AI-SUPPORTED CENTRALIZED POWER DISTRIBUTION AND MANAGEMENT UNIT FOR ROBOTIC PLATFORMS AND RELATED METHODOLOGY.

Abstract

The invention is a power distribution method and unit designed for autonomous mobile robots and humanoid platforms, integrating hardware protection, telemetry, embedded artificial intelligence, and standard robot operating system (ROS2) communication. Within the scope of the invention, in case of short circuit or overcurrent, ultra-fast cutting at the hardware level isolates only the faulty channel, ensuring uninterrupted robot operation (phased reduction of function). High-frequency collected current and voltage data are converted into standard robot operating system messages, enabling dynamic power management and predicting failures before they occur using machine learning models. Furthermore, the system manages the autonomous charging process by monitoring only thermal and electrical parameters, without the need for an external communication protocol between the robot and the charging station, and prevents arc / fire risk by cutting off the voltage at the charging contacts.

Description

1 TARIFF AI-POWERED CENTRALIZED POWER FOR ROBOTIC PLATFORMS DISTRIBUTION AND MANAGEMENT UNIT AND RELATED METHODS TECHNICAL FIELD The invention relates to industrial robotics and automation industries, as well as embedded software and intelligent control. It is related to systems and robot operating systems (ROS2). The invention is particularly relevant to Autonomous Mobile Robots (AMR / AGV) and humanoid robots. LIDAR, NPU-based AI processors, cameras, and high-performance systems are included in these platforms. Current, voltage, and temperature regulation of critical peripherals such as torque actuators providing integrated, autonomous charging processes via an external communication protocol. Managing based on current and temperature monitoring without needing to be monitored, in microseconds 15 hardware short-circuit protection, phased reduction of functionality, and artificial intelligence at this level. a centralized intelligent power distribution system with intelligence-based predictive maintenance capabilities and It relates to the management unit and the associated method. STATE OF THE ART 20 In current technology, autonomous mobile systems are used in industrial logistics and production lines. Robots and next-generation humanoid platforms incorporate numerous sensitive sensors, high-performance artificial intelligence units and high-torque actuators It contains 25 components used to nourish and manage these complex sub-components. Traditional distributed and passive power architectures have serious technical problems and shortcomings. It brings with it: 1. Cabling and Hardware Complexity: In current distributed architectures, each Individual lines are drawn from the central battery to the sensor, processor, and motor; 30 A dense bundle of wiring between fuses, relays, and DC-DC converters. It consists of this structure, which occupies a large volume within the robot body and is movable. friction in robotic joints, insulation wear, assembly / maintenance 2 due to difficulties and long lines, energy losses caused by voltage drop and overheating. This leads to losses. 2. Insufficiency of Passive Protection Elements and High MTTR: ​​Traditional Protection in buildings is passively provided by melting fuses or terminal-magnetic circuit breakers. This is provided. The response times of these elements are on the order of milliseconds. 5 Furthermore, the fuse only cuts off the system after the malfunction occurs and the fault is detected. It does not generate anomaly warnings beforehand. The most critical disadvantage is the error. The lack of isolation; of a critical peripheral device (e.g., security) When the fuse of the LIDAR blows, all the robots become inoperable, causing unplanned production. This causes it to stop. 10 3. Lack of Data Collection and Dynamic Power Management: The most advanced power systems currently available... The modules are either unable to measure current / voltage at all, or only perform low measurements. Frequencies (~10 Hz – 100 Hz) offer simple threshold controls. Microseconds Because telemetry data could not be collected at that level, according to the robot's mission scenario... Dynamic “Role-Based Power Management” strategies cannot be implemented. 15 4. Dependencies and Fire Risks During Charging Processes: Charging autonomous robots For this to be possible, a brand-specific, complex connection must be made between the charging station and the robot. There are protocol dependencies. Also, charging continues even if the robot leaves the station. The presence of continuously active voltage in the contacts can cause arcs and fires in industrial areas. It triggers risks. 