A multi-sensor fusion-based intelligent wheelchair autonomous navigation and obstacle avoidance mechanism and use method

Abstract

This invention discloses an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism and its usage method based on multi-sensor fusion. The device includes a wheelchair frame and integrated environmental perception module, core control module, drive execution module, and human-machine interaction module. The environmental perception module includes front and rear radars; the core control module includes a host, a hierarchical power supply management system, and a network switch; the drive execution module includes two wheelchair motors and corresponding motor drive boards; and the human-machine interaction module includes a touchscreen. This invention constructs an omnidirectional horizontal perception field through front and rear dual radars, eliminating blind spots; the host runs synchronous positioning and map building algorithms to achieve autonomous mapping and positioning; and combines global path planning (A / B) and dynamic window method for local obstacle avoidance to achieve full-process autonomous navigation; the hierarchical power supply ensures stable system operation. This invention enables intelligent wheelchairs to achieve autonomous navigation and safe obstacle avoidance in complex environments, improving the independence and safety of user movement.

Description

Technical Field

[0001] This invention belongs to the field of intelligent wheelchair navigation and control technology, and in particular relates to an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism and its usage method based on multi-sensor fusion. Background Technology

[0002] With the deepening of global population aging and the increasing demand for independent travel among people with disabilities, intelligent wheelchairs, as an important mobility assistive device, have received widespread attention from all sectors of society. While traditional electric wheelchairs have alleviated some of the user's physical strain, they are still essentially human-operated assistive tools, requiring continuous directional control and speed adjustment. However, for the elderly, those with limited mobility, and those with cognitive impairment, maintaining concentration for extended periods to operate a wheelchair is a significant challenge. Especially in complex indoor environments such as hospitals, nursing homes, and large shopping malls, users must simultaneously navigate narrow corridors, frequent pedestrians, temporarily piled items, and changing entrance and exit layouts. This places extremely high demands on their environmental perception, path judgment, and emergency response capabilities. Misjudgment or delayed reaction can easily lead to collisions, rollovers, and other safety accidents, seriously threatening the user's life and health. Therefore, how to equip intelligent wheelchairs with autonomous environmental understanding, intelligent path decision-making, and reliable obstacle avoidance capabilities, enabling them to safely move from start to finish without continuous user intervention, has become a core technical problem urgently needing to be solved in this field.

[0003] To address the aforementioned technical issues, those skilled in the art have conducted some beneficial explorations. Currently, some high-end electric wheelchairs or laboratory prototypes on the market are beginning to incorporate obstacle avoidance sensors to improve safety. A common technical solution involves installing ultrasonic or infrared sensors in front of the wheelchair to detect nearby obstacles. When the sensor detects an obstacle at a distance less than a preset threshold, the system will issue an audible and visual alarm or automatically execute emergency braking to prevent a collision. This solution is essentially a passive safety protection method, intervening only when a collision is imminent. However, its intervention method is relatively simple and crude, often merely involving emergency stopping, and it cannot achieve proactive detour or path replanning.

[0004] Another improvement involves installing a single-line LiDAR in front of the wheelchair. Leveraging the wide scanning range and high ranging accuracy of LiDAR, a two-dimensional contour map of the surrounding environment can be constructed. Based on this, some studies have attempted to introduce simple path planning algorithms, enabling the wheelchair to travel along a pre-set route and perform simple detours when encountering obstacles. Furthermore, some solutions employ depth cameras as visual sensors, using image recognition technology to identify specific targets such as door frames and pedestrians, assisting navigation decisions. These technological solutions, to some extent, enhance the intelligence level of wheelchairs, giving them a basic ability to perceive their environment.

[0005] Despite the progress made in existing technologies, several shortcomings remain. First, current solutions generally employ a single type of sensor, resulting in significant blind spots in the field of view. For example, a wheelchair equipped only with a front-facing LiDAR can only perceive the environment in front, completely failing to detect pedestrians suddenly approaching from behind or obstacles in the side blind spots, posing a serious safety hazard in scenarios requiring backward movement or U-turns. Second, the navigation functions in existing technologies mostly remain at the level of "preset route tracking" or "simple obstacle avoidance," lacking true autonomous positioning and map-building capabilities. They cannot autonomously explore and build environmental maps in unknown environments, nor can they dynamically adjust the global path based on real-time environmental changes. Third, the power supply and communication architecture of existing systems is relatively simple. When integrating multiple sensors and computing units, problems such as unstable power supply and data transmission delays can easily occur, affecting the reliability and real-time performance of the system. Furthermore, existing technologies are weak in human-computer interaction; users find it difficult to intuitively understand the wheelchair's perception results and navigation intentions, and cannot easily set target points.

