Simulated Kalman filter circuit based on molybdenum disulfide memristor transistor and application thereof

Abstract

The invention discloses a simulated Kalman filter circuit based on a molybdenum disulfide (MoS2) memristive transistor. Efficient fusion of sensor data is realized in the field of automatic driving. According to the circuit, the Kalman filtering gain can be accurately adjusted by adjusting the conductance value of the MoS2 memristor transistor, so that the circuit is suitable for various automatic driving scenes. The simulated Kalman filter circuit can effectively fuse a laser radar and a millimeter wave radar, can accurately estimate the coordinates of a target vehicle, and can significantly reduce power consumption while realizing a data processing speed equivalent to that of a digital filter.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to an analog Kalman filter circuit based on a molybdenum disulfide (MoS2) memristor transistor, which is particularly suitable for positioning a target vehicle. Background Art

[0002] The digital Kalman filter is a widely used technology in autonomous driving systems. It controls the confidence in measurement data by adjusting the K value (Kalman gain), affecting the accuracy and stability of the final estimate. The main features of this technology include:

[0003] (1) Implementation method: The Kalman filter algorithm is usually implemented in a digital processor.

[0004] (2) Signal conversion requirements: Analog-to-digital converters (ADCs) are needed to convert the analog signals from the sensors into digital signals.

[0005] (3) Time form: The algorithm is expressed in discrete time form.

[0006] (4) Computational complexity: Multiple algorithm cycles need to be executed continuously to reach a steady state.

[0007] (5) Resource consumption: Complex computing requirements lead to a large amount of computing resource consumption, which in turn causes power consumption problems and response time delays.

[0008] Accordingly, the digital Kalman filter has the following technical problems:

[0009] (1) Power consumption: Digital Kalman filters require a lot of computing resources, which leads to significant energy consumption in a continuously running autonomous driving environment.

[0010] (2) Response delay: Due to the need to continuously execute the algorithm to maintain steady state and the conversion of analog signals to digital signals, the digital Kalman filter exhibits response delay in fast response environments.

[0011] (3) Complexity of algorithm implementation: Digital Kalman filter requires complex digital calculations and frequent state updates, which increases the complexity of implementation and the instability of the system.

[0012] The primary limitation of digital Kalman filters is their reliance on digital processors and ADCs. This reliance forces the algorithm to execute in the digital domain, increasing conversion and processing time, thus impacting the overall efficiency and responsiveness of the system. Furthermore, the complexity of the digital computations and the need for constant state updates further increase the system's energy consumption.

[0013] In summary, while digital Kalman filters perform well in terms of accuracy and stability, they have significant limitations in terms of power consumption, response time, and algorithmic complexity. These issues are particularly prominent in autonomous driving applications that require real-time data processing and rapid response, limiting the effectiveness of digital Kalman filters in highly dynamic environments. Summary of the Invention

[0014] The present invention aims to address the problems of digital Kalman filters in autonomous driving systems, mainly including high power consumption, response delay and algorithm complexity. These limitations are caused by the reliance on digital processors and the need for analog-to-digital conversion, especially in autonomous driving applications that require real-time data processing and fast response. The present invention provides an analog Kalman filter circuit based on MoS2 memristor transistors, which is designed to accelerate sensor fusion to achieve precise positioning in autonomous driving applications. The non-volatile storage characteristics of memristor transistors allow the storage of fixed Kalman gains, thereby eliminating data convergence and accelerating processing speed. By modulating multiple conduction states through the gate terminal, multiple Kalman filter gains can be adjusted to quickly adapt to different autonomous driving scenarios.

