Robotic variable foot end and building interior adaptive stair climbing robot

Abstract

The application relates to a robot variable foot end and a building interior adaptive stair climbing robot, and belongs to the technical field of robots. A main rotary joint comprises a main rotary connecting arm and a first rotary power element. One end of the main rotary connecting arm is hinged to a body, and a hook claw is arranged at the other end of the main rotary connecting arm. A foot end folding joint comprises a folding connecting arm and a second rotary power element. One end of the folding connecting arm is hinged to one side of the main rotary connecting arm close to the hook claw, and a suction disc is arranged at the other end of the folding connecting arm. In the application, when the friction coefficient of a stair face of a stair step is greater than or equal to 0.5, the first rotary power element drives the main rotary connecting arm to rotate downward, so that the hook claw can tightly hold the damaged stair face to avoid unstable contact. When the friction coefficient of the stair face of the stair step is less than 0.5, the second rotary power element drives the folding connecting arm to rotate downward, and the suction disc can be adsorbed to the stair face to avoid slipping, so that the foot end can adapt to stair steps in different states.

Description

Technical Field

[0001] This invention belongs to the field of robotics technology, specifically relating to a robot with variable feet and an adaptive stair-climbing robot for building interiors. Background Technology

[0002] Currently, there are three main types of leg structures for robots: wheeled, tracked, and multi-legged. The most common is the wheeled structure. While highly efficient on flat ground, it performs poorly on stairs. Wheels cannot overcome vertical drops in steps, easily getting stuck on edges or slipping. Even with large-diameter wheels or special wheel sets, its adaptability and stability to different step sizes remain poor. Tracked structures have a large ground contact area, giving them an advantage on uneven terrain. They can also utilize the length of their tracks to "hover" over steps. However, the rigid structure of tracked robots makes it difficult to conform to the right-angle turns of stairs, resulting in inconvenient turning. At the edges of steps, contact is often point or line contact, leading to insufficient grip and a tendency to slip or spin on stairs, resulting in low climbing efficiency. Furthermore, tracked structures struggle with weight management when climbing steep stairs, easily resulting in forward or backward tilting, leading to lower robot stability. The multi-legged structure employs a biomimetic design with multiple jointed legs. The feet are equipped with flat or curved plates with friction pads, allowing for point or surface contact with the ground. By precisely controlling the movement of each leg, the feet can be placed directly in the middle of a step, providing support and enabling movement. However, when facing damaged or smooth steps, the feet are prone to instability or slippage, making it difficult for the robot to climb stairs stably. Summary of the Invention

[0003] In view of this, the present invention provides a robot with a variable foot and an adaptive stair-climbing robot for use inside buildings, in order to overcome the shortcomings of the prior art. The robot foot of the present invention can adapt to stair steps in different states, ensuring that the robot can climb stairs stably in different stair environments.

[0004] The technical solution of this invention is: a variable foot end for a robot, configured and connected to both sides of the robot's body. The variable foot end includes: a main rotary joint comprising a main rotary connecting arm and a first rotary power element; one end of the main rotary connecting arm is hinged to the robot body, and the hinge axis is parallel to the width direction of the robot body; the first rotary power element is disposed on the robot body, and its output end is connected to the main rotary connecting arm to drive the main rotary connecting arm to rotate around its hinge axis; a pawl is disposed at the other end of the main rotary connecting arm, and the tail of the pawl is hinged to the main rotary connecting arm; the hinge axes at both ends of the main rotary connecting arm are parallel to each other; when the coefficient of friction of the stair step surface is greater than or equal to 0.5, the pawl is used. The device is placed on the step surface and grips its surface. The foot-end folding joint includes a folding connecting arm and a second rotating power element. One end of the folding connecting arm is hinged to the side of the main rotating connecting arm near the hook, and the hinge axis is parallel to the width direction of the body. The second rotating power element is set on the main rotating connecting arm, and the output end of the second rotating power element is connected to the folding connecting arm to drive the folding connecting arm to rotate around its hinge axis. The suction cup is set at the other end of the folding connecting arm and is hinged to the top of the folding connecting arm. The hinge axes at both ends of the folding connecting arm are parallel to each other. The suction cup is equipped with a negative pressure element. When the friction coefficient of the stair step surface is less than 0.5, the suction cup is placed on the step surface and adsorbs and connects with it.

[0005] Preferably, the hook includes: a claw arm, a hook tip, and an elastic hinge. One end of the claw arm is hinged to the main rotating connecting arm, and the other end is hinged to the tail of the hook tip. The hinge axis is parallel to the width direction of the machine body. The elastic hinge is connected between the claw arm and the hook tip to drive the hook tip to rotate towards the vertical side closer to the stair step.

[0006] Preferably, the hook claw is made of carbon fiber composite material, and the surface of the hook tip is coated with tungsten carbide.

[0007] Preferably, the suction cup is made of polymer rubber, a buffer rubber layer is fixed on the top of the suction cup, and an annular flexible pad is fixed at the mouth of the suction cup and is coaxial with its center line.

