Wheel-foot type robot leg device based on reinforcement learning algorithm

Abstract

The invention discloses a wheel-foot type robot leg device based on a reinforcement learning algorithm, and belongs to the technical field of artificial intelligence and mechatronics. The device comprises a frame assembly, a shank motor assembly, a thigh plate assembly, a clutch assembly, a rotating wheel assembly, a shifting fork assembly, a clutch driving assembly, a suspension floating assembly, a thigh motor assembly and a shank assembly. Wherein the shank motor assembly penetrates through the frame assembly, the thigh plate assembly and the clutch assembly; the rotating wheel assembly is mounted between the shank assemblies; the shifting fork assembly is installed on the clutch assembly. The clutch driving assembly is mounted above the frame assembly; the suspension floating assembly is installed below the rear portion of the frame assembly. The thigh motor assembly is installed on the side of the frame assembly. The shank assembly and the clutch assembly jointly form a crank rocker mechanism. Three motion modes of wheel-foot switching, wheel-type motion and foot-type motion can be realized, and self-adaptive switching of the motion modes is realized according to different terrain environments in combination with a reinforcement learning algorithm, so that the motion flexibility and the environment self-adaptive capability of the robot under complex terrains are improved.

Description

Technical Field

[0001] This invention relates to a leg device for a wheel-footed robot based on a reinforcement learning algorithm, belonging to the fields of artificial intelligence and mechatronics. Background Technology

[0002] With the rapid development of mobile robot technology, wheeled-legged robots have gained widespread attention in fields such as complex terrain inspection, disaster relief, and field operations due to their combination of the high-speed and efficient movement capabilities of wheeled robots and the strong mobility and terrain adaptability of legged robots. Compared to single-wheeled or legged robots, wheeled-legged robots can flexibly transition between flat ground and unstructured complex terrain by switching between structural or motion modes, and have high application potential.

[0003] Traditional wheeled robot leg devices generally suffer from structural design flaws such as complex mechanisms, large size, power redundancy, and high control coupling, which limit their application in miniaturized, lightweight, and highly maneuverable scenarios. Therefore, a wheeled robot leg device with a compact structure, diverse motion modes, and the ability to adaptively switch motion modes by combining reinforcement learning algorithms has been developed to improve the motion flexibility, environmental adaptability, and overall motion performance of wheeled robots in complex terrain environments. Summary of the Invention

[0004] This invention provides a wheel-legged robot leg device based on reinforcement learning algorithms. This device centers on combined wheel-leg motion, and through integrated and modular design of the leg mechanism, wheel drive and leg support functions are achieved collaboratively within the same leg structure. The overall structure is compact, the transmission path is clear, and the degree of freedom is rationally configured. This leg device can achieve wheel rolling, leg support, and smooth switching between the two using only two motors within a limited space, significantly reducing the problems of multiple mechanism superposition and power redundancy common in traditional wheel-legged robot leg devices. Simultaneously, it can combine reinforcement learning algorithms to optimize motion patterns under different terrain conditions, exhibiting both good terrain adaptability and motion stability, effectively improving the overall performance of the wheel-legged robot.

[0005] The technical solution of this invention is: a leg device for a wheel-legged robot based on a reinforcement learning algorithm, comprising a frame assembly 1, a lower leg motor assembly 2, a thigh plate assembly 3, a clutch assembly 4, a rotating wheel assembly 5, a shift fork assembly 6, a clutch drive assembly 7, a suspension floating assembly 8, a thigh motor assembly 9, and a lower leg assembly 10; the lower leg motor assembly 2 is installed on the left side of the frame assembly 1 by an internal hexagonal flathead screw, and passes through the frame assembly 1, the thigh plate assembly 3, and the clutch assembly 4; the lower leg motor assembly 2 cooperates with the clutch assembly 4 and the shift fork assembly 6 to selectively transmit the power of the lower leg motor 2-2 to the lower leg 10-2 or the rotating wheel 5-1; the rotating wheel assembly 5 is installed between the lower leg assemblies 10, providing stable support for the lower leg assemblies 10; the shift fork assembly 6 is installed on the first sliding coupling 4-3. The first sliding coupling 4-3 moves left and right, thus determining whether it engages with the second sliding coupling 4-2, thereby disengaging and engaging the leg movement power. The clutch drive assembly 7 is mounted on the upper side plate 1-3 of the frame assembly 1 via hexagonal flat head screws, providing the fork assembly 6 with the power to move left and right. The suspension floating assembly 8 is mounted on the right rear of the left side plate 1-2 of the frame assembly 1 via hexagonal cylindrical head screws, providing suspension and multi-directional elasticity for the lower leg 10-2 in wheel-like motion. The thigh motor assembly 9 is mounted on the left side plate 1-2 of the frame assembly 1 via hexagonal flat head screws, providing power to the thigh plate 3-2 via keyless synchronous pulley 9-1, thigh synchronous pulley 3-3, and synchronous belt. The lower leg assembly 10 and the clutch assembly 4 are connected by a pin, forming a crank-rocker mechanism.