[0006] Therefore, it is necessary to propose an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism, usage method, and system based on multi-sensor fusion to solve the above problems. Summary of the Invention

[0007] The technical problem to be solved by this invention is to provide an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism and its usage method based on multi-sensor fusion. By configuring front and rear dual LiDARs, a 360-degree horizontal perception field is constructed, eliminating blind spots. Through synchronous positioning and map building algorithms integrated into the host unit, autonomous mapping and real-time positioning in unknown environments are achieved. Through A... The combination of global path planning algorithm and dynamic window method for local obstacle avoidance planning enables fully autonomous navigation from the starting point to the target point. A hierarchical power supply management system ensures stable power supply to each module, and a network switch guarantees high-speed transmission of radar point cloud data. A touch screen enables visualization of the environmental map and intuitive setting of target points. This approach aims to solve the problems of blind spots, insufficient navigation capabilities, and poor system reliability in existing technologies, and truly realize fully autonomous navigation and safe obstacle avoidance for intelligent wheelchairs.

[0008] To achieve the above technical solution, the technical solution adopted by the present invention is as follows: An intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion includes a wheelchair frame and an integrated environmental perception module, core control module, drive execution module, and human-machine interaction module. The environmental perception module includes a front radar mounted at the front of the frame and a rear radar mounted at the rear. The core control module includes a host computer serving as the main control computing unit, a hierarchical power supply management system, and a network switch. The hierarchical power supply management system includes a power supply, an air switch in the main power output circuit, and a voltage regulator. The drive execution module includes two wheelchair motors and motor drive boards connected to each motor, with the motors driving the rear wheels. The human-machine interaction module includes a touchscreen. The front and rear radars are connected to the network switch via network cables, and the network switch is connected to the host computer via network cables. The host computer is connected to the motor drive boards via a communication module. The power supply sequentially powers the host computer, front radar, rear radar, and network switch via the air switch and voltage regulator. The system also includes a front wheel and a rear auxiliary wheel for the omnidirectional wheels.

[0009] Preferably, the voltage regulator includes a first DC-DC voltage regulator and a second DC-DC voltage regulator. The first DC-DC voltage regulator converts the output voltage of the power supply into a first voltage to power the host, the front radar, and the rear radar. The second DC-DC voltage regulator converts the output voltage of the power supply into a second voltage to power the network switch.

[0010] Preferably, the human-computer interaction module also includes a mouse and keyboard, which are connected to the host computer; the communication module is a USB to CAN bus communication module; the front radar is a forward-facing lidar, and the rear radar is a rearward-facing lidar; the host computer is installed at the center of the bottom of the wheelchair frame, the front radar is installed on the front support of the frame, and the rear radar is installed at the back of the frame; the two wheelchair motors drive the left rear wheel and the right rear wheel respectively, and the two motor drive boards are connected to the two wheelchair motors one-to-one.

[0011] Furthermore, the power supply is a DC battery with a rated voltage of 24V, and an air switch is connected in series in the total output circuit of the power supply as the main system switch and short-circuit protection device.

[0012] Furthermore, a touchscreen is mounted at the front of the vehicle frame to display navigation maps, system status, and receive destination input.

[0013] Preferably, a method for using an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion includes the following steps: S1, System power-on and initialization, close the air switch, the hierarchical power supply management system supplies power to each module, the host, front radar and rear radar start; S2, Environmental perception and mapping localization: The front and rear radars continuously scan the surrounding environment and send point cloud data to the host through a network switch. The host fuses the dual radar data, calculates the wheelchair's real-time pose in the current environment, and builds or updates the environmental map. S3, Task setting and path planning: Users can view the environment map built by the host through the touch screen and set the target point on the touch screen. The host generates a global path from the current location to the target point based on the current map and real-time pose. S4, Real-time Obstacle Avoidance and Motion Control: As the wheelchair moves along the global path, the main unit plans the local motion trajectory at the current moment under the guidance of the global path based on the obstacle information detected in real time by the front and rear radars, and generates speed control commands. S5, Command Execution and Movement: The host sends speed control commands to the motor drive board through the communication module. The motor drive board drives the corresponding wheelchair motor to rotate, and the wheelchair moves through differential drive of the rear wheels until it reaches the target point.

[0014] Preferably, in step S2, the specific method for constructing an environmental map by fusing dual radar data on the host is as follows: The host computer will collect the first point cloud data from the front-mounted radar. The second point cloud data collected by the rear radar The fusion is performed, where is the two-dimensional coordinate vector of the i-th point in the first point cloud on the horizontal plane. Let be the two-dimensional coordinate vector of the j-th point in the second point cloud on the horizontal plane; The fused point cloud data is projected onto the occupying raster map, with each raster cell in the map... The occupancy probability is iteratively updated using the Bayesian update formula: ; in, This represents the grid cell in the u-th row and v-th column of the map. This represents all radar observation data from time 1 to t. This represents the posterior probability of a raster cell being occupied, derived from all historical observation data. The probability of a grid cell being occupied, based on radar observation data at the current time t, is calculated using an inverse sensor model based on radar ranging values.