[0008] Preferably, the main rotating connecting arm includes: two first arms and a third rotating power element. One end of each of the two first arms is hinged to the hook and the machine body respectively, and the other end is hinged to each other. The hinge axes are parallel to the width direction of the machine body respectively. The first rotating power element drives the first arm near the machine body to rotate around its hinge axis. The third rotating power element is disposed at the connection of the two first arms to drive the first arm near the hook to rotate around its hinge axis.

[0009] Preferably, the folding connecting arm includes: two second arms and a fourth rotary power element. One end of each of the two second arms is hinged to the main rotary connecting arm and the suction cup, respectively, and the other end is hinged to each other. The hinge axes are parallel to the width direction of the body. The second rotary power element drives the second arm near the main rotary connecting arm to rotate around its hinge axis. The fourth rotary power element is disposed at the connection of the two second arms to drive the second arm near the suction cup to rotate around its hinge axis.

[0010] An adaptive stair-climbing robot for building interiors includes a body and multiple variable feet. The multiple variable feet are symmetrically arranged at equal intervals on both sides of the body along its length. The end of the main rotating connecting arm of the variable feet away from the hook or suction cup is hinged to the body, and the hinge axis is parallel to the width direction of the body.

[0011] Preferably, the robot body is equipped with an environmental sensing component, which includes a lidar, an IMU, and a tactile sensor. The lidar is fixed at the front end of the robot body to scan the stair steps and obtain the friction coefficient of the step surface using the information obtained from the scan. The IMU is fixed on the robot body to collect the robot body's posture information. The tactile sensors are respectively fixed on the bottom of the hook or suction cup to collect the contact force between the hook or suction cup and the step surface when the robot walks.

[0012] Preferably, the robot body is equipped with a dynamic point cloud filter and a gait generator. The dynamic point cloud filter is connected to the LiDAR signal and constructs the robot's movement path based on the information obtained from the scan. The gait generator is connected to the LiDAR and IMU signals respectively and obtains the robot's gait based on the information obtained from the scan and the collected posture information.

[0013] Preferably, the function used by the dynamic point cloud filter to construct the robot's movement path is:

[0014] ;

[0015] in, For the actual total cost of node n, g( n Let be the path cost from the starting point to node n. It is the angle between the current node's movement direction and the staircase normal. It is the angle between the direction of movement of the previous node and the normal to the staircase, (x g y g , z g (x) represents the coordinates of the target point. n y n , z n ) represents the coordinates of the current node.

[0016] Compared with existing technologies, this invention provides a variable-foot robot and an adaptive stair-climbing robot for buildings. The variable-foot robot uses a main rotary joint, a main rotary connecting arm, a first rotary power element, and a grappling hook in cooperation. When the friction coefficient of the stair step surface is greater than or equal to 0.5, the first rotary power element drives the main rotary connecting arm to rotate downward, so that the grappling hook can firmly grasp the damaged step surface and avoid unstable contact. Then, the foot folding joint, a folding connecting arm, a second rotary power element, and a suction cup are used in cooperation. When the friction coefficient of the stair step surface is less than 0.5, the first rotary power element drives the main rotary connecting arm to move upward, while the second rotary power element drives the folding connecting arm to rotate downward, so that the suction cup can adhere to the step surface and avoid slippage. By switching between the grappling hook and the suction cup, the foot can adapt to different stair steps, thereby ensuring that the robot can climb stairs stably in different stair environments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the variable foot end of the present invention;

[0018] Figure 2 This is a schematic diagram of the hook claw of the present invention;

[0019] Figure 3 This is a schematic diagram of the structure of the stair-climbing robot of the present invention;

[0020] Figure 4 This is a flowchart illustrating the workflow of the stair-climbing robot of this invention.

[0021] Explanation of reference numerals in the attached figures:

[0022] 1. Main rotating connecting arm; 3. Folding connecting arm; 4. Claw; 5. Suction cup; 6. Material compartment; 7. Fuselage; 11. First arm; 31. Second arm; 41. Claw arm; 42. Hook tip; 43. Flexible hinge. Detailed Implementation

[0023] This invention provides a robot with variable-length feet and an adaptive stair-climbing robot for building interiors, which will be described below in conjunction with... Figures 1 to 4 The present invention is illustrated by the structural diagram shown below.

