation determination result is a positive determination.
[0209] (Item 11) A main system comprising a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of an autonomous robot, the main system comprising the first processing unit and the operating unit, and a subsystem comprising the second processing unit and storing a judgment model, the subsystem being isolated as an independent control system from the main system, and at least a part of the second processing unit being isolated as an independent circuit system, the signals comprising a first signal, a second signal and a third signal, the judgment model comprising a first model and a second model, the first processing unit generating the first signal based on the control model to perform control including operation and stopping of the operating unit, the second processing unit inputting the first signal to the first model and executing a first deviation judgment process from the behavioral norms of the autonomous robot relating to the first signal, the second processing unit, if the first deviation judgment result is a positive judgment, executes a second deviation judgment process based on the second model instead of the first model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation judgment process from the behavioral norms, the processing unit is second An autonomous robot, wherein if the deviation judgment result is a positive judgment, the first processing unit generates the second signal that suppresses the behavioral action, and if the second deviation judgment result is a positive judgment, the first processing unit generates the third signal based on the control model that performs the control to transition the autonomous robot to a stationary state using the first processing unit.
[0210] (Item 12) The second model is an autonomous robot described in any of items 9 to 11, which has a higher estimated cost than the first model.
[0211] (Item 13) The second model is an autonomous robot described in any of items 9 to 11, which has a higher estimated speed than the first model.
[0212] (Item 14) A method for controlling an autonomous robot, wherein the autonomous robot comprises a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a judgment model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals comprises a first signal, a second signal and a third signal, the first processing unit generates the first signal based on a control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the judgment model and performs deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates the second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination A method for controlling an autonomous robot, wherein the processing unit, when the deviation determination result is a positive determination, modifies at least some resources to improve the performance of the autonomous robot.
[0213] (Item 15) A method for controlling an autonomous robot, wherein the autonomous robot comprises a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, wherein the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal, and a third signal, the determination model has a first model and a second model, the first processing unit generates the first signal based on the control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the first model and performs a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal, and if the first deviation determination result is a positive determination, the second processing unit performs a second deviation determination process based on the second model instead of the first model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation determination process from the behavioral norms. A method for controlling an autonomous robot, wherein the processing unit generates a second signal to suppress the behavioral action if the first deviation determination result or the second deviation determination result is positive, and the first processing unit generates a third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the second deviation determination result is positive.
[0214] (Item 16) A control program for an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a judgment model, wherein the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the control program causes the first processing unit to generate the first signal based on the control model to perform control including operation and stopping of the operating unit, causes the second processing unit to input the first signal to the judgment model and perform deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, causes the processing unit to generate the second signal that suppresses the behavioral action if the deviation determination result is a positive determination, causes the first processing unit to generate the third signal based on the control model to perform the control that transitions the autonomous robot to a stationary state if the deviation determination result is a positive determination. The processing unit includes an autonomous robot control program that, if the deviation determination result is positive, modifies at least some resources to improve the performance of the autonomous robot.
[0215] (Item 17) A control program for an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, the subsystem is isolated as an independent control system from the main system, and at least a part of it including the second processing unit is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the determination model has a first model and a second model, the control program causes the first processing unit to generate the first signal based on the control model, which includes controlling the operation and stopping of the operating unit, and causes the second processing unit to input the first signal to the first model and execute a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal. A control program for an autonomous robot, wherein the second processing unit is instructed to perform a second deviation judgment process based on the second model instead of the first model if the first deviation judgment result is a positive judgment, thereby improving one or more of the effectiveness, energy efficiency, and responsiveness of the deviation judgment process from the behavioral norm; the processing unit is instructed to generate the second signal that suppresses the behavioral action if the first or second deviation judgment result is a positive judgment; and the first processing unit is instructed to generate the third signal that performs the control to transition the autonomous robot to a stationary state based on the control model if the second deviation judgment result is a positive judgment.