[0015] Preferably, in step S2, the specific method for the host computer to calculate the real-time pose of the wheelchair is as follows: The host computer uses an iterative nearest-point algorithm to register the fused point cloud at the current time step with the fused point cloud at the previous time step, obtaining the pose change of the wheelchair; let the pose of the wheelchair at the previous time step be... ,in The x-axis is... The vertical axis is , The heading angle is used; the pose increment relative to the previous moment is obtained through registration calculation. Then the real-time position of the wheelchair at the current moment. Calculated using the following formula: ; in, This represents the displacement increment along the X-axis. This represents the displacement increment in the Y-axis direction. The change in heading angle is denoted by , and all three are obtained by solving the point cloud registration results.

[0016] Preferably, in step S3, the specific method for the host to generate the global path is as follows: The host uses A Path planning algorithms generate paths from the wheelchair's current position. To the target point The global path, and the evaluation function of the algorithm is: ； Where n represents the grid node occupied in the grid map, and g(n) is the actual movement cost from the current position of the wheelchair to node n, calculated using the following formula: ; In the formula, K is the number of nodes traversed on the path from the starting point to node n, and X... k Let be the coordinates of the k-th node on the path. The Euclidean distance between adjacent nodes; h(n) represents the distance from node n to the target point. The heuristic cost estimation is calculated using Euclidean distance: ; in, Let n be the coordinates of node n. The coordinates of the target point are given. The algorithm maintains an open list and a closed list, iteratively searches for the node with the smallest evaluation function value until it extends to the target point, and generates a global path node sequence by backtracking the parent node.

[0017] Preferably, in step S4, the specific method for the host computer to plan the local motion trajectory and generate speed control commands is as follows: The host computer uses a dynamic window method for local path planning, within the speed space. Medium-sampled multiple sets of linear velocity v and angular velocity Each speed group corresponds to a simulated trajectory, and the optimal speed combination is selected through an evaluation function; the evaluation function is: ; in, This is the azimuth evaluation item, used to measure the consistency between the orientation of the simulated trajectory's endpoint and the target point's direction. The calculation formula is: ; In the formula, To simulate the heading angle of the wheelchair at the end of the trajectory, The azimuth angle of the target point relative to the endpoint; This is the obstacle distance evaluation item, used to measure the minimum distance between the simulated trajectory and the nearest obstacle. If the trajectory collides with an obstacle, this item is set to zero. This is a speed evaluation item, used to encourage high-speed movement, and its calculation formula is: ; In the formula, v is the sampling linear velocity; These are the weighting coefficients. Normalization factor; Host (select evaluation function) The speed combination with the highest value This serves as the optimal speed control command for the current moment.

[0018] Preferably, in step S5, the specific method for the motor drive board to drive the wheelchair motor to rotate is as follows: The host will use the linear velocity in the optimal speed control command. and angular velocity The data is transmitted to the motor drive board via the communication module. The motor drive board converts the linear velocity and angular velocity into the target rotational speeds of the left and right wheels based on the differential drive model. The conversion formula is as follows: ; in, The target linear velocity of the left rear wheel. Let L be the target linear velocity of the right rear wheel, and L be the wheelbase between the left and right rear wheels. The motor drive board uses a PID control algorithm to adjust the speed of the wheelchair motor, minimizing the error between the actual speed and the target speed. The PID control law is as follows: ; Where u(t) is the control quantity output by the motor drive board, and e(t) is the error between the target speed and the actual speed. These are the proportional coefficient, integral coefficient, and differential coefficient, respectively.

[0019] Preferably, a computer device includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the method of using the intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion.

[0020] Preferably, a computer-readable storage medium stores computer instructions that cause a computer to execute the method of using the intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion.

[0021] The beneficial effects of this invention are as follows: 1. This invention constructs a 360° horizontal environmental perception field with no blind spots through the coordinated configuration of front and rear radars in the environmental perception module. The front radar is used for obstacle detection and scene scanning in front, while the rear radar is used for environmental perception behind. The data from both are fused and processed by the host computer. Compared with the existing technology that only uses a single forward sensor, this invention effectively eliminates blind spots in the rear and sides, enabling the wheelchair to perceive dynamic and static obstacles in the surrounding environment in real time during various movement modes such as forward, backward, and U-turn. Based on the data fusion of the front and rear radars, the host computer can more accurately calculate the real-time pose of the wheelchair and construct a more complete environmental map, thereby achieving all-round active obstacle avoidance during navigation, significantly reducing the risk of collisions caused by blind spots, and significantly improving the user's travel safety.