[0024] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0025] Reference Figure 1 , Figure 1 This is a schematic diagram of the variable foot structure of this embodiment. A variable foot for a robot is configured and connected to both sides of the robot's body 7. The variable foot includes: a main rotary joint including a main rotary connecting arm 1 and a first rotary power element. One end of the main rotary connecting arm 1 is hinged to the body 7, and the hinge axis is parallel to the width direction of the body 7. The first rotary power element is disposed on the body 7, and its output end is connected to the main rotary connecting arm 1 to drive the main rotary connecting arm 1 to rotate around its hinge axis. A pawl 4 is disposed on the other end of the main rotary connecting arm 1, and the tail of the pawl 4 is hinged to the main rotary connecting arm 1. The hinge axes at both ends of the main rotary connecting arm 1 are parallel to each other. When the friction coefficient of the stair step surface is greater than or equal to 0.5, the pawl is used... The claw 4 is placed on the step surface and grips its surface. The foot-end folding joint includes a folding connecting arm 3 and a second rotating power element. One end of the folding connecting arm 3 is hinged to the side of the main rotating connecting arm 1 near the claw 4, and the hinge axis is parallel to the width direction of the body 7. The second rotating power element is set on the main rotating connecting arm 1, and the output end of the second rotating power element is connected to the folding connecting arm 3 to drive the folding connecting arm 3 to rotate around its hinge axis. The suction cup 5 is set at the other end of the folding connecting arm 3. The suction cup 5 is hinged to the top of the folding connecting arm 3. The hinge axes at both ends of the folding connecting arm 3 are parallel to each other. The suction cup 5 is equipped with a negative pressure element. When the friction coefficient of the stair step surface is less than 0.5, the suction cup 5 is placed on the step surface and adsorbs and connects with it.

[0026] In this embodiment, when the friction coefficient of the stair step surface is greater than or equal to 0.5, i.e., when the step surface is damaged, the first rotary power element drives the main rotary connecting arm 1 to rotate downward (while the second rotary power element drives the folding connecting arm 3 to rotate upward and fold), so that the hook 4 steps on the step surface. The hook 4 improves the grip of the foot and prevents the foot from shaking when stepping on the damaged step surface. When the friction coefficient of the stair step surface is less than 0.5 (such as tile or metal stairs), the second rotary power element drives the folding connecting arm 3 to flip downward so that the suction cup 5 steps on the step surface. This can prevent the foot from slipping on the smooth (wet) step surface. The main rotary joint and the foot folding joint are used to switch between the hook 4 and the suction cup 5, so that the foot can adapt to the step surface with different friction states, enabling the robot to climb stairs stably. This allows the robot to transport vertical materials in high-rise buildings when there is no elevator or the elevator is malfunctioning.

[0027] Specifically, in this embodiment, the negative pressure element is a miniature negative pressure vacuum pump, which extracts air from the suction cup 5 to create a negative pressure inside, thereby achieving a stable connection between the suction cup 5 and the step surface. When the foot needs to move to the next step, the miniature negative pressure vacuum pump runs in reverse to release the negative pressure in the suction cup 5, thereby allowing the suction cup 5 to quickly separate from the stair step surface.

[0028] Reference Figure 2 , Figure 2 This is a schematic diagram of the hook claw structure in this embodiment. As a further optimization, the hook claw 4 in this embodiment includes: a claw arm 41, a hook tip 42, and an elastic hinge 43. One end of the claw arm 41 is hinged to the main rotating connecting arm 1, and the other end is hinged to the tail of the hook tip 42. The hinge axis is parallel to the width direction of the body 7. The elastic hinge 43 is connected between the claw arm 41 and the hook tip 42 to drive the hook tip 42 to rotate towards the vertical side closer to the stair step.

[0029] In this embodiment, a claw 4 is formed by a claw arm 41, a hook tip 42, and an elastic hinge 43. When the claw 4 steps on the stair step, the weight of the body 7 and the foot causes the hook tip 42 to flip upward relative to the claw arm 41. The elastic hinge 43 enables the claw 4 to fit tightly against the step surface, and the hook tip 42 can be inserted into the grooves and gaps of the broken step. This not only improves the grip of the claw 4 on the broken step surface, but also improves the cushioning of the foot landing, making the robot's climbing and walking more stable. When the foot is lifted, the elastic hinge 43 causes the hook tip 42 to flip downward and return to its original position.

[0030] Specifically, the flexible hinge 43 is made of alloy material, which improves its service life.

[0031] In addition, the arc-shaped structure formed by the claw arm 41 and the hook tip 42 can be stepped on the side of the step surface near the edge. The elastic hinge 43 allows the hook tip 42 to automatically adjust its angle, so that the hook 4 is locked at the junction of the step surface and the vertical surface of the stair step, in order to adapt to the edge of the step with different heights, thicknesses and shapes, and further improve the stability of the foot.

[0032] As a further optimization, in this embodiment, the hook 4 is made of carbon fiber composite material, and the surface of the hook tip 42 is coated with tungsten carbide.

[0033] In this embodiment, the carbon fiber composite material hook 4 has the characteristics of lightweight, high strength and high wear resistance, which can enhance the grip and its own strength. At the same time, the tungsten carbide coating makes the hook tip 42 have strong wear resistance, further preventing the hook 4 from shaking when stepping on the step surface.

[0034] As a further optimization, in this embodiment, the suction cup 5 is made of polymer rubber, a buffer rubber layer is fixed on the top of the suction cup 5, and an annular flexible pad is fixed at the opening of the suction cup 5 and is coaxial with its center line.

[0035] In this embodiment, the high-temperature resistance, anti-aging, and anti-slip adsorption properties of the polymer rubber enable the suction cup 5 to have sufficient strength and friction performance. When the suction cup 5 is stepped on the step surface, it maintains a stable shape, allowing the negative pressure element to stably fix the suction cup on the step surface and avoid slippage. The buffer rubber layer on the top of the suction cup 5 can provide cushioning when the foot steps on the step surface, and the annular flexible pad can fit tightly with the step surface, adjusting the fit to minor unevenness on the step surface, further improving the stability of the suction cup adsorbed on the step surface.