20 Ultimately, the problems mentioned above, which cannot be solved with current technology, are related. This has made it necessary to make an innovation in the technical field. A BRIEF DESCRIPTION OF THE INVENTION The present invention eliminates the aforementioned disadvantages and the related technical an AI-powered robotic power distribution unit to bring new advantages to the field and it relates to the method associated with it. The main purpose of the invention is to improve power distribution, protection, measurement, and advanced robotics. By offering integrated communication in a single compact chassis, it eliminates internal robot cabling and to reduce hardware size by at least 30%, prevent voltage drops, and The goal is to minimize the risks of physical wear and tear. 3 Another objective of the invention is intelligent power switches with integrated diagnostic functions. Through this, in short circuit and overcurrent situations, an ultra-high response time of <5 microseconds. rapidly intervening at the hardware level and "Incremental Function Degradation" This method allows for uninterrupted robot operation by isolating only the faulty channel. to provide. 5 Another purpose of the invention is to collect current and voltage from the output channels at a frequency of 1 kHz. and by extracting features from temperature data, it runs on the robot's main processor. quantized machine learning models detect failures before they even occur. First, the goal is to detect with a success rate of 70% or higher. 10 Another objective of the invention is to transmit precise time-stamped power data to a dedicated ROS2 node. broadcasting via a high-level computer with a delay of <10 ms, "Role It enables "Power Management Based" and secure OTA remote updates. The best way to utilize the advantages of the existing invention, together with its structure and additional elements. For understanding, it should be evaluated together with the figures explained below. is necessary. BRIEF DESCRIPTION OF THE FIGURES 20 Figure 1, Figure 2, REFERENCE NUMBERS 25 Main Processing Unit Smart Power Switch Battery Monitoring and Interface Unit 33 Robot Main Processors 30 40 ROS2 Communication and Software Nodes 45 Charging Contacts 50 Web-Based Monitoring and Remote Management Dashboard 89 Charging Stations 4 DETAILED DESCRIPTION OF THE INVENTION This detailed explanation aims to provide a better understanding of the innovation in the invention. This is explained with examples that will not create any limiting effect. 5 The invention is a hardware-based system designed for autonomous mobile robots and mobile humanoid platforms. protection, real-time telemetry, embedded AI-powered anomaly detection, and robotics. Operating systems (ROS2) communication infrastructure on a single hardware architecture. an AI-powered robotic power distribution method that offers an integrated solution, and these 10 The central power management and distribution unit (PMU) is the one that executes the method. The subject of the invention is the hardware and software of the central power management and distribution unit. The elements are described below: Main Processing Unit (10): Multiple ADCs compliant with industrial automation standards. (Analog-to-Digital Converter) channels, advanced timers, and real-world applications. It is an STM32-based microcontroller with real-time processing capability. The entire system... It manages cyclical data management, digital filtering, and embedded inferences. Unlike known solutions in the technical field, traditional data reading of 10 Hz to 100 Hz is 20 by increasing the sampling rate to a high sampling frequency level of 1 kHz, the data is stored in a cyclical buffer. It accumulates continuously in memory (circular buffer). Hardware-based at the embedded level. mathematical derivation of instantaneous rate of change from current signals It performs dI / dt slope analysis, which provides the following data: Quantized Machine Learning algorithms in the robot main processor (33) 25 by operating it, anomalies and malfunctions such as increased mechanical friction are not yet physically apparent. It predicts before it happens. Smart Power Switches (20): Replacing traditional passive fuses and relays, It has integrated hardware protection against short circuits, overcurrents, overtemperatures, and undervoltages. 