Robot mapping and navigation method based on bionic beard touch sense

Abstract

The invention provides a robot mapping and navigation method based on bionic beard touch, and aims to solve the problem that a traditional vision and laser radar system fails in a complex and low-visibility environment. The specific process comprises the following steps: initializing and self-checking the bionic beard sensor to ensure normal work of each module; the motion of the beards is controlled through the pulse neural network, and the motion rhythm of the beards is regulated and controlled by depending on the central mode generator, so that the touch perception of the surrounding environment is realized; the collected tactile data is subjected to reconstruction filtering and low-pass filtering processing, position information of an object is extracted, and a complete environment map is generated through fusion of a grid map construction algorithm; and performing path planning on the generated tactile map by adopting an A * algorithm, calculating an optimal path and executing a navigation task. The system is suitable for robot autonomous navigation in a complex environment, and efficient and accurate environment perception and navigation can be realized through the bionic beard sensor and the neural network technology.

Description

Technical Field

[0001] The present invention belongs to the field of bionic technology and artificial intelligence, and specifically relates to a robot mapping and navigation method based on bionic whisker tactile sensation. Background Art

[0002] With the rapid development of artificial intelligence and robotics, robot bionics and artificial intelligence have become the core research directions in the field of intelligent robots. Traditional robot navigation systems mainly rely on technologies such as vision and lidar, which achieve environmental perception and path planning by acquiring visual or laser data of the environment. However, these technologies have obvious limitations in certain specific scenarios, especially in complex or restricted environments, such as smoke, dust, insufficient light or complete darkness. The perception capabilities of vision and lidar will drop significantly or even fail completely. In these cases, robot navigation systems that rely solely on vision or lidar may not be able to provide sufficiently accurate environmental information, thus affecting the robot's autonomy and the reliability of task execution.

[0003] Tactile perception systems have gradually become an effective alternative. Tactile perception systems, similar to animal whiskers, can perceive information such as the shape, position, and texture of objects through physical contact with environmental objects, thereby providing environmental perception capabilities that are completely different from vision and lidar. Especially in extreme environments where vision and lidar cannot work effectively, tactile perception systems have shown great potential. For example, in the dark, visual perception can hardly work, while tactile sensors can continue to detect and perceive environmental information through physical contact.

[0004] At present, traditional robot navigation systems usually rely on the following two main perception technologies: The first is visual perception technology, which obtains visual information of the environment through monocular, binocular or RGB-D cameras, and uses computer vision algorithms for image processing, feature extraction and matching, so as to achieve environmental mapping and navigation; this technology performs well in environments with sufficient light and no obstructions, but the performance will be significantly reduced in poor lighting conditions or visual obstruction. The second is laser radar perception technology, which measures the distance and shape of objects in the environment by emitting laser beams and receiving reflected signals through laser radar (LIDAR), so as to achieve environmental mapping; this technology has high accuracy and robustness, especially in outdoor or large-scale scenes. However, in smoke, dust or complex geometric structure environments, the perception ability of laser radar will also be limited.

[0005] Therefore, providing a robot navigation system that can effectively work in complex or restricted environments is a technical problem that needs to be urgently solved by those skilled in the art. Summary of the invention

[0006] The present invention aims to provide a robot mapping and navigation system based on whisker-like tactile perception, which can achieve efficient autonomous navigation by simulating the tactile perception mechanism of animal whiskers and combining signal processing and other technologies. Its main purpose is to enhance the robot's autonomous ability and adaptability in complex and low-visibility environments, such as disaster relief, industrial applications, and home environments. The invention aims to solve the following three key problems:

[0007] 1. Addressing the limitations of traditional navigation systems in extreme environments: When conventional perception technologies such as vision or lidar fail in environments such as smoke, darkness or dust, the tactile system provides a reliable means of perception, ensuring that the robot can continue to effectively perceive the surrounding environment.

[0008] 2. Improve environmental adaptability and flexibility: The system uses tactile perception technology to achieve adaptive navigation in complex environments, enhance the robot's real-time response capabilities to obstacles and dynamic environments, and improve the flexibility of autonomous navigation.

[0009] 3. Provide low-power, cost-effective solutions: Compared with other high-energy-consuming perception technologies, such as lidar and vision systems, whisker-like tactile sensors can significantly reduce energy consumption while maintaining efficient navigation, making them suitable for resource-constrained environments.