[0036] As a further optimization, in this embodiment, the main rotating connecting arm 1 includes two first arms 11 and a third rotating power element. One end of each of the two first arms 11 is hinged to the hook 4 and the body 7 respectively, and the other end is hinged to each other. The hinge axes are parallel to the width direction of the body 7 respectively. The first rotating power element drives the first arm 11 near the body 7 to rotate around its hinge axis. The third rotating power element is set at the connection of the two first arms 11 to drive the first arm 11 near the hook 4 to rotate around its hinge axis.

[0037] In this embodiment, the main rotating connecting arm 1, which is composed of two first arms 11 and a third rotating power element, can drive the two first arms 11 to rotate using the first and third rotating power elements. This allows the position of the hook 4 to be flexibly adjusted according to the width of the step surface and the height of the step, so that the hook 4 can adapt to different stair environments.

[0038] In addition, the main rotating connecting arm 1 of this structure can also adjust the position and height of the folding connecting arm 3 and the suction cup 5 by using the independent rotation of the two first arms 11, further improving the flexibility of the robot's foot climbing stairs.

[0039] As a further optimization, the folding connecting arm 3 in this embodiment includes two second arms 31 and a fourth rotational power element. One end of each of the two second arms 31 is hinged to the main rotating connecting arm 1 and the suction cup 5, respectively, and the other end is hinged to each other. The hinge axes are parallel to the width direction of the body 7. The second rotational power element drives the second arm 31 near the main rotating connecting arm 1 to rotate around its hinge axis. The fourth rotational power element is set at the connection of the two second arms 31 to drive the second arm 31 near the suction cup 5 to rotate around its hinge axis.

[0040] In this embodiment, the folding connecting arm 3, which is composed of two second arms 31 and a fourth rotating power element, can not only flexibly adjust the position of the suction cup 5 according to the width of the step surface and the height of the step, so that the suction cup 5 can adapt to different stair environments, but also improve the folding effect of the folding connecting arm 3 when using the hook 4, and avoid the folding connecting arm 3 and the suction cup 5 affecting the walking of the feet when using the hook 4.

[0041] Specifically, in the above embodiments, the first, second, third, and fourth rotating power elements of the variable foot end adopt harmonic geared motors. These harmonic geared motors have high peak torque and light weight, which can effectively drive the robot to climb stairs. Furthermore, by reducing energy loss through efficient transmission, the harmonic geared motor can provide 3 to 5 times more torque than ordinary motors of the same size, ensuring that the robot can climb stairs stably.

[0042] The reduction structure of a harmonic geared motor mainly consists of the following three parts:

[0043] Rigid wheel: outer ring gear, fixed and stationary;

[0044] Flexible gear: Inner ring gear, thin-walled structure, deforms with wave action;

[0045] Wave generator: An elliptical cam structure drives a flexible wheel to generate waves, thereby achieving speed reduction transmission.

[0046] Traditional planetary reducers typically consist of multiple gears, while harmonic reducers comprise only three core components, reducing overall weight by approximately 30%, effectively lowering robot load and improving energy efficiency. Due to its backlash-free transmission characteristics, the harmonic reducer structure provides angular control accuracy at the 0.01° level, ensuring precise adjustment of the robot's foot contact angle when walking on stairs, preventing slippage. Traditional gear drives suffer from frictional losses, while harmonic drives boast a mechanical efficiency exceeding 80%, reducing energy waste and extending robot endurance.

[0047] Reference Figure 3 , Figure 3 This is a schematic diagram of the structure of the stair-climbing robot of this embodiment. An adaptive stair-climbing robot for building interiors includes a body 7 and multiple variable feet. The multiple variable feet are symmetrically arranged at equal intervals on both sides of the body 7 along its length direction. The end of the main rotating connecting arm 1 of the variable feet away from the hook 4 or suction cup 5 is hinged to the body 7, and the hinge axis is parallel to the width direction of the body 7.

[0048] In this embodiment, the building interior adaptive stair-climbing robot performs execution control through intelligent decision-making, realizing energy management, foot mode selection, material management, and joint control.

[0049] In addition, when the stair-climbing robot moves down stairs or down slopes, it can convert potential energy into electrical energy, thereby improving its battery life.

[0050] Regenerative braking: When the robot descends the stairs, its feet touch the ground, and the harmonic geared motor is in a passive rotation state. Traditional motors would need to consume energy for active control in this situation, but this system utilizes the principle of motor braking to allow the motor to generate electricity in reverse, recovering gravitational potential energy and converting it into electrical energy for storage.

[0051] Supercapacitor energy storage: Robots may generate a large amount of energy recovery in a short period of time, such as when rapidly descending from a high floor. Supercapacitors are used to store the energy recovered in a short time, avoiding battery overload during instantaneous charging and improving energy utilization.