30 They are semiconductor switching elements that incorporate protection mechanisms. Each one It controls an independent output channel (at least 6 channels) and has an internal current mirror. It provides instantaneous analog current feedback through its circuits. Passive fusible fuse and Unlike thermal-magnetic circuit breakers, in the event of a short circuit or overcurrent, the main processor without waiting for the software algorithm cycle or intervention of the unit (10), directly at the hardware level at an ultra-high speed of <5 microseconds It is performing a cut. This affects the entire system and sensitive sensors (LIDAR, NPU cards). without causing damage, by simply shutting down and isolating the faulty channel within milliseconds. The "Graceful Degradation" function is physically 5 It is carrying out. Battery Monitoring and Interface Unit (30): Robot main battery and battery management voltage, current, state of charge (SoC), and connector temperature data of the building materials system (BMS). It is the hardware and software interface layer that reads and monitors charging safety. Charging 10 an external wired / wireless hardware or software between the station (89) and the robot charging contacts (45) without the need for communication infrastructure / protocol via connector heat monitoring and amperage monitoring (charging current) for full charge. It performs the autonomous charging and disconnection process. The robot leaves the charging area. By autonomously cutting off the voltage at the charging terminals the moment it shows a tendency to disconnect or breaks down, 15 Technically, the risks of arc (spark) and fire in industrial facilities are completely eliminated. It prevents. ROS2 Communication and Software Node (40): Main Processing Unit (10) and the robot's top physical data line (USB, 20 between the level central computer / robot main processor (33) Modbus or CANbus) manages and standardizes low-level hardware telemetry data. software that converts robot operating system (sensor_msgs, std_msgs) message types. It is the end-to-end data layer from the STM32 hardware to the ROS2 topic. By keeping the transmission delay at a critical level of <10 ms, the robot's mission scenarios and dynamic "Role-Based Power Distribution" according to the current operating mode (e.g., standby 25 software that enables disabling unnecessary mapping sensors in operating mode It is the interface. Web-based Monitoring and Remote Management Dashboard (50): Connected to the ROS2 network, Power and anomaly detection by field operators or central Fleet Management Software (FMS) 30 to monitor data with real-time graphs, configure it, and remotely manage service calls. It is an interface platform that allows intervention. Threat modeling. Designed according to standards, dual-zone encryption and roll-back protection. 6 Integrated dual-bank secure OTA remote software update support. It offers. The invention concerns an artificial intelligence-powered robotic power distribution and management method, as follows: This is accomplished by executing the steps sequentially or simultaneously: 5 1. Data Collection, Filtering, and Cyclical Storage Step: The main processing unit (10) receives the current and voltage signals from the smart power switches (20). It reads continuously at a high sampling frequency of 1 kHz. This raw analog reading is then read. The data uses embedded moving average or Kalman 10 for noise suppression. passed through filters into a cyclic buffer within the main processing unit (10) It is stored in memory (circular buffer) in real time. 2. Hardware-Based Ultra-Fast Protection and Cutting Step: In case of a short circuit or sudden overcurrent in the output channels; smart power switches (20), software algorithm cycle of the main processor unit (10) or 15 without waiting for intervention, directly at the hardware level <5 microseconds It protects the system by cutting off power to the faulty channel. 3. Low-Level Telemetry Transfer and ROS2 Conversion Step: Precise time-stamped current, voltage, and temperature are stored in a cyclic buffer memory. Telemetry data is transmitted via a ROS2 communication and software node (40) using standard 20 The robot operating system converts the data into message formats. The converted data is then processed end-to-end. It is transmitted to the robot main processor (33) with a transmission delay as low as <10 ms. Step 4: Autonomous Charging Management Without Communication (Smart Disconnect): When the robot approaches the charging station (89); the battery monitoring and interface unit (30) charges external hardware or software communication via contacts (45) 25 connector heat and charging current (amperes) only, without needing a protocol It monitors changes such as battery charging or the robot leaving the station. In case of a current drop as a result, the charging line is activated via smart power switches (20). It is switched off autonomously and prevents a constant active voltage from remaining on the charging contacts (45). The risk of arcing (sparks) is prevented. 