[0010] In order to solve the above problems, the present invention proposes a robot mapping and navigation system based on whisker-like touch. First, a whisker-like tactile sensor is designed, and a pulse neural network is used to imitate the perception mechanism of animal whiskers, and the position information of objects is obtained through physical contact to ensure that the robot can effectively perceive in complex environments. Secondly, a tactile signal processing and mapping algorithm is developed to reduce the noise of the sensor data, and then the coordinates of the contact points are calculated to generate a tactile-based environmental map to improve environmental adaptability. Finally, a low-power autonomous navigation and path planning system is designed, and path planning is performed through the A* algorithm to ensure that the robot selects the optimal path and avoids obstacles in a dynamic environment. This system not only solves the limitations of traditional navigation, but also improves the robot's autonomy and reliability in extreme environments.

[0011] The specific workflow of the present invention is as follows:

[0012] Step 1: Start the system. The robot will initialize the bionic whisker sensor, and then the system will perform a self-check to confirm that all sensors and modules are in normal working condition.

[0013] Step 2: The bionic beard sensor is controlled by a pulse neural network to realize the movement of the beard.

[0014] Step 3: perceive the surrounding tactile information in real time, control the movement of the whisker based on the tactile information feedback, and transmit the data to the signal processing module.

[0015] Step 4: preprocess the collected tactile data (including reconstruction filtering and low-pass filtering) to extract the position information of the object.

[0016] Step 5: Fuse the processed data to generate a complete environment map. The map generation process uses a technology similar to SLAM.

[0017] Step 6: After generating the tactile map, the system uses the A* algorithm to calculate the optimal path, and the robot performs the navigation task according to the path planning results.

[0018] The specific steps are as follows:

[0019] Step 1: System startup and sensor initialization

[0020] When the robot starts, the initialization of the bionic whisker sensor involves zero point calibration and gain calibration.

[0021] 1. Zero point calibration: Calibrate the initial state of the sensor so that the sensor output is zero when there is no external interference.

[0022] S calibrated =S measured -ΔS

[0023] Among them, S measured is the actual measured signal of the sensor, and ΔS is the calibrated offset.

[0024] 2. Gain calibration: Calibrate the gain factor G to ensure that the sensor output is linearly related to the displacement.

[0025] S output =G·x

[0026] Where x is the input physical quantity (displacement), and G is the gain coefficient of the sensor, which is a constant.

[0027] Step 2: Beard Motion Control

[0028] The whisker pad, as the main receiver of sensory information, transmits environmental tactile information to the nervous system, involving the trigeminal ganglion (TG), trigeminal nucleus (TN) and facial nucleus (FN). In this process, the central pattern generator (CPG) autonomously generates rhythmic neural activity and sends signals to the FN, thereby regulating whisker movement.

[0029] The whisker movement is controlled by a spiking neural network (SNN), which simulates the discharge mechanism of biological neurons. The neuron membrane potential V(t) is calculated as follows:

[0030]

[0031] Among them C m is the capacitance of the cell membrane surface, G L 、E L is the leakage conductance, which are the parameters of the neuron model itself, I app is the external input. When the membrane potential V(t) exceeds the threshold V th When =-55mV, the neuron sends a pulse signal.

[0032] Whisker motor control is generated by the Central Pattern Generator (CPG), which controls the movement of the whisker pad muscles. Neurons are located in the facial nucleus (FN) and send motor commands through the facial nerve. About 80% of the FN neurons are protract neurons, which cause the extension of a single whisker, and about 20% are retract neurons, which cause the retraction of multiple whiskers.

[0033] The firing rate of FN neurons determines the angle of whisker movement

[0034] When the protract neuron fires, the whiskers extend forward at an angle of:

[0035] θ pro =R pro *θ max

[0036] When the retract neuron fires, the whiskers extend forward at an angle of:

[0037] θ ret =R ret *θ min

[0038] Where R pro , R ret are the firing rates of the protract neurons and the retract neurons, respectively. max and θ min Take π / 3 and -π / 4 respectively.

[0039] Step 3: Environmental Perception

[0040] The bionic whisker sensor collects environmental information by sensing the force and displacement when an object contacts the whisker. The formulas involved include Hooke's law and acceleration perception formula.

[0041] 1. Hooke's Law: The force on the beard is proportional to the displacement, which conforms to Hooke's Law:

[0042] F=k·x

[0043] Where F is the force felt by the whiskers, k is the stiffness coefficient of the whiskers, and x is the displacement.