[0015] To achieve the above objectives, the technical solutions of the present invention are as follows:

[0016] An analog Kalman filter circuit based on MoS2 memristor transistors, by integrating LiDAR and millimeter-wave radar, can accurately estimate the position coordinates of target vehicles and successfully solve the practical problem of unsignaled intersections. Specifically, it includes:

[0017] The differential ratio operation module includes at least one operational amplifier and multiple resistors, which is used to compare the collected laser radar signal with the gain adjustment signal W (t) By adjusting the resistance value of the resistor, an amplifier circuit with an adjustable gain factor is created. This gain factor amplifies the input signal to generate a lidar differential amplification signal U (t) ;

[0018] The integral operation module includes at least one operational amplifier and a larger resistor and capacitor connected in parallel, which is used to phase shift the differential amplified signal of the laser radar and update the state estimation value x (t) ;

[0019] The subtraction operation module includes at least one operational amplifier and multiple resistors, which is used to perform differential operation between the collected millimeter-wave radar signal and the output of the analog Kalman filter circuit, and output the millimeter-wave radar state deviation signal V (t) ;

[0020] The integral operation module based on the memristor transistor includes at least an operational amplifier, a memristor transistor, and a large resistor and capacitor connected in parallel. By adjusting the gate voltage of the memristor transistor, the conductance value of the memristor transistor is changed, thereby adjusting the gain of the Kalman filter in different driving scenarios and adjusting the state deviation signal V according to the millimeter wave radar. (t) Output gain adjustment signal W (t) To the differential proportional operation module.

[0021] Each of the lidar signals and millimeter-wave radar signals is fused via a Kalman gain, which is updated during the Kalman filtering process via a memristor transistor whose conductivity corresponds to the value of the Kalman gain, thereby being adjustable according to the state of the system.

[0022] On the other hand, the present invention also provides an analog sensor fusion system, which is characterized in that it integrates the above-mentioned analog Kalman filter circuit based on MoS2 memristor.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1) Each lidar signal and millimeter-wave radar signal are fused using a Kalman gain, which is updated during the Kalman filtering process via a memristor. The conductivity of the memristor corresponds to the value of the Kalman gain, which can be adjusted according to the state of the system.

[0025] 2) Compared with digital circuits, the present invention achieves a significant 1000-fold improvement in energy efficiency, demonstrating the feasibility of memristor transistors in achieving fast, energy-efficient real-time sensing and continuous signal processing in advanced sensor fusion technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a schematic diagram of an analog Kalman filter circuit based on MoS2 memristor.

[0027] Figure 2 It is a distribution diagram of different conduction states, which describes the distribution of 11 different conduction states. These states are clearly separated and do not overlap.

[0028] Figure 3 Figure 1 shows the reconfigurable output of the Kalman filter circuit for different K values, showing the output of the Kalman filter circuit in different conduction states modulated by adjusting the gate voltage. This figure demonstrates the adaptability of the memristor-based Kalman filter in various autonomous driving scenarios.

[0029] Figure 4 This is a comparison of the power consumption of the memristor analog Kalman filter circuit and the traditional digital Kalman filter circuit.

[0030] Figure 5 This is a schematic diagram of the structure of the MoS2 memristor.

[0031] Figure 6 This is a schematic diagram of a built analog Kalman filter hardware platform

[0032] Figure 7 This is a schematic diagram of the function verification of the simulated Kalman filter circuit proposed in this invention DETAILED DESCRIPTION

[0033] The present invention will be further described below with reference to the accompanying drawings and examples, but the scope of protection of the present invention shall not be limited thereto.

[0034] According to the continuous Kalman filter theory, the update equation of the vehicle position is:

[0035] x (t) =L (t) +K(R (t) -x′ (t) ) (1)

[0036] Where, L (t) Represents the position coordinates obtained from the lidar, R (t) Indicates the position coordinates obtained from the millimeter wave radar. (t) Represents the previously estimated position result, x (t) represents the output of the Kalman filter, that is, the estimated position coordinates, and K represents the Kalman gain, which becomes a constant after reaching a stable state.

[0037] Figure 1 This is a schematic diagram of an analog Kalman filter circuit based on MoS2 memristor transistor. As shown in the figure, this circuit can be used for laser radar (L (t) ) and millimeter wave radar (R (t) The data collected by the system is processed by Kalman filtering. It consists of four main modules: two differential proportional operation modules, an integral operation module, and an integral operation module based on MoS2 memristor transistors.