30 5. AI-Based Predictive Maintenance and Anomaly Detection Step: 7 Variance and peaks were obtained from current data at a frequency of 1 kHz from a cyclic buffer memory. Statistical parameters such as crest factor, kurtosis, and rate of change in time (dI / dt) Features are extracted. These extracted features are used by the robot main processor (33) quantized and truncated machine learning models that are being worked on by feeding, mechanical friction increases or electrical weaknesses, the failure is not yet 5 It is predicted with a success rate of 70% or higher before it happens. Step 6: Graceful Degradation and Isolation Step: When the hardware protection step is triggered, the entire system is not shut down; only the hardware is turned on. The faulty channel is isolated within milliseconds. The robot's walking motors and navigation system... 10 This enables the process of graceful degradation. In case of transient current fluctuations, the main processing unit (10) uses a predefined Based on the number of attempts, it autonomously reactivates the relevant channel (smart retry). 7. Heat-Compensated High-Precision Calibration Step: Temperature coefficient of smart power switches (20) 𝑅𝐷𝑆(𝑜𝑛), ADC quantization error and 15 Reference voltage deviations are mathematically modeled. These fault sources... The effect on total uncertainty is determined using the Root Sum of Squares (RSS) method. By analyzing, a temperature range for industrial operating conditions of -20°C to +70°C was determined. A compensation polynomial is applied, and the current measurement accuracy is + / - 1%. It is calibrated within the range of 20. Step 8: Secure Remote Update (OTA): New encrypted data coming through the network via the web-based interface (50) or ROS2 layer. By verifying the software packages, the main processing unit's (10) dual-bank memory It is written into the architecture. In case of interruptions or errors that may occur during loading, the system It automatically reverts to the old stable version (roll-back protection). 25 The subject of the invention is an intelligent power distribution and management unit; motion mechanisms (motors) drivers), sensors (LIDAR, cameras) and high-end processors (NPU-based Edge It can be combined with other robotics techniques such as AI boards. This structural integration... As a result, telemetry at the microsecond level can detect malfunctions even before they become apparent. 30 capable of predicting through analysis, and completely halting the task if a subcomponent fails. instead of continuing to operate in safe mode, gradual function reduction (graceful 8 integrated autonomous mobile robot (AMR / AGV) with degradation mechanism and Humanoid platforms are being obtained. The aforementioned integrated robotics... Thanks to their uninterrupted operation capabilities, these systems reduce downtime in industry. They are positioned as end products with high commercialization potential, minimizing the number of products they contain. The methodology and architecture developed within the scope of the invention enable autonomous charging stations and smart power. It can also be implemented in the form of cutting integration kits. Located within the unit... The area features "Smart Disconnect" and battery monitoring algorithms, among the best on the market. together with existing autonomous charging stations (89) and battery management systems (BMS) When combined as a hardware or software kit; robot and charging station (89) 10 among them are expensive, dependent on a specific brand, and require complex communication protocols. Autonomous charging management infrastructures that do not detect noise are being established. These integration kits, by autonomously cutting off the voltage the moment the robot disconnects from the charging contacts (45) technically eliminates all fire and arc risks in industrial facilities They are offered as commercial packages. 15 The ROS2 communication and software node (40) provided by the invention Microsecond-level instantaneous current and voltage telemetry data, fleet management also as a predictive maintenance and energy optimization module for software (FMS) It can be integrated. Centralized fleet management in cloud or on-premises server architecture 20 these are included in their software either as an add-on or as a standalone software layer. The module artificially detects motor wear and sensor malfunctions in dozens of robots in the field. Detecting malfunctions before they even occur using an intelligence-based anomaly detection method. The system also implements "Role-Based Power Management" strategies through this module. 25 that reduce the total energy costs of factories and increase operational efficiency by operating them The employer views a commercial SaaS (Software as a Service) product.