[0216] (Item 18) An autonomous robot having an operating unit and a processing unit, wherein the operating unit converts signals from the processing unit into behavioral actions of the autonomous robot, the processing unit has a first processing unit and a second processing unit, and the signals have a first signal, a second signal and a third signal, the first processing unit uses a first processing processor to generate the first signal based on a control model for controlling the operating unit, including the operation and stopping of the operating unit, the second processing unit uses a second processing processor to input the first signal to a judgment model and execute a deviation determination process from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates a second signal to suppress the behavioral action if the deviation determination result is a positive determination, the system comprises a main system including the first processing processor and the operating unit, and a subsystem including the second processing processor and storing the judgment model, the first processing unit changes the operating frequency of at least a part of the main system if the deviation determination result is a positive determination, and generates the third signal based on the control model for performing the control to transition the autonomous robot to a stationary state using the first processing processor. An autonomous robot in which the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing processor, is isolated as an independent circuit system.
[0217] (Item 19) The autonomous robot according to Item 18, wherein the first processing unit changes the operating frequency of the first processing processor when the deviation determination result is a positive determination.
[0218] (Item 20) The autonomous robot according to Item 18, wherein the main system includes a plurality of first processing processors with different operating frequencies, and the first processing unit generates the third signal by using the first processing processor with a higher operating frequency instead of the first processing processor with a lower operating frequency when the deviation determination result is a positive determination.
[0219] (Item 21) An autonomous robot having an operating unit and a processing unit, wherein the operating unit converts signals from the processing unit into behavioral actions of the autonomous robot, the processing unit has a first processing unit and a second processing unit, and the signals have a first signal, a second signal and a third signal, the first processing unit uses a first processing processor to generate the first signal based on a first control model for controlling the operating unit, including operation and stopping, the second processing unit uses a second processing processor to input the first signal to a judgment model and execute a deviation determination process from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates a second signal to suppress the behavioral action if the deviation determination result is a positive determination, the system comprises a main system including the first processing processor and the operating unit, and a subsystem including the second processing processor and storing the judgment model, the first processing unit, if the deviation determination result is a positive determination, generates the third signal based on the second control model instead of the first control model, using the first processing processor to perform the control to transition the autonomous robot to a stationary state, An autonomous robot in which the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing processor, is isolated as an independent circuit system.
[0220] (Item 22) The autonomous robot described in Item 21, wherein the first and second control models are trained models, and the second control model has a larger parameter size than the first control model.
[0221] (Item 23) The autonomous robot described in Item 21, wherein the first and second control models are trained models, and the second control model has a higher estimated speed than the first control model.
[0222] (Item 24) An autonomous robot having an operating unit and a processing unit, wherein the operating unit converts signals from the processing unit into behavioral actions of the autonomous robot, the processing unit has a first processing unit and a second processing unit, and the signals have a first signal, a second signal and a third signal, the first processing unit generates a first signal based on a control model that performs control including operation and stopping of the operating unit using a first processing processor, the second processing unit inputs the first signal to a judgment model using a second processing processor and performs deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates a second signal that suppresses the behavioral action if the deviation determination result is a positive determination, the first processing unit generates a third signal based on the control model that performs the control to transition the autonomous robot to a stationary state using the first processing processor, and comprises a main system including the first processing processor and the operating unit, and a subsystem including the second processing processor and storing the judgment model. An autonomous robot in which, if the deviation determination result is a positive determination, the second processing unit changes the operating frequency of at least a part of the subsystem and generates the second signal, the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing processor, is isolated as an independent circuit system.
[0223] (Item 25) The autonomous robot according to Item 24, wherein the second processing unit changes the operating frequency of the second processing processor when the deviation determination result is a positive determination.
[0224] (Item 26) The autonomous robot according to Item 24, wherein the subsystem includes a plurality of second processing processors with different operating frequencies, and the second processing unit generates the second signal by using the second processing processor with a higher operating frequency instead of the second processing processor with a lower operating frequency when the deviation determination result is a positive determination.
[0225] (Item 27) An autonomous robot having an operating unit and a processing unit, wherein the operating unit converts signals from the processing unit into behavioral actions of the autonomous robot, the processing unit has a first processing unit and a second processing unit, and the signals have a first signal, a second signal and a third signal, the first processing unit uses a first processing processor to generate the first signal based on a control model for controlling the operating unit, including operation and stopping, the second processing unit uses a second processing processor to input the first signal to a first judgment model and execute a deviation determination process from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates a second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit uses the first processing processor to generate the third signal based on the control model for performing the control to transition the autonomous robot to a stationary state, and comprises a main system including the first processing processor and the operating unit, and a subsystem including the second processing processor and storing the judgment model. The second processing unit, if the deviation determination result is a positive determination, re-executes the deviation determination process based on the second determination model instead of the first determination model, and the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing processor, is isolated as an independent circuit system, in an autonomous robot.