[0052] Intelligent energy allocation: The robot monitors motor load, battery level, and current task requirements in real time, and determines the energy allocation through an intelligent decision management system. When the battery level is low, it prioritizes storing recovered electrical energy into the battery to extend the driving time. When the battery is fully charged, it automatically switches to supercapacitor storage to reduce energy waste. It also manages energy according to the movement status: downhill / braking and climbing / flat road. In addition, it monitors the safety of the supercapacitor.

[0053] In this embodiment, multiple variable feet are symmetrically arranged on both sides of the robot's body 7, enabling the robot to climb stairs in different stair environments, improving the robot's stability during stair climbing, and thus facilitating vertical material transportation in high-rise buildings where there are no elevators or the elevators are malfunctioning.

[0054] Reference Figure 4 , Figure 4 The flowchart of the stair-climbing robot is shown below. As a further optimization, an environmental perception component is provided on the body 7 in this embodiment. The environmental perception component includes a lidar, an IMU, and a tactile sensor. The lidar is fixed at the front end of the body 7 to scan the stair steps and obtain the friction coefficient of the step surface using the information obtained from the scan. The IMU is fixed on the body 7 to collect the attitude information of the body 7. The tactile sensors are respectively fixed on the bottom of the hook 4 or the suction cup 5 to collect the contact force between the hook 4 or the suction cup 5 and the step surface when the robot walks.

[0055] In this embodiment, the staircase is scanned by a LiDAR, and the scanned data is processed by an AI algorithm to obtain the friction system of the staircase steps. The robot performs terrain scanning during the climbing process and switches between the hook 4 or the suction cup 5 based on the friction coefficient of the step surface obtained from the terrain scanning. During the climbing process, the IMU can collect the posture information of the robot body 7 to monitor the posture of the climbing robot. Combined with the tactile sensor to collect the contact force between the hook 4 or the suction cup 5 and the step surface, the robot performs contact force detection to monitor the gripping force of the feet in real time. If the feet are detected to be loose from the step surface, the clamping force of the hook will be adjusted to increase the gripping force or the negative pressure of the suction cup 5 will be adjusted.

[0056] Specifically, when it is necessary to adjust the gripping force of the claw 4, a micro linear drive element can be set between the claw arm 41 and the hook tip 42. The micro linear drive element is used to drive the hook tip 42 to flip to the ground side, thereby further improving the gripping force of the claw 4.

[0057] Specifically, the lidar in this embodiment uses a 905nm wavelength, which can penetrate interfering environments such as dense smoke and dust, significantly improving the accuracy of modeling. The lidar collects real-time 3D point cloud data of the stairwell environment for environmental modeling and navigation.

[0058] Compared to other wavelengths of laser light, 905nm laser light has strong penetrating power and performs exceptionally well in penetrating dense smoke and dust. This allows lidar to acquire clear 3D point cloud data even in smoky or dusty environments. LiDAR can collect 3D data of the surrounding environment in real time and generate accurate environmental models. This is crucial for robot navigation in complex stairwell environments, helping to plan safe and efficient travel paths. LiDAR can also operate stably in dynamically changing environments, reducing data errors caused by environmental changes.

[0059] In this embodiment, the use of LiDAR effectively solves the problem of traditional visual sensors failing under low visibility conditions such as smoke and dust, ensuring the stability and safety of the robot in various complex stairwell environments.

[0060] As a further optimization, in this embodiment, the body 7 is equipped with a dynamic point cloud filter and a gait generator. The dynamic point cloud filter is connected to the LiDAR signal and constructs the robot's movement path based on the information obtained from the scan. The gait generator is connected to the LiDAR and IMU signals respectively and obtains the robot's gait when moving based on the information obtained from the scan and the collected posture information.

[0061] In this embodiment, a dynamic point cloud filter is connected to the LiDAR signal. Based on the information acquired through scanning, a robot movement path is constructed. The dynamic point cloud filter, based on a 2.5D elevation map, dynamically analyzes the point cloud data to remove the influence of moving obstacles in real time, further improving modeling accuracy and ensuring precise robot navigation in complex environments. A 2.5D elevation map is a way to represent three-dimensional terrain on a two-dimensional plane, where each pixel contains a height value representing the vertical height of that point relative to a reference plane. This representation simplifies data processing and storage requirements while preserving terrain undulation information.

[0062] In this embodiment, the stair-climbing robot is equipped with a dynamic gait generator. The gait generator is pre-trained using a reinforcement learning algorithm to generate x different gaits, including special gaits for handling 45° turns and navigating narrow 70cm stairs. During operation, the gait generator can adjust in real time based on environmental feedback to optimize the robot's walking efficiency and stability.

[0063] The working process of a dynamic point cloud filter:

[0064] 1. Point cloud data acquisition: The robot acquires point cloud data of the surrounding environment in real time through LiDAR or other sensors, forming a high-density three-dimensional point cloud set;

[0065] 2. Elevation map construction: The collected point cloud data is converted into a 2.5D elevation map, generating a two-dimensional map containing elevation information to reflect the terrain undulations of the environment;

[0066] 3. Dynamic Analysis and Filtering: Dynamic analysis of point cloud data using elevation maps is performed to identify points with significant differences from the ground elevation. These points usually represent moving obstacles. Through filtering algorithms, these dynamic obstacle points are removed, while static environmental features are preserved.