Claims

9 REQUESTS 1. AI-powered systems developed for autonomous mobile robots and humanoid platforms. It is a robotic power distribution center unit whose feature is;  Current and 5 of external peripherals via multiple analog-to-digital converter channels digital filtering and collecting voltage data in a cyclical buffer memory. a main processing unit (10) that performs embedded data extraction operations,  Independent of the control instructions from the main processing unit (10) operating as; short circuit, overcurrent, overtemperature or low voltage 10 that autonomously isolates the relevant line by performing a hardware-level interrupt under these conditions. and provides instantaneous analog current feedback via internal current mirror circuits. smart power switches (20),  Monitoring the voltage, current, and charge of the robot's main battery and battery management system. Reading status data; external communication between the charger and the robot. Heat and charging current on the charging contacts (45) without needing a protocol 15 Autonomous charging cut-off based on changes and charging upon disconnection A battery monitoring and interface unit that prevents arc risk by cutting off the voltage at its terminals. (30),  Main processing unit (10) and robot's high-level computer / robot main processor (33) low-level hardware telemetry that enables physical data transmission between them 20 by converting the data into standard robot operating system interface message types, the role Enables the dynamic execution of power distribution commands based on power distribution. a ROS2 communication and software node (40) It includes.

2. Power distribution processes of autonomous mobile robots and humanoid robotic platforms. an AI-powered robotic power distribution method developed for management and its characteristic is;  A semiconductor that controls the power output lines via a main processing unit (10). 1 kHz sampling of current and voltage data from smart power switches (20) 30 real-time reading from a cyclical buffer memory with a constant frequency. storage,  In the event of a short circuit or sudden overcurrent on any output line, software of the main processor unit (10) of the smart power switches (20) without waiting for intervention, directly at the hardware level <5 microseconds interrupting the relevant channel during that time,  Telemetry data stored in the transformation buffer is processed by a ROS2 Standard robot operating system via communication and software node (40) converted into message types and transmitted end-to-end with a transmission delay of <10 ms to the robot master 5 transfer to the processor (33),  Through a battery monitoring and interface unit (30), the robot can be connected to the charging station (89) no external hardware or software communication line is needed between them Without being heard, the connector heat and charging current on the charging contacts (45) Managing the autonomous charging process by monitoring changes: 10 steps It includes.

3. A robotic power distribution method that complies with claim 2 and features a cyclical buffer. Variance and crest factor are calculated from current data at a frequency of 1 kHz retrieved from memory. Extraction of features such as factor), kurtosis and instantaneous rate of change (dI / dt) 15 and these attributes are quantized and run on the robot main processor (33) By feeding this information into a machine learning model, mechanical and electrical anomalies and malfunctions are identified. It involves the step of predicting it before it happens.

4. A robotic power distribution method that complies with Claim 2, and whose characteristic is; 20 at the hardware level. Following the closure of the relevant channel; only the faulty channel is isolated and the robot is then activated. a system that ensures the uninterrupted operation of navigation and safety components graceful degradation step and temporary autonomous operation according to a programmed number of attempts on the relevant line during fluctuations. a smart retry step that enables it to be reactivated as 25 It includes.

5. A robotic power distribution method that complies with Claim 2, and its feature is autonomous charging. when the battery is full during the process or from the robot charging station (89) When disconnected, the charging line is autonomously cut off from the smart power switches (20) and 30 Thus, arcing is prevented by preventing a constant active voltage from remaining in the charging contacts (45). It includes the step of preventing the risk. 11 6. A robotic power distribution method that complies with Claim 2, and its feature is intelligent power. Sum of Squares of temperature coefficient and quantization errors of the keys (20) Analysis using the (RSS) method and a temperature range of -20°C to +70°C. Current measurement accuracy is + / - 1% through compensation polynomial. This includes the step of calibrating it with high precision. 5 7. A robotic power distribution method compliant with Claim 2, characterized by its incoming encrypted software. by verifying the packets, the main processing unit (10) dual-bank memory its architecture is written to and in case of loading interruptions, it autonomously reverts to the old stable version. Secure remote control 10 with roll-back protection that allows for reversal. It includes the over-the-air (OTA) update step.

Images