[0022] 2. This invention achieves fully autonomous navigation from the starting point to the target point through synchronous positioning and map building algorithms, global path planning algorithms, and local dynamic obstacle avoidance algorithms running on the host computer within the core control module. Users only need to click on the target location on the real-time constructed map via the touchscreen, and the system can autonomously complete a series of complex tasks such as environmental perception, pose localization, global path planning, local trajectory tracking, and dynamic obstacle avoidance. In particular, during navigation, the host computer uses a dynamic window method to plan local motion trajectories in real time. Based on the dynamic obstacle information detected in real time by the front and rear radars, it can instantly adjust the wheelchair's speed and steering to achieve smooth and safe dynamic obstacle avoidance. This "global planning + local adjustment" architecture enables this invention to adapt to various emergencies in complex dynamic environments such as hospitals, nursing homes, and shopping malls, truly freeing the user's hands and improving the wheelchair's intelligence and ease of use.

[0023] 3. This invention significantly improves the operational reliability of the entire system through a hierarchical power supply management system and an efficient communication architecture. In the hierarchical power supply management system, an air switch serves as the main circuit protection device, a first DC-DC regulator provides stable voltage to the host and dual radars, and a second DC-DC regulator powers the network switch. This effectively isolates the high-voltage power components from the low-voltage precision electronic components, preventing power interference to the computing unit and sensors caused by motor start-stop. Simultaneously, the front and rear radars are connected to the network switch via network cables, and then to the host via network cables, forming an independent local area communication network. This ensures high-speed, low-latency transmission of large-capacity radar point cloud data, providing a data foundation for real-time navigation decisions. Regarding human-computer interaction, this invention displays the navigation map, wheelchair position, planned path, and system status in real time via a touchscreen. Users do not need professional knowledge to intuitively understand the wheelchair's perception results and navigation intentions, and can issue navigation commands through simple touch operations. This significantly lowers the barrier to entry for using smart wheelchairs, making it easy for the elderly and disabled to use. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the rear side of the device structure in an embodiment of the present invention; Figure 2 This is a front view of the device structure in an embodiment of the present invention; Figure 3 This is a schematic diagram of the method flow in an embodiment of the present invention; In the diagram: 1: Touchscreen; 2: Mouse; 3: Keyboard; 4: Front radar; 5: Rear radar; 6: Air switch; 7: Voltage regulator; 8: Motor driver board; 9: Wheelchair motor; 10: Switch; 11: Power supply; 12: Main unit; 13: Rear wheel; 14: Rear wheel auxiliary wheel; 15: Front wheel. Detailed Implementation

[0025] Example 1: like Figures 1-2As shown, an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion includes a wheelchair frame and an integrated environmental perception module, core control module, drive execution module, and human-machine interaction module. The environmental perception module includes a front radar 4 installed at the front of the frame and a rear radar 5 installed at the rear of the frame. The core control module includes a host 12 as the main control computing unit, a hierarchical power supply management system, and a network switch 10. The hierarchical power supply management system includes a power supply 11, an air switch 6 located in the main power output circuit, and a voltage regulator 7. The execution module includes two wheelchair motors 9 and motor drive boards 8 connected to each motor respectively. The wheelchair motors 9 drive the rear wheels 13. The human-machine interaction module includes a touch screen 1. The front radar 4 and the rear radar 5 are connected to a network switch 10 via a network cable. The network switch 10 is connected to the host 12 via a network cable. The host 12 is connected to the motor drive board 8 via a communication module. The power supply 11 supplies power to the host 12, the front radar 4, the rear radar 5 and the network switch 10 in sequence through an air switch 6 and a voltage regulator 7. It also includes a front wheel 15 and a rear auxiliary wheel 14 of the omnidirectional wheel.

[0026] Preferably, the voltage regulator 7 includes a first DC-DC voltage regulator and a second DC-DC voltage regulator. The first DC-DC voltage regulator converts the output voltage of the power supply 11 into a first voltage to power the host 12, the front radar 4 and the rear radar 5. The second DC-DC voltage regulator converts the output voltage of the power supply 11 into a second voltage to power the network switch 10.

[0027] Preferably, the human-computer interaction module further includes a mouse 2 and a keyboard 3, which are connected to the host 12; the communication module is a USB to CAN bus communication module; the front radar 4 is a forward-facing lidar, and the rear radar 5 is a rearward-facing lidar; the host 12 is installed at the center of the bottom of the wheelchair frame, the front radar 4 is installed on the front support of the frame, and the rear radar 5 is installed behind the back of the frame; the two wheelchair motors 9 drive the left rear wheel and the right rear wheel respectively, and the two motor drive boards 8 are connected to the two wheelchair motors 9 one-to-one.

[0028] Furthermore, power supply 11 is a DC storage battery with a rated voltage of 24V, and air switch 6 is connected in series in the total output circuit of power supply 11 as the system main switch and short-circuit protection device.

[0029] Furthermore, the touchscreen 1 is mounted on the front of the vehicle frame to display navigation maps, system status, and receive destination input.