[0067] 4. Environment Modeling and Navigation: The processed point cloud data is used to build an accurate environment model for use by the navigation algorithm, ensuring that the robot can accurately locate and plan its path in complex environments.

[0068] Dynamic point cloud filters have the following advantages:

[0069] By combining with a 2.5D elevation map, the dynamic point cloud filter can effectively identify and remove moving obstacles, improving the accuracy of environmental modeling and reducing the interference of dynamic obstacles on the navigation system.

[0070] Real-time performance and adaptability: This dynamic point cloud filter can process point cloud data in real time, adapt to environmental changes, and ensure that the robot maintains efficient and stable navigation performance in dynamic environments.

[0071] Simplified data representation: The terrain is represented by a 2.5D elevation map, which reduces the complexity of data processing and improves the system's response speed and reliability.

[0072] The traditional robot path evaluation function is: ;

[0073] in, f ( n ) is the total cost of evaluating node n. g ( n ) represents the path cost (along the path length) from the starting point to node n. h ( n The heuristic cost estimate from the current node n to the target must satisfy acceptability, i.e. h ( n )≤True residual cost.

[0074] The path evaluation function mentioned above frequently turns in the horizontal and vertical directions, increasing energy consumption and time. In stairwell environments, the priority of vertical movement (such as going up and down stairs) is not optimized separately, resulting in energy waste and long gait adjustment time caused by frequent robot turns.

[0075] As a further optimization, the function used by the dynamic point cloud filter to construct the robot's movement path in this embodiment is:

[0076] ;

[0077] in, Let n be the actual total cost of evaluation. g ( n Let be the path cost from the starting point to node n. It is the angle between the current node's movement direction and the staircase normal. It is the angle between the direction of movement of the previous node and the normal to the staircase, (x g y g , z g (x) represents the coordinates of the target point. n y n , z n ) represents the coordinates of the current node.

[0078] In this embodiment, to reduce sudden changes in path direction, the path cost... g ( n Introducing an angle penalty term for optimization :

[0079] ;

[0080] in, It is the angle between the current node's movement direction and the staircase normal. It is the angle between the direction of movement of the previous node and the normal to the staircase. The angle penalty coefficient can be set to 0.8.

[0081] In this embodiment, by introducing an angle penalty term to optimize the directional abrupt changes of adjacent nodes in the path, the algorithm is guided to generate a smooth path, which can reduce the energy waste and gait adjustment time caused by the robot's frequent turning.

[0082] Vertical movement (going up and down steps) in a stairwell environment has a significant impact on path efficiency. The improved heuristic function is divided into horizontal and vertical components:

[0083] ;

[0084] The vertical weight can be set to 1.5.

[0085] The heuristic function in this embodiment can increase the priority of vertical (Z-axis) movement, prioritize the shortest path up and down stairs, and reduce path redundancy caused by horizontal detours (such as detours when avoiding a single stair).

[0086] The improved path evaluation function is:

[0087] ;

[0088] ;

[0089] For the actual cost, This is a heuristic cost.

[0090] The movement path construction process in the above embodiments is as follows:

[0091] 1. Environmental Modeling: A 2.5D elevation map of the stairs is constructed using LiDAR, with each grid including coordinates (x, y, z). A dynamic point cloud filter is used to remove moving obstacles to ensure the accuracy of the map's static features.

[0092] 2. Node expansion rules: Each node is allowed to move in 8 directions (4 horizontal directions and 4 directions for moving up and down stairs), and moving up and down stairs must meet physical constraints;

[0093] 3. Priority queue management: Use a min-heap (priority queue) to manage priorities. Sort, expand each time The smallest node; maintain the visited nodes recorded in the closed table to avoid duplicate calculations;

[0094] 4. Path backtracking and optimization: After reaching the target node, the original path is generated by backtracking through the parent node pointer; B-spline curve smoothing technology is applied to further reduce the path turning points.

[0095] The stair-climbing robot system of this invention is equipped with a stair feature library. The robot has a built-in parameter library of n types of standard staircases, including different step heights (15-25cm) and tread widths (20-30cm). It can automatically adjust its gait pattern according to the actual situation of the staircase, ensuring smooth passage on different types of staircases. Through the built-in standard staircase parameter library, the robot can perceive the specific dimensions of the current staircase in real time and dynamically adjust gait parameters, such as step length, step height, and step frequency, to achieve adaptation to diverse staircase environments. Combined with sensor data, the robot can measure the actual dimensions of the staircase in real time and compare them with the built-in standard staircase parameters. This real-time perception and feedback mechanism enables the robot to react quickly and adjust its gait in complex and changing staircase environments, avoiding imbalance or falls caused by size differences.