[0038] Part I is a differential ratio operation module composed of an operational amplifier and resistors, which is used to perform the LiDAR (L (t) ) measured value and gain adjustment signal W (t) Subtraction operation between:

[0039]

[0040] Where W (t) is the gain adjustment signal, L (t) Represents the position coordinates obtained from the lidar; U (t) Differentially amplify the LiDAR signal.

[0041] Part II is an integration module consisting of resistors, capacitors, and operational amplifiers, which is responsible for integrating the signal. It processes the signal from the lidar and updates the state estimate x (t) .

[0042]

[0043] Part III is a proportional-integral module consisting of resistors, capacitors, operational amplifiers, and memristor transistors, which is used to further process the signal. It adjusts the Kalman gain according to the state of the memristor transistor.

[0044]

[0045] W (t) =KV (t) (5)

[0046] Where V (t) is the millimeter-wave radar state deviation signal, K is the Kalman filter gain, K = Gk / C, C represents the value of capacitor C2, and Gk is the conductance of the memristor. By modulating the gate voltage, the conductance of the memristor (Gk) is adjusted accordingly. In analog Kalman filtering, by modulating Gk, the K value is adjusted for different driving scenarios. This invention can minimize the inherent defects of analog circuits, such as leakage current and bias voltage of operational amplifiers.

[0047] Part IV is another subtraction module composed of a resistor and an operational amplifier, which processes the raw data from the millimeter-wave radar (R) and uses the previous state estimate x (t) Perform signal processing.

[0048]

[0049] Where R (t) Indicates the position coordinates obtained from the millimeter wave radar; / (t) is the millimeter wave radar state deviation signal

[0050] Each input signal (L and R) is fused using a Kalman gain, which is updated during the Kalman filtering process via a memristor. The conductivity of the memristor corresponds to the value of the Kalman gain, allowing it to be adjusted according to the state of the system.

[0051] The present invention achieves the following effects:

[0052] (1) Direct processing of analog signals. The MoS2-based analog Kalman filter circuit can directly process analog signals from sensors, eliminating the need for ADCs.

[0053] (2) The analog Kalman filter circuit based on memristor can quickly adapt to different scenarios. In the above MoS2 memristor-based Kalman filter circuit, the Kalman filter gain (K) can be expressed as K = Gk / C, where C represents the value of capacitor C2 and Gk represents the conductance of the memristor. Figure 2 As shown, by modulating the gate voltage, the memristor can obtain multiple discrete conductance values Gk, thereby achieving multiple Kalman filter gains. Figure 3 The curves in the figure represent the Kalman filter results for Gk values of 5μS, 10μS, and 20μS, respectively. Unlike digital circuits, the non-volatile memory function of memristors can pre-store the final Kalman filter gains corresponding to different driving scenarios. This effectively avoids the time delay and increased power consumption caused by matrix iteration in digital circuits when switching between scenarios, thereby improving sensor fusion speed and reducing power consumption.

[0054] (3) Low power consumption and high precision. Figure 3 A quantitative comparison of the power consumption of an analog Kalman filter circuit and a traditional digital Kalman filter circuit was performed. A constant voltage source (V = ±12V) was supplied to the analog Kalman filter circuit, and the current was monitored in real time for approximately 60 seconds using an integrated analysis module. The measured current in the analog circuit was approximately 6mA, resulting in an energy consumption of 72mJ. Simultaneously, a digital Kalman filter algorithm was implemented using a field-programmable logic device (FPGA), also powered by the same 12V supply. As can be seen, the digital implementation draws approximately 250mA, resulting in an energy consumption of 3J. The analog circuit draws only 1 / 40 of the current of the digital circuit, demonstrating its reduced energy consumption. Overall, our analog Kalman filter circuit achieves reliable sensor fusion while reducing energy consumption.