[0030] Example 2: like Figure 3 As shown, a method for using an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion includes the following steps: S1, system power-on and initialization, close air switch 6, the hierarchical power supply management system supplies power to each module, host 12, front radar 4 and rear radar 5 start; S2, Environmental perception and mapping localization: The front radar 4 and the rear radar 5 continuously scan the surrounding environment and send the point cloud data to the host 12 through the network switch 10. The host 12 fuses the dual radar data, calculates the real-time pose of the wheelchair in the current environment, and builds or updates the environmental map. S3, Task setting and path planning: The user views the environment map built by the host 12 through the touch screen 1 and sets the target point on the touch screen 1. The host 12 generates a global path from the current location to the target point based on the current map and real-time pose. S4, Real-time obstacle avoidance and motion control: As the wheelchair moves along the global path, the host 12 plans the local motion trajectory at the current moment under the guidance of the global path based on the obstacle information detected in real time by the front radar 4 and the rear radar 5, and generates speed control commands. S5, Command execution and movement: The host 12 sends the speed control command to the motor drive board 8 through the communication module. The motor drive board 8 drives the corresponding wheelchair motor 9 to rotate, and the wheelchair moves through the differential drive rear wheel 13 until it reaches the target point.

[0031] Preferably, in step S2, the specific method for constructing an environmental map by fusing dual radar data on the host is as follows: The host 12 will collect the first point cloud data from the front radar 4 The second point cloud data collected by the rear radar 5 The fusion is performed, where is the two-dimensional coordinate vector of the i-th point in the first point cloud on the horizontal plane. Let be the two-dimensional coordinate vector of the j-th point in the second point cloud on the horizontal plane; The fused point cloud data is projected onto the occupying raster map, with each raster cell in the map... The occupancy probability is iteratively updated using the Bayesian update formula: ; in, This represents the grid cell in the u-th row and v-th column of the map. This represents all radar observation data from time 1 to t. This represents the posterior probability of a raster cell being occupied, derived from all historical observation data. The probability of a grid cell being occupied, based on radar observation data at the current time t, is calculated using an inverse sensor model based on radar ranging values.

[0032] Preferably, in step S2, the specific method by which the host computer 12 calculates the real-time pose of the wheelchair is as follows: Host 12 uses an iterative nearest-point algorithm to register the fused point cloud at the current moment with the fused point cloud at the previous moment, obtaining the pose change of the wheelchair; let the pose of the wheelchair at the previous moment be... ,in The x-axis is... The vertical axis is , The heading angle is used; the pose increment relative to the previous moment is obtained through registration calculation. Then the real-time position of the wheelchair at the current moment. Calculated using the following formula: ; in, This represents the displacement increment along the X-axis. This represents the displacement increment in the Y-axis direction. The change in heading angle is denoted by , and all three are obtained by solving the point cloud registration results.

[0033] Preferably, in step S3, the specific method for host 12 to generate the global path is as follows: Host 12 uses A Path planning algorithms generate paths from the wheelchair's current position. To the target point The global path, and the evaluation function of the algorithm is: ； Where n represents the grid node occupied in the grid map, and g(n) is the actual movement cost from the current position of the wheelchair to node n, calculated using the following formula: ; In the formula, K is the number of nodes traversed on the path from the starting point to node n, and X... k Let be the coordinates of the k-th node on the path. The Euclidean distance between adjacent nodes; h(n) represents the distance from node n to the target point. The heuristic cost estimation is calculated using Euclidean distance: ; in, Let n be the coordinates of node n. The coordinates of the target point are given. The algorithm maintains an open list and a closed list, iteratively searches for the node with the smallest evaluation function value until it extends to the target point, and generates a global path node sequence by backtracking the parent node.

[0034] Preferably, in step S4, the specific method for the host 12 to plan the local motion trajectory and generate speed control commands is as follows: Host 12 uses the dynamic window method for local path planning, within the speed space Medium-sampled multiple sets of linear velocity v and angular velocity Each speed group corresponds to a simulated trajectory, and the optimal speed combination is selected through an evaluation function; the evaluation function is: ; in, This is the azimuth evaluation item, used to measure the consistency between the orientation of the simulated trajectory's endpoint and the target point's direction. The calculation formula is: ; In the formula, To simulate the heading angle of the wheelchair at the end of the trajectory, The azimuth angle of the target point relative to the endpoint; This is the obstacle distance evaluation item, used to measure the minimum distance between the simulated trajectory and the nearest obstacle. If the trajectory collides with an obstacle, this item is set to 0. This is a speed evaluation item, used to encourage high-speed movement, and its calculation formula is: ; In the formula, v is the sampling linear velocity; These are the weighting coefficients. The normalization factor is used; host 12 selects the evaluation function. The speed combination with the highest value This serves as the optimal speed control command for the current moment.