[0096] By incorporating multiple built-in stair parameters, the robot achieves stair parameter matching, enabling it to adapt not only to common standard staircases but also to non-standard staircases with parameters falling within a given range through interpolation or extrapolation. This significantly improves the robot's environmental adaptability and expands its application scope. Combined with an improved path planning function, the robot can efficiently calculate the optimal path in stairwell environments, avoiding unnecessary detours. The built-in stair parameter library provides accurate environmental information for path planning, further enhancing planning efficiency and walking stability.

[0097] Additionally, physical simulation platforms (such as Gazebo and MuJoCo) can be used to create realistic staircase environments, allowing robots to train repeatedly in a virtual environment. These environments include: 45° corner stairs (common in emergency stairwells), 70cm narrow stairs (simulating stairs in old or special buildings), and irregular steps (with randomly set height and width to enhance generalization ability). In this simulation environment, the robot can safely conduct numerous gait trials, gradually learning how to walk stably on different types of stairs.

[0098] For example, this invention uses the Proximal Policy Optimization (PPO) algorithm for training. PPO continuously optimizes the gait strategy through a policy network, enabling the robot to gradually learn how to adjust its gait to adapt to different stair environments. A random gait strategy is initialized, and the robot randomly tries different gaits in the stair environment (such as large strides, small steps, and changes in foot grip). After the robot executes a gait, the system records key parameters such as stability, energy consumption, and speed. Different reward values ​​are given based on whether the robot smoothly traverses the stairs, falls, or consumes excessive energy. The PPO algorithm uses a gradient update strategy to enable the robot to gradually learn the optimal gait and achieve gait planning until the robot can autonomously adapt to different stair environments.

[0099] The robot employs a 45° turning gait, learning the appropriate leg swing angle through trial and error training to ensure smooth passage through corners. The tilt angle of the robot body (7) is monitored in real-time via an IMU, adjusting the gait when entering a corner. A reward mechanism is used during training to encourage the robot to adopt a more stable gait (such as reducing stride frequency and adjusting the center of gravity). A typical quadruped robot gait may not be suitable for stairs with only 70cm steps, easily leading to missteps or excessive stride length causing instability. Therefore, the robot continuously adjusts its stride length during training to find a gait pattern suitable for narrow stairs. Tactile sensors detect the edges of the stair steps to prevent missteps. The robot undergoes millions of training iterations in a simulation environment to optimize its gait, enabling it to walk smoothly on narrow 70cm stairs.

[0100] In actual operation, the stair-climbing robot not only relies on pre-trained gait strategies, but also makes adjustments based on real-time environmental feedback.

[0101] 1. Environmental perception and state estimation

[0102] The robot's multimodal environmental perception components monitor the surrounding environment in real time.

[0103] LiDAR: Scans the shape of the staircase to obtain information such as step height, width, and slope;

[0104] Inertial Measurement Unit (IMU): Detects changes in robot posture and prevents instability;

[0105] Tactile sensor: Monitors the contact force at the foot end to determine whether the current foot grip and gait are suitable for the current stair environment.

[0106] 2. Adaptive Gait Planning

[0107] The robot adjusts its gait in real time based on data collected by its environmental perception components, including: stride length adjustment (automatically shortening the stride to prevent slipping if a narrowing stair tread is detected); cadence optimization (increasing the stride frequency to improve walking stability if an excessive tilt angle is detected (e.g., on a corner staircase); and foot mode selection (automatically switching to rubber suction cup mode for enhanced grip on low-friction surfaces like tiled stairs, and to grappling hooks for improved stability on high-friction surfaces).

[0108] 3. Reinforcement learning strategies are updated online.

[0109] During its stair-climbing operation, the robot not only uses pre-trained gaits but also performs online fine-tuning, calculating reward values ​​in real time to determine if the current gait is optimal. By combining dynamic programming, gait parameters are optimized during execution. A local fine-tuning strategy is employed to continuously improve the robot's adaptability.

[0110] In the above embodiment, a material compartment 6 is provided on the top of the fuselage 7 to facilitate the storage of materials to be transported. The material compartment 6 is equipped with an electromagnetic locking mechanism to realize the detachable connection between the material compartment 6 and the fuselage 7. The material compartment 6 has the function of constant temperature and shock resistance protection to realize material management.

[0111] The above-disclosed embodiments are merely preferred embodiments of the present invention. However, the embodiments of the present invention are not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A robot variable foot end configured to be connected to both sides of a body (7) of a robot, characterized in that, The variable foot end comprises: a main rotary joint, comprising a main rotary connecting arm (1) and a first rotary power element, one end of the main rotary connecting arm (1) is hinged to the fuselage (7), and the hinging axes are parallel to the width direction of the fuselage (7), the first rotary power element is arranged on the fuselage (7), and the output end of the first rotary power element is connected to the main rotary connecting arm (1) to drive the main rotary connecting arm (1) to rotate around the hinging axis thereof; a hook claw (4) arranged at the other end of the main rotary connecting arm (1), the tail of the hook claw (4) is hinged to the main rotary connecting arm (1), the hinging axes of the two ends of the main rotary connecting arm (1) are parallel to each other, when the friction coefficient of the stair step surface is greater than or equal to 0.5, the hook claw (4) is placed on the stair step surface and clutches the surface thereof; a foot end folding joint, comprising a folding connecting arm (3) and a second rotary power element, one end of the folding connecting arm (3) is hinged to the side of the main rotary connecting arm (1) close to the hook claw (4), and the hinging axes are parallel to the width direction of the fuselage (7), the second rotary power element is arranged on the main rotary connecting arm (1), and the output end of the second rotary power element is connected to the folding connecting arm (3) to drive the folding connecting arm (3) to rotate around the hinging axis thereof; a suction disc (5) arranged at the other end of the folding connecting arm (3), the top of the suction disc (5) is hinged to the folding connecting arm (3), the hinging axes of the two ends of the folding connecting arm (3) are parallel to each other, and a negative pressure element is arranged in the suction disc (5), when the friction coefficient of the stair step surface is less than 0.5, the suction disc (5) is placed on the stair step surface and is adsorbed and connected to the stair step surface.