[0044] 2. Acceleration perception: When the whiskers sense the acceleration of an object, Newton's second law is used:

[0045] F=m·a

[0046] Among them, F is the force sensed by the whisker sensor, m is the mass of the object, and a is the acceleration of the object.

[0047] When the whiskers sense an object, they activate the contact cells in the TG, which are connected to the FN neurons through the contact cells in the TN, thereby inhibiting the discharge of protract neurons and stimulating the discharge of retract neurons, causing the whiskers to retract and reduce the degree of curvature of the whiskers.

[0048] Step 4: Data preprocessing and data collection

[0049] Tactile data preprocessing mainly includes reconstruction filtering and low-pass filtering.

[0050] 1. Reconstruction filtering:

[0051] This process is equivalent to correcting the displacement x and y of the Hall sensor according to the sensor's whisker angle input θ to reduce noise and error, and the final output is x q ,y q .

[0052] The output of the filter f(θ(t)) is obtained by passing a series of filter coefficients b i The weighted sum of the input signal θ is obtained as follows:

[0053] f(θ(t))=b0·θ(t)+b1·θ(t-1)+b2·θ(t-2)+…+b n ·θ(tn)

[0054] Where θ is the beard angle and the filter coefficient b i Generated by a filter design tool such as the filter design module in MATLAB or SciPy.

[0055] The sensor output signal is obtained by subtracting the result of the filter from the original signal:

[0056] x q (t) = x(t) - f(θ(t))

[0057] y q (t) = y(t) - f(θ(t))

[0058] 2. Low pass filtering:

[0059] Low-pass filters are usually used to remove high-frequency noise from sensor signals and retain relatively stable and important signal features. Here, frequencies below 50 Hz are retained, and the calculation formula for the filter output signal y(t) is:

[0060]

[0061] where q i and c i is the filter coefficient, generated by a filter design tool (such as the filter design module in MATLAB or SciPy), t is the discrete time, ti represents the time before the i-th time step, and x(t) is the input, that is, the filter output x of the previous step. q ,y q , y(t) is the filter output, expressed as x f ,y f .

[0062] 3. Data collection:

[0063] According to x f ,y f Signal calculation whisker contact force:

[0064]

[0065] macon=(lT)*K

[0066] Where l is the Euclidean distance, T is the contact threshold, which is generally 0.3, K is the contact gain, which is generally 5.0, and finally the contact force macon is limited to the range of [0.0, 1.0].

[0067] In order to reduce the amount of data and high-frequency noise, a simple average is performed on the whisker angle signal θ and the contact force signal macon.

[0068] When macon is greater than 0, the corresponding end coordinate point of the whisker is recorded as the contact point. Finally, all contact points are saved as a .csv file.

[0069] Step 5: Environment Mapping

[0070] At this stage, the system uses a method similar to SLAM (Simultaneous Localization and Mapping) to generate an environmental map through tactile data.

[0071] The processed data is fused through a grid map construction algorithm to generate a complete environmental map. This method first converts the sensor data into a grid form, then fills the information in the grid through interpolation and data fusion techniques, and finally generates a high-resolution environmental map, providing an accurate basis for subsequent navigation and path planning.

[0072] Step 6: Path Planning

[0073] After generating the tactile map, the system uses the A* algorithm (for details, please refer to http: / / theory.stanford.edu / ~amitp / GameProgramming / AStarComparison.html#the-a-star-algorithm) for path planning and calculates the optimal path.

[0074] The A* algorithm finds the optimal path by evaluating the heuristic function f(n):

[0075] f(n)=g(n)+h(n)

[0076] Among them, g(n) is the actual cost from the starting point to node n, and h(n) is the heuristic estimated cost from node n to the target point.

[0077] The present invention has the following advantages:

[0078] (1) Adapt to complex environment:

[0079] The Whisker tactile system can work effectively in low-visibility and extreme environments, making up for the limitations of traditional vision and lidar. In environments such as smoke-filled disaster sites and dark underground pipelines, the physical contact information provided by tactile sensors is irreplaceable.

[0080] (2) Real-time feedback and response:

[0081] The real-time feedback provided by the tactile sensor enables the robot to respond quickly to changes in the environment. This real-time response capability is crucial for obstacle avoidance and path adjustment, ensuring that the robot can complete navigation tasks safely and efficiently.