[0226] (Item 28) The autonomous robot described in Item 27, wherein the first and second decision models are trained models, and the second decision model has a larger parameter size than the first decision model.
[0227] (Item 29) The autonomous robot described in Item 27, wherein the first and second decision models are trained models, and the second decision model has a higher estimation speed than the first decision model.
[0228] The present invention is not limited to the embodiments described above, and various modifications and variations are possible without departing from the spirit and scope of the invention.
[0229] 10 Autonomous robot 20 Human 100 Main system 110 First processing chip (first processing unit) 1101 Motion control unit 1102 Extraction unit 1103 Resource control unit 120 Main memory 130 Auxiliary memory 140 Sensor (detection unit) 150 Actuator (motion unit) 200 Enclave (subsystem) 210 Second processing chip (second processing unit) 2101 Learning unit 2102 Judgment unit 2103 Extraction unit 2104 Generation unit 2105 Resource control unit 220 Dedicated main memory (dedicated area) 230 Dedicated auxiliary memory 240 AI / ML accelerator 250 Encryption accelerator 260 Dedicated boot ROM 270 Memory protection 310 Common bus interface 320 Dedicated bus interface 600 Information processing unit 610 Processing unit 620 Memory Unit 630 Communication Unit
Claims
1. A main system comprising a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of an autonomous robot, and a subsystem comprising the second processing unit and a judgment model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals comprising a first signal, a second signal and a third signal, the first processing unit generates the first signal based on a control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the judgment model and performs deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates the second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination. The processing unit improves the performance of the autonomous robot by modifying at least some resources if the deviation determination result is a positive determination.
2. The autonomous robot according to claim 1, wherein the first processing unit, when the deviation determination result is a positive determination, modifies at least some of the resources of the main system to improve the performance of the main system and generates the third signal using the modified resources.
3. The autonomous robot according to claim 1, wherein, if the deviation determination result is a positive determination, the second processing unit modifies at least some of the resources of the subsystem to improve the performance of the subsystem and generates the second signal using the modified resources.
4. The autonomous robot according to claim 1, wherein the processing unit includes a plurality of processing processors with different maximum operating frequencies, and when the deviation determination result is a positive determination, it performs control to use the processing processor with a higher maximum operating frequency instead of the processing processor with a lower maximum operating frequency, thereby improving the performance.
5. The autonomous robot according to claim 1, wherein the processing unit includes a processing processor having a plurality of processing processor cores with different maximum operating frequencies, and when the deviation determination result is a positive determination, the processing unit performs control that utilizes one or more processing processor cores with a higher maximum operating frequency in place of, or in addition to, one or more of the processing processor cores with a lower maximum operating frequency, thereby improving the performance.
6. The autonomous robot according to claim 1, wherein the processing unit includes a processing processor, and when the deviation determination result is a positive determination, the processing unit performs control to increase the operating frequency of the processing processor to improve the performance.
7. The autonomous robot according to claim 1, wherein the first processing unit, when the deviation determination result is a positive determination, executes control to improve the responsiveness of the operating unit and speeds up the conversion to the behavioral action based on the third signal.
8. The autonomous robot according to claim 1, wherein the autonomous robot is equipped with a detection unit for detecting environmental information, and the first processing unit performs one or more of the following actions when the deviation determination result is a positive determination: control to increase the operating frequency of the detection unit, control to switch the detection unit for detecting the environmental information to a high-performance unit, or control to change the analysis logic of the environmental information.
9. The autonomous robot according to claim 1, wherein the control model comprises a first model and a second model, the first processing unit generates the first signal based on the first model, and the first processing unit generates the third signal based on the second model when the deviation determination result is a positive determination, thereby improving one or more of the effectiveness, energy efficiency, and responsiveness of the operation task of transitioning the autonomous robot to a stationary state.
10. The autonomous robot according to claim 1, wherein the determination model comprises a first model and a second model, the second processing unit inputs the first signal to the first model of the determination model and executes the deviation determination process, the second processing unit, if the first deviation determination result is a positive determination, executes the second deviation determination process based on the second model of the determination model instead of the first model of the determination model, thereby improving one or more of the effectiveness, energy efficiency, and responsiveness of the deviation determination process from the behavioral norm, the processing unit generates the second signal if the second deviation determination result is a positive determination, and the first processing unit generates the third signal if the second deviation determination result is a positive determination.