[0035] Preferably, in step S5, the specific method by which the motor drive board 8 drives the wheelchair motor 9 to rotate is as follows: The host 12 will control the linear velocity in the optimal speed control command. and angular velocity The data is transmitted to the motor drive board 8 via the communication module. The motor drive board 8 converts the linear velocity and angular velocity into the target rotational speeds of the left and right wheels based on the differential drive model. The conversion formula is as follows: ; in, The target linear velocity of the left rear wheel. Let L be the target linear velocity of the right rear wheel, and L be the wheelbase between the left and right rear wheels. The motor drive board 8 uses a PID control algorithm to adjust the speed of the wheelchair motor 9, minimizing the error between the actual speed and the target speed. The PID control law is as follows: ; Where u(t) is the control quantity output by the motor drive board 8, and e(t) is the error between the target speed and the actual speed. These are the proportional coefficient, integral coefficient, and differential coefficient, respectively.

[0036] Preferably, a computer device includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the method of using the intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion.

[0037] Preferably, a computer-readable storage medium stores computer instructions that cause a computer to execute the method of using the intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion.

[0038] Example 3: like Figure 1 and Figure 2 As shown, this embodiment provides a specific assembly structure for an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion, including: This system is a highly integrated mechatronics platform. Its core lies in sensing the environment through multi-sensor fusion, making intelligent decisions and plans through a central computing unit, and ultimately controlling the actuators to achieve autonomous and safe movement.

[0039] I. System Hardware Composition and Layout: The wheelchair robot is based on a traditional wheelchair frame but has been intelligently modified. As shown in Figures 1 and 2, its hardware layout fully considers functional zoning and signal flow; Human-Computer Interaction Area: Located directly in front of the wheelchair for easy user operation. It mainly includes a touchscreen, mouse, and keyboard. The touchscreen serves as the primary interface, displaying real-time navigation maps, system status such as battery level, speed, and sensor information, and receiving user-inputted destinations or navigation commands. The mouse and keyboard are for system debugging, parameter setting, or as backup input devices.

[0040] Environmental perception module: This includes a front radar and a rear radar. The front radar is preferably a forward-facing two-dimensional LiDAR, mounted on the front support of the wheelchair, scanning horizontally forward to construct an environmental contour of more than 10° in front. It is a key sensor for achieving Simultaneous Localization and Mapping (SLAM) and obstacle detection. The rear radar is preferably a rearward-facing two-dimensional LiDAR, mounted behind the back of the wheelchair, scanning horizontally backward to perceive the rear environment. After fusion with the data from the front radar, it forms a horizontal perception field of 10°, greatly improving positioning accuracy, mapping integrity, and navigation safety in narrow spaces such as corridors and U-turns.

[0041] The core control and power supply area is mainly integrated into the space under the wheelchair seat, with a compact layout.

[0042] The main control computing unit, a high-performance embedded computer such as an industrial PC based on x86 or ARM architecture, is fixedly installed in the center of the area. The main unit runs the robot operating system ROS, which carries core algorithms such as SLAM, global path planning, and local dynamic obstacle avoidance.

[0043] A hierarchical power supply management system consists of a power supply, an air switch, and a voltage regulator. The power supply is a high-capacity lithium-ion DC battery with a rated voltage of V, serving as the system's main energy source. The air switch is connected in series in the power supply's main output circuit, acting as the main switch and short-circuit protection device for the entire electrical system. The voltage regulator includes multiple DC-DC conversion modules: the first module stabilizes the 24V output voltage of the power supply to 19V, powering the main unit and the front and rear radars; the second module converts 24V to 9V or 12V, powering the switch and other low-voltage equipment.

[0044] Communication Hub: The switch is located in this area. The front and rear radars are connected to the switch via network cables, and the host is also connected to the switch via a network cable, thereby acquiring the raw scanning data of the two radars through high-speed and stable network communication.

[0045] Drive actuators include two wheelchair motors, two motor drive boards, rear wheels, rear auxiliary wheels, and front wheels.

[0046] The two wheelchair motors are preferably high-torque brushless DC motors, which directly drive the left and right rear wheels via reduction mechanisms. The rear wheels are drive wheels.

[0047] The two motor drive boards are each connected to one of the two wheelchair motors, receiving control commands and driving the motors to operate precisely. The rear auxiliary wheels are installed near the rear wheels, providing auxiliary support and preventing tilting. The front wheels are swivel wheels, providing flexible steering.

[0048] The host computer communicates via a USB-to-CAN bus module (not shown in the diagram), which can be integrated near the host computer and connected to two motor drive boards to form a closed-loop control circuit.

[0049] II. System Workflow and Interaction Relationships: The intelligent navigation workflow in this embodiment is as follows: System power-on and initialization: Close the air switch, and the hierarchical power supply management system starts working, providing stable voltage to each module. The main unit, front and rear radars, and other equipment start up.

[0050] Environmental perception and mapping / localization: The front and rear radars continuously scan the surrounding environment, sending point cloud data to the host computer via a network switch. The host computer fuses the dual radar data in real time to calculate the wheelchair's precise pose and orientation in the current environment, and simultaneously builds or updates a high-precision map.