[0055] The following provides the construction and application scenarios of the present invention

[0056] Step 1: Preparation of MoS2 memristor. Memristor is prepared by micro-nano processing technology. Figure 5Schematic diagram of the top-gate memristor structure. The top-gate memristor was fabricated on a continuous MoS2 film grown directly on a sapphire substrate using chemical vapor deposition (CVD) without using a transfer process. The memristor fabrication process can be described as follows: The sample was first spin-coated with PMMA 950 at 4000 RPM for 45 seconds and then baked at 180°C for 120 seconds. Subsequently, the source and drain electrodes were precisely patterned using electron beam lithography. The exposed areas were developed using a 1:3 mixture of 4-methyl-2-pentanone (MIBK) and isopropyl alcohol (IPA) for 90 seconds, followed by a 60-second rinse in pure IPA. To complete the device, Pt / Au (10 nm / 40 nm) metal electrodes were deposited onto the patterned areas using electron beam evaporation. The lift-off process was facilitated by immersing the sample in N-methyl-2-pyrrolidone (NMP) for 30 minutes, followed by immersion in IPA for 15 minutes. As a next step, a 50-nm-thick aluminum oxide (Al2O3) layer, used as the gate dielectric, was grown by atomic layer deposition (ALD) at 300°C. Electron beam lithography was then used to precisely define the gate, followed by the deposition of Ti / Au (10nm / 35nm) metal, which was subsequently lifted off.

[0057] Step 2: Hardware construction. Figure 6 As shown, according to Figure 1 Design the circuit, select appropriate resistors, capacitors, operational amplifiers, and the MoS2 memristor transistor prepared in step 1, and integrate them onto a printed circuit board. Use a dual-channel waveform generator to simulate data collected by the lidar and millimeter-wave radar, powered by a precision power supply, and an oscilloscope to capture the filtering results of the simulated Kalman filter circuit.

[0058] Step 3: Functional verification. Figure 7 The target vehicle's trajectory is assumed to be a sine wave with an amplitude of 1V. The data collected by the lidar and millimeter-wave radar, represented as L(t) and R(t), are sinusoidal waves with a 90° phase difference. This data is input into the hardware platform prepared in step 2 via a signal generator, and the filtered results of the Kalman filter circuit are acquired using an oscilloscope. It can be seen that the filtered output signal (x(t)) accurately reproduces the vehicle's trajectory and is very similar to the trajectory obtained through simulation, demonstrating the feasibility of the proposed simulated Kalman filter for sensor fusion.

[0059] The above are only preferred embodiments of the present invention and do not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. An analog Kalman filter circuit based on MoS2 memristor, characterized in that: include: A differential proportional operation module, comprising at least one operational amplifier and a plurality of resistors, for performing a differential operation between the collected lidar signal and the gain adjustment signal. By adjusting the resistance values of the resistors, an amplifier circuit with an adjustable gain factor is created. This gain factor amplifies the input signal, thereby generating a lidar differential amplified signal. an integral operation module, comprising at least one operational amplifier and a relatively large resistor and capacitor connected in parallel, for performing a phase shift on the differential amplified signal of the laser radar and updating a state estimation value; a subtraction operation module, comprising at least one operational amplifier and a plurality of resistors, for performing a differential operation between the collected millimeter-wave radar signal and the output of the analog Kalman filter circuit, and outputting a millimeter-wave radar state deviation signal; The integral operation module based on the memristor transistor includes at least one operational amplifier, a memristor transistor, and a larger resistor and capacitor connected in parallel. By adjusting the gate voltage of the memristor transistor, the conductance value of the memristor transistor is changed, thereby adjusting the Kalman filter gain in different driving scenarios, and outputting the gain adjustment signal to the differential proportional operation module according to the millimeter-wave radar state deviation signal.

2. The analog Kalman filter circuit based on MoS2 memristor according to claim 1, characterized in that: Each of the lidar signals and millimeter-wave radar signals is fused via a Kalman gain, which is updated during the Kalman filtering process via a memristor transistor whose conductivity corresponds to the value of the Kalman gain, thereby being adjustable according to the state of the system.

3. An analog sensor fusion system, characterized in that: An analog Kalman filter circuit based on a MoS2 memristor transistor comprising the method of claim 1 or 2.

4. Application of the analog Kalman filter circuit based on MoS2 memristor according to claim 1 or 2 in autonomous driving.

Images