2. The robotic variable foot end of claim 1, wherein, The hook claw (4) comprises a claw arm (41), a hook tip (42) and an elastic hinge (43), one end of the claw arm (41) is hinged to the main rotary connecting arm (1), the other end is hinged to the tail of the hook tip (42), and the hinging axes are parallel to the width direction of the fuselage (7), the elastic hinge (43) is connected between the claw arm (41) and the hook tip (42) to drive the hook tip (42) to rotate towards the side close to the stair step surface.

3. The robotic variable foot end of claim 2, wherein, The material of the hook claw (4) is carbon fiber composite material, and the surface of the hook tip (42) is provided with a tungsten carbide coating.

4. The robotic variable foot end of claim 1, wherein, The material of the suction disc (5) is high polymer rubber, a buffer rubber layer is fixedly arranged on the top of the suction disc (5), and a ring-shaped flexible pad is fixedly arranged on the mouth of the suction disc (5) and coaxial with the center line thereof.

5. The robotic variable foot end of claim 1, wherein, The main rotary connecting arm (1) comprises two first branch arms (11) and a third rotary power element, one end of each of the two first branch arms (11) is hinged to the hook claw (4) and the fuselage (7) respectively, the other ends are hinged to each other, and the hinging axes are parallel to the width direction of the fuselage (7) respectively, the first rotary power element drives the first branch arm (11) close to the fuselage (7) to rotate around the hinging axis thereof, and the third rotary power element is arranged at the connection position of the two first branch arms (11) to drive the first branch arm (11) close to the hook claw (4) to rotate around the hinging axis thereof.

6. The robotic variable foot end of claim 1, wherein, The folding connecting arm (3) comprises two second branch arms (31) and a fourth rotating power element, one end of each of the two second branch arms (31) is hingedly connected with the main rotating connecting arm (1) and the suction cup (5), the other end is hingedly connected with each other, and the hinging shafts are parallel to each other in the width direction of the fuselage (7), the second rotating power element drives the second branch arm (31) close to the main rotating connecting arm (1) to rotate around the hinging shaft thereof, and the fourth rotating power element is arranged at the connection of the two second branch arms (31) to drive the second branch arm (31) close to the suction cup (5) to rotate around the hinging shaft thereof.

7. A building interior adaptive stair climbing robot comprising: The fuselage (7) further comprises a plurality of variable foot ends according to claim 1, the plurality of variable foot ends are symmetrically arranged on both sides of the fuselage (7) at equal intervals along the length direction of the fuselage (7), and one end of the main rotating connecting arm (1) of the variable foot end away from the hook claw (4) or the suction cup (5) is hingedly connected with the fuselage (7), and the hinging shaft is parallel to the width direction of the fuselage (7).

8. The building interior adaptive stair climbing robot of claim 7, wherein, An environment sensing assembly is arranged on the fuselage (7), the environment sensing assembly comprises a laser radar, an IMU and a tactile sensor, the laser radar is fixedly arranged at the front end of the fuselage (7) to scan the stair step, the friction coefficient of the step surface is obtained by using the information obtained by scanning, the IMU is fixedly arranged on the fuselage (7) to collect the attitude information of the fuselage (7), and the tactile sensor is fixedly arranged at the bottom of the hook claw (4) or the suction cup (5) to collect the contact force between the hook claw (4) or the suction cup (5) and the step surface when the robot walks.

9. The building interior self-adapting climbing robot according to claim 8, characterized in that, A dynamic point cloud filter and a gait generator are arranged on the fuselage (7), the dynamic point cloud filter is connected with the laser radar signal, the moving path of the robot is constructed according to the information obtained by scanning, the gait generator is connected with the laser radar and the IMU signal respectively, and the gait when the robot moves is obtained according to the information obtained by scanning and the collected attitude information.

10. The building interior self-adapting climbing robot according to claim 9, characterized in that, The function used by the dynamic point cloud filter when constructing the moving path of the robot is: ； wherein, is the actual evaluation total cost of the node n, g ( n ) is the path cost from the start point to the node n, is the angle between the current node moving direction and the stair normal, is the angle between the previous node moving direction and the stair normal, (x g , y g , z g ) is the target point coordinate, (x n , y n , z n ) is the current node coordinate.

Images