[0051] Task setting and path planning: Users view a map built in real time by the host computer via a touchscreen and click or input target points on the map. After receiving the target, the host computer generates an optimal coarse path from the starting point to the target point based on the current map and real-time pose.

[0052] Real-time obstacle avoidance and motion control: During movement along the global path, the host's local planner, using dynamic window method (DWA) and time elastic band (TEB) inputs from forward and backward radars and real-time detected dynamic and static obstacle information, plans a local trajectory, including linear velocity and angular velocity, that satisfies kinematic constraints and safely avoids obstacles at the current moment, guided by the global path. The host sends this speed command to the two motor drive boards via a USB-to-CAN bus module.

[0053] Command execution and movement: Based on the received speed command, the motor drive board uses a closed-loop control method, such as PID control, to drive the corresponding wheelchair motor to rotate. Through differential speed, the wheelchair can move forward, backward, turn, and rotate in place, thereby accurately tracking the planned local trajectory until it safely reaches the target point.

Claims

1. A smart wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion, characterized in that, It includes a wheelchair frame and an integrated environmental perception module, core control module, drive execution module, and human-machine interaction module; the environmental perception module includes a front radar (4) installed at the front of the frame and a rear radar (5) installed at the rear of the frame; the core control module includes a host (12) as the main control computing unit, a hierarchical power supply management system, and a network switch (10); the hierarchical power supply management system includes a power supply (11), an air switch (6) located in the main power output circuit, and a voltage regulator (7); the drive execution module includes two wheelchair motors (9) and motor drives connected to each motor. The moving plate (8) and the wheelchair motor (9) drive the rear wheel (13); the human-machine interaction module includes a touch screen (1); the front radar (4) and the rear radar (5) are connected to the network switch (10) via a network cable, and the network switch (10) is connected to the host (12) via a network cable; the host (12) is connected to the motor drive board (8) via a communication module; the power supply (11) supplies power to the host (12), the front radar (4), the rear radar (5) and the network switch (10) in sequence via an air switch (6) and a voltage regulator (7); it also includes the front wheel (15) of the omnidirectional wheel and the rear auxiliary wheel (14).

2. The intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion according to claim 1, characterized in that, The voltage regulator (7) includes a first DC-DC voltage regulator and a second DC-DC voltage regulator. The first DC-DC voltage regulator converts the output voltage of the power supply (11) into a first voltage to power the host (12), the front radar (4) and the rear radar (5). The second DC-DC voltage regulator converts the output voltage of the power supply (11) into a second voltage to power the network switch (10).

3. The intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion according to claim 1, characterized in that, The human-computer interaction module also includes a mouse (2) and a keyboard (3), which are connected to the host (12); the communication module is a USB to CAN bus communication module; the front radar (4) is a forward-facing laser radar, and the rear radar (5) is a rear-facing laser radar; the host (12) is installed at the center of the bottom of the wheelchair frame, the front radar (4) is installed on the front support of the frame, and the rear radar (5) is installed at the back of the frame; the two wheelchair motors (9) drive the left rear wheel and the right rear wheel respectively, and the two motor drive boards (8) are connected to the two wheelchair motors (9) one by one.

4. A method for using an intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion according to any one of claims 1-3, characterized in that, Includes the following steps: S1, system power-on and initialization, close the air switch (6), the hierarchical power supply management system supplies power to each module, the host (12), front radar (4) and rear radar (5) start; S2, Environmental perception and mapping localization: The front radar (4) and rear radar (5) continuously scan the surrounding environment and send the point cloud data to the host (12) through the network switch (10). The host (12) fuses the dual radar data, calculates the real-time pose of the wheelchair in the current environment, and builds or updates the environmental map. S3, Task setting and path planning: The user views the environment map built by the host (12) through the touch screen (1) and sets the target point on the touch screen (1). The host (12) generates a global path from the current location to the target point based on the current map and real-time pose. S4, Real-time obstacle avoidance and motion control: During the movement of the wheelchair along the global path, the host (12) plans the local motion trajectory at the current moment under the guidance of the global path based on the obstacle information detected in real time by the front radar (4) and the rear radar (5), and generates speed control commands. S5, Command execution and movement: The host (12) sends the speed control command to the motor drive board (8) through the communication module. The motor drive board (8) drives the corresponding wheelchair motor (9) to rotate. The wheelchair moves by driving the rear wheel (13) through differential speed until it reaches the target point.