11. A main system comprising a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of an autonomous robot, and a subsystem comprising the second processing unit and a judgment model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal, and a third signal, the judgment model has a first model and a second model, the first processing unit generates the first signal based on the control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the first model and performs a first deviation judgment process from the behavioral norms of the autonomous robot relating to the first signal, and if the first deviation judgment result is a positive judgment, the second processing unit performs a second deviation judgment process based on the second model instead of the first model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation judgment process from the behavioral norms. An autonomous robot wherein the processing unit generates the second signal to suppress the behavioral action when the second deviation determination result is positive, and the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state using the first processing unit when the second deviation determination result is positive.
12. The autonomous robot according to any one of claims 9 to 11, wherein the second model has a higher estimated cost than the first model.
13. The autonomous robot according to any one of claims 9 to 11, wherein the second model has a higher estimated speed than the first model.
14. A method for controlling an autonomous robot, wherein the autonomous robot comprises a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a judgment model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals comprises a first signal, a second signal and a third signal, the first processing unit generates the first signal based on a control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the judgment model and performs deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, the processing unit generates the second signal to suppress the behavioral action if the deviation determination result is a positive determination, the first processing unit generates the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination A method for controlling an autonomous robot, wherein the processing unit, when the deviation determination result is a positive determination, modifies at least some resources to improve the performance of the autonomous robot.
15. A method for controlling an autonomous robot, wherein the autonomous robot has a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, wherein the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal, and a third signal, the determination model has a first model and a second model, the first processing unit generates the first signal based on the control model to perform control including operation and stopping of the operating unit, the second processing unit inputs the first signal to the first model and performs a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal, and if the first deviation determination result is a positive determination, the second processing unit performs a second deviation determination process based on the second model instead of the first model, thereby improving one or more of the effectiveness, energy saving, and responsiveness of the deviation determination process from the behavioral norms. A method for controlling an autonomous robot, wherein the processing unit generates a second signal to suppress the behavioral action if the first deviation determination result or the second deviation determination result is positive, and the first processing unit generates a third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the second deviation determination result is positive.
16. A control program for an autonomous robot, wherein the autonomous robot comprises a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a judgment model, wherein the subsystem is isolated as an independent control system from the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the control program causes the first processing unit to generate the first signal based on the control model to perform control including operation and stopping of the operating unit, causes the second processing unit to input the first signal to the judgment model and perform deviation determination processing from the behavioral norms of the autonomous robot relating to the first signal, causes the processing unit to generate the second signal to suppress the behavioral action if the deviation determination result is a positive determination, and causes the first processing unit to generate the third signal based on the control model to perform the control to transition the autonomous robot to a stationary state if the deviation determination result is a positive determination. The processing unit includes an autonomous robot control program that, if the deviation determination result is positive, modifies at least some resources to improve the performance of the autonomous robot.
17. A control program for an autonomous robot, wherein the autonomous robot comprises a processing unit having a first processing unit and a second processing unit, and an operating unit that converts signals from the processing unit into behavioral actions of the autonomous robot, and comprises a main system including the first processing unit and the operating unit, and a subsystem including the second processing unit and storing a determination model, wherein the subsystem is isolated as a control system independent of the main system, and at least a part of it, including the second processing unit, is isolated as an independent circuit system, the signals have a first signal, a second signal and a third signal, the determination model has a first model and a second model, the control program causes the first processing unit to generate the first signal based on the control model, which includes controlling the operation and stopping of the operating unit, and causes the second processing unit to input the first signal to the first model and execute a first deviation determination process from the behavioral norms of the autonomous robot relating to the first signal. A control program for an autonomous robot, wherein the second processing unit is instructed to perform a second deviation judgment process based on the second model instead of the first model if the first deviation judgment result is a positive judgment, thereby improving one or more of the effectiveness, energy efficiency, and responsiveness of the deviation judgment process from the behavioral norm; the processing unit is instructed to generate the second signal that suppresses the behavioral action if the first or second deviation judgment result is a positive judgment; and the first processing unit is instructed to generate the third signal that performs the control to transition the autonomous robot to a stationary state based on the control model if the second deviation judgment result is a positive judgment.