[0082] (3) Low power consumption and high efficiency:

[0083] Compared to vision and lidar systems, the whisker tactile sensor consumes less energy and is suitable for scenarios with limited resources. The system's low power consumption allows it to maintain high efficiency during long-term operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 Workflow diagram of a robot mapping and navigation system based on bionic whisker tactile sensation

[0085] Figure 2 Simulation robot based on NRP platform

[0086] Figure 3 Contact point scatter plot

[0087] Figure 4 Raster Map

[0088] Figure 5 Navigation using the A* algorithm DETAILED DESCRIPTION

[0089] In order to describe the technical solution of the invention in more detail, the present invention is further described below with reference to the accompanying drawings and examples.

[0090] like Figure 1 As shown, the specific steps of the present invention are as follows:

[0091] 1. Bionic beard structure design

[0092] The bionic whiskers are made of flexible materials and can bend slightly when they touch an object. Sensors embedded in the whiskers can detect this physical deformation and output the force information of the whiskers. Each whisker sensor has multiple tactile nodes, which improves the spatial resolution. The choice of flexible materials and the structural design of the whiskers can be adjusted according to the needs of different application scenarios. For example, in scenarios that require higher tactile accuracy, the whiskers can be designed to be more sensitive and delicate.

[0093] 2. Beard motion control

[0094] Whisker movements are controlled by spiking neural networks (SNNs), which simulate the firing mechanism of biological neurons. Whisker movements are generated by the central pattern generator (CPG), a control mechanism that can oscillate autonomously and does not rely on the external environment. Motor neurons in the facial nucleus (FN) send movement commands through the facial nerve. The firing rate of FN neurons determines the angle of whisker movement, with the angle of extension and retraction being regulated by the firing rate of the corresponding neurons, respectively.

[0095] 3. Tactile data processing method

[0096] The tactile data processing method is one of the cores of the present invention. The tactile data from the whisker sensor first undergoes a preprocessing stage, including reconstruction filtering and low-pass filtering. Then, based on the tactile information, the contact point coordinates are calculated to obtain the location of the obstacle. The goal of tactile data processing is to convert the information of physical contact into actionable environmental perception data to help the robot make navigation decisions.

[0097] 4. Tactile map generation

[0098] The tactile mapping module is another key component of the system. This module gradually builds a complete map of the environment by fusing the tactile data of the robot at different positions. The algorithm uses a similar SLAM approach, but builds the map based on tactile data.

[0099] 5. Tactile navigation algorithm

[0100] Based on the constructed tactile map, the robot uses the A* algorithm for path planning to ensure that the robot can move autonomously in complex environments. The design goal of the tactile navigation algorithm is to maximize navigation efficiency and adapt to different environmental changes while ensuring safety.

[0101] The above description of the specific implementation method of the present invention is only used for those skilled in the art to understand the present invention and is not used to limit the scope of the present invention.

Claims

1. A robot mapping and navigation method based on bionic whisker tactile sensation, characterized in that: The steps include: Step 1, start the bionic beard sensor; Step 2, the bionic beard sensor is controlled by a pulse neural network to realize the movement of the beard; Step 3, sensing the surrounding tactile information in real time, controlling the movement of the whisker according to the tactile information feedback, and transmitting the data to the signal processing module; Step 4, preprocessing the collected tactile data, including reconstruction filtering and low-pass filtering, to extract the position information of the object; Step 5: Fuse the processed data to generate a complete environment map; the map generation process uses a technology similar to SLAM; Step 6: After generating the tactile map, the system uses the A* algorithm to calculate the optimal path and performs the navigation task based on the path planning results.