5. The method for using a multi-sensor fusion-based intelligent wheelchair autonomous navigation and obstacle avoidance mechanism according to claim 4, characterized in that, In step S2, the specific method for constructing an environmental map by fusing dual radar data on the host is as follows: The host (12) will collect the first point cloud data from the front radar (4). The second point cloud data collected by the rear radar (5) The fusion is performed, where is the two-dimensional coordinate vector of the i-th point in the first point cloud on the horizontal plane. Let be the two-dimensional coordinate vector of the j-th point in the second point cloud on the horizontal plane; The fused point cloud data is projected onto the occupying raster map, with each raster cell in the map... The occupancy probability is iteratively updated using the Bayesian update formula: ； in, This represents the grid cell in the u-th row and v-th column of the map. This represents all radar observation data from time 1 to t. This represents the posterior probability of a raster cell being occupied, derived from all historical observation data. The probability of a grid cell being occupied, based on radar observation data at the current time t, is calculated using an inverse sensor model based on radar ranging values.

6. The method for using a multi-sensor fusion-based intelligent wheelchair autonomous navigation and obstacle avoidance mechanism according to claim 5, characterized in that, In step S2, the specific method by which the host (12) calculates the real-time pose of the wheelchair is as follows: The host (12) registers the fused point cloud at the current moment with the fused point cloud at the previous moment using the iterative nearest point algorithm to obtain the pose change of the wheelchair; let the pose of the wheelchair at the previous moment be... ,in The x-axis is... The vertical axis is , The heading angle is used; the pose increment relative to the previous moment is obtained through registration calculation. Then the real-time position of the wheelchair at the current moment. Calculated using the following formula: ； in, This represents the displacement increment in the X-axis direction. This represents the displacement increment in the Y-axis direction. The change in heading angle is denoted by , and all three are obtained by solving the point cloud registration results.

7. The method for using a multi-sensor fusion-based intelligent wheelchair autonomous navigation and obstacle avoidance mechanism according to claim 6, characterized in that, In step S3, the specific method for the host (12) to generate the global path is as follows: The host (12) adopts A Path planning algorithms generate paths from the wheelchair's current position. To the target point The global path, and the evaluation function of the algorithm is: ； Where n represents the grid node occupied in the grid map, and g(n) is the actual movement cost from the current position of the wheelchair to node n, calculated using the following formula: ； In the formula, K is the number of nodes traversed on the path from the starting point to node n, and X... k Let be the coordinates of the k-th node on the path. The Euclidean distance between adjacent nodes; h(n) represents the distance from node n to the target point. The heuristic cost estimation is calculated using Euclidean distance: ； in, Let n be the coordinates of node n. The coordinates of the target point are given. The algorithm maintains an open list and a closed list, iteratively searches for the node with the smallest evaluation function value until it extends to the target point, and generates a global path node sequence by backtracking the parent node.

8. The method for using a multi-sensor fusion-based intelligent wheelchair autonomous navigation and obstacle avoidance mechanism according to claim 7, characterized in that, In step S4, the specific method by which the host (12) plans the local motion trajectory and generates speed control commands is as follows: The host (12) uses the dynamic window method for local path planning in the speed space. Medium-sampled multiple sets of linear velocity v and angular velocity Each speed group corresponds to a simulated trajectory, and the optimal speed combination is selected through an evaluation function; the evaluation function is: ； in, This is the azimuth evaluation item, used to measure the consistency between the orientation of the simulated trajectory's endpoint and the target point's direction. The calculation formula is: ； In the formula, To simulate the heading angle of the wheelchair at the end of the trajectory, The azimuth angle of the target point relative to the endpoint; This is the obstacle distance evaluation item, used to measure the minimum distance between the simulated trajectory and the nearest obstacle. If the trajectory collides with an obstacle, this item is set to 0. This is a speed evaluation item, used to encourage high-speed movement, and its calculation formula is: ； In the formula, v is the sampling linear velocity; These are the weighting coefficients. The normalization factor is used; the host (12) selects the evaluation function. The speed combination with the highest value This serves as the optimal speed control command for the current moment.

9. The method of using the intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion according to claim 8, characterized in that, In step S5, the specific method by which the motor drive board (8) drives the wheelchair motor (9) to rotate is as follows: The host (12) will use the linear velocity in the optimal speed control command. and angular velocity The data is sent to the motor drive board (8) via the communication module. The motor drive board (8) converts the linear velocity and angular velocity into the target rotational speeds of the left and right wheels according to the differential drive model. The conversion formula is as follows: ； in, The target linear velocity of the left rear wheel. Let L be the target linear velocity of the right rear wheel, and L be the wheelbase between the left and right rear wheels. The motor drive board (8) uses a PID control algorithm to adjust the speed of the wheelchair motor (9) to minimize the error between the actual speed and the target speed. The PID control law is: ； Where u(t) is the control quantity output by the motor drive board (8), and e(t) is the error between the target speed and the actual speed. These are the proportional coefficient, integral coefficient, and differential coefficient, respectively.

10. A computer device, characterized in that, It includes a memory and a processor, which are interconnected and communicate with each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the method of using the intelligent wheelchair autonomous navigation and obstacle avoidance mechanism based on multi-sensor fusion as described in any one of claims 4 to 9.

Images