2. A robot mapping and navigation system based on bionic whisker tactile sensation as claimed in claim 1, characterized in that: The movement of the beard is achieved by controlling the pulse neural network as described in step 2. The specific steps are as follows: The whisker pad WP is the main receiver of sensory information, transmitting environmental tactile information to the nervous system, involving the trigeminal ganglion TG, trigeminal nucleus TN and facial nerve nucleus FN; in this process, the central pattern generator CPG autonomously generates rhythmic neural activity and sends signals to FN, thereby achieving the regulation of whisker movement; The whisker movement is controlled by a spiking neural network, SNN, which simulates the discharge mechanism of biological neurons; the input pulse u(t) is calculated by the neuron membrane potential: Among them C m is the capacitance of the cell membrane surface, G L is the first leakage conductance, E L is the second leakage conductance, I app is the external input. When the membrane potential V(t) exceeds the threshold V th =-55mV, the neuron sends a pulse signal; Whisker movement control is generated by the central pattern generator CPG, which controls the movement of the whisker pad muscles; the neurons are located in the facial nucleus FN, which sends movement commands through the facial nerve; 80% of the FN neurons are protract neurons that cause a single whisker to extend, and 20% are retract neurons that cause multiple whiskers to retract; The firing rate of FN neurons determines the angle of whisker movement; When the protract neuron fires, the whiskers extend forward at an angle of: i pro =R pro *i max When the retract neuron fires, the whiskers extend forward at an angle of: i ret =R ret *i min Where R pro , R ret are the firing rates of the protract neurons and the retract neurons, respectively. max and θ min Take π / 3 and -π / 4 respectively.

3. A robot mapping and navigation system based on bionic whisker tactile sensation as claimed in claim 1, characterized in that: The real-time perception of surrounding tactile information in step 3 and the control of the beard movement according to the tactile information feedback are as follows: The bionic whisker sensor collects environmental information by sensing the force and displacement when an object contacts the whisker; the formulas involved include Hooke's law and acceleration perception formula; (1) Hooke's Law: The force on the whisker is proportional to the displacement, which conforms to Hooke's Law: F=k·x Where F is the force felt by the whiskers, k is the stiffness coefficient of the whiskers, and x is the displacement; (2) Acceleration perception: When the whiskers sense the acceleration of an object, Newton’s second law is used: F=m·a Where F is the force sensed by the whisker sensor, m is the mass of the object, and a is the acceleration of the object; When the whiskers sense an object, they activate the contact cells in the TG, which connect to the FN neurons through the contact cells in the TN, thereby inhibiting the discharge of protract neurons and stimulating the discharge of retract neurons, causing the whiskers to retract and reduce the degree of curvature of the whiskers.

4. A robot mapping and navigation system based on bionic whisker tactile sensation as claimed in claim 1, characterized in that: The reconstruction filtering and low-pass filtering described in step 4, and the position information of the object are extracted, are as follows: (1) Reconstruction filtering: This process is equivalent to correcting the displacement x and y of the Hall sensor according to the sensor's whisker angle input θ to reduce noise and error, and the final output is x q ,y q ; The output of the filter f(θ(t)) is obtained by passing a series of filter coefficients b i The weighted sum of the input signal θ is obtained as follows: f(θ(t))=b0·θ(t)+b1·θ(t-1)+b2·θ(t-2)+…+b n ·θ(t-n) The sensor output signal is obtained by subtracting the result of the filter from the original signal: x q (t)=x(t)-f(θ(t)) y q (t)=y(t)-f(θ(t)) (2) Low-pass filtering: The low-pass filter is used to remove high-frequency noise in the sensor signal and retain the frequency below 50Hz. The calculation formula of the filter output signal y(t) is: where a i and c i is the filter coefficient, generated by the filter design tool, t is the discrete time, ti represents the time before the i-th time step, and x(t) is the input, that is, the filter output x of the previous step. q ,y q , y(t) is the filter output, expressed as x f ,y f ; (3) Data collection: According to x f ,y f Signal calculation whisker contact force: macon=(lT)*K Where l is the Euclidean distance, T is the contact threshold, which is 0.3, K is the contact gain, which is 5.0, and finally the contact force macon is limited to the range of [0.0, 1.0]; When macon is greater than 0, the corresponding end coordinate point of the whisker is recorded as the contact point. Finally, all contact points are saved as a .csv file.

5. A robot mapping and navigation system based on bionic whisker tactile sensation as claimed in claim 1, characterized in that: Step 5 is as follows: the processed data is fused through a grid map construction algorithm to generate a complete environmental map; this method first converts the sensor data into a grid form, then fills the information in the grid through interpolation and data fusion techniques, and finally generates an environmental map.

6. A robot mapping and navigation system based on bionic whisker tactile sensation as claimed in claim 1, characterized in that: Step 6 is as follows: After generating the tactile map, the system uses the A* algorithm for path planning and calculates the optimal path; The A* algorithm finds the optimal path by evaluating the heuristic function f(n): f(n)=g(n)+h(n) Among them, g(n) is the actual cost from the starting point to node n, and h(n) is the heuristic estimated cost from node n to the target point.

Images