[0006] Specifically, the frame assembly 1 includes a first circular flange single-row bearing 1-1, a left side plate 1-2, an upper side plate 1-3, a limiting support plate 1-4, a right side plate 1-5, and a support rod 1-6; two first circular flange single-row bearings 1-1 are respectively installed on the left side plate 1-2 and the right side plate 1-5 using hexagon head screws; the limiting support plate 1-4 is installed between the left side plate 1-2 and the right side plate 1-5 using hexagon head screws; the left side plate 1-2, the upper side plate 1-3, and the right side plate 1-5 are spliced ​​and fixed together using hexagon head screws; the support rod 1-6 is installed between the left side plate 1-2 and the right side plate 1-5 using hexagon head screws; that is, the first circular flange The single-row bearing 1-1, left side plate 1-2, upper side plate 1-3, limiting support plate 1-4, right side plate 1-5, and support rod 1-6 are all fixedly connected by screws; the calf motor assembly 2 and thigh motor assembly 9 are respectively installed at the same height on the left side plate 1-2 by hexagonal flathead screws; the guide rod 6-2 in the shift fork assembly 6 is installed between the left side plate 1-2 and the right side plate 1-5 by hexagonal flathead screws; the clutch drive assembly 7 is installed on the upper side plate 1-3 by hexagonal flathead screws; and the suspension floating assembly 8 is installed on the left side plate 1-2 by hexagonal cylindrical head screws, and its installation position is lower than that of the calf motor assembly 2 and thigh motor assembly 9.

[0007] Specifically, the calf motor assembly 2 includes a calf motor adapter plate 2-1, a calf motor 2-2, a first spacer 2-3, a second spacer 2-4, a third spacer 2-5, a first rotating wheel synchronous pulley 2-6, a fourth spacer 2-7, and a calf motor drive shaft 2-8. The calf motor adapter plate 2-1 is connected to the calf motor 2-2 via countersunk hexagonal screws. The first spacer 2-3 is connected to the side of the calf motor adapter plate 2-1 away from the calf motor 2-2 via flat-head hexagonal screws. The second spacer 2-4, the third spacer 2-5, and the fourth spacer 2-7 are coaxially arranged with the calf motor drive shaft 2-8 and are sleeved on the calf motor drive shaft 2-8. The calf motor drive shaft 2-8 is connected to the calf motor 2-2 via countersunk Phillips head screws. The first rotating wheel synchronous pulley 2-6 is sleeved on the calf motor drive shaft 2-8. That is, the calf motor adapter plate 2-1... -1. The calf motor 2-2 and the calf motor drive shaft 2-8 are all fixedly connected by screws. The second spacer 2-4, the third spacer 2-5, the first rotating wheel synchronous pulley 2-6, and the fourth spacer 2-7 are sleeved on the calf motor drive shaft 2-8. The first spacer 2-3 is connected to the left side plate 1-2 in the frame assembly 1 by hexagonal flat head screws. The first rotating wheel synchronous pulley 2-6 is connected to the second rotating wheel synchronous pulley 5-5 in the rotating wheel assembly 5 by a synchronous belt. The calf motor drive shaft 2-8 is connected to the first circular flange single-row bearing 1-1 in the frame assembly 1 and the second circular flange single-row bearing 3-1 in the thigh plate 3 by a small clearance fit. The transmission sleeve 4-6 in the clutch assembly 4 is sleeved on the calf motor drive shaft 2-8. The two are fixedly connected by a convex end set screw.

[0008] Specifically, the thigh plate assembly 3 includes a second circular flange single-row bearing 3-1, a first thigh plate 3-2, a thigh timing pulley 3-3, and a second thigh plate 3-4; four second circular flange single-row bearings 3-1 are respectively mounted on the first thigh plate 3-2 and the second thigh plate 3-4 using hexagon socket head cap screws, and the thigh timing pulley 3-3 is mounted on the side of the first thigh plate 3-2 near the left side plate 1-2 in the frame assembly 1 using hexagon socket head cap screws; that is, the second circular flange single-row bearings 3-1, the thigh timing pulley 3-3, and the first thigh plate 3-2 and the second thigh plate All parts 3-4 are fixedly connected by screws; the two second circular flange single-row bearings 3-1 of the first thigh plate 3-2 and the second thigh plate 3-4 near the frame assembly 1 are connected to the calf motor drive shaft 2-8 in the calf motor assembly 2 through a small clearance fit; the two second circular flange single-row bearings 3-1 of the other end of the first thigh plate 3-2 and the second thigh plate 3-4 are connected to the rotating wheel shaft 5-4 in the rotating wheel assembly 5 through a small clearance fit; the thigh synchronous pulley 3-3 is connected to the keyless synchronous pulley 9-1 in the thigh motor assembly 9 through a synchronous belt.

[0009] Specifically, the clutch assembly 4 includes a composite bushing 4-1, a second sliding coupling 4-2, a first sliding coupling 4-3, a hole retaining ring 4-4, a shaft retaining ring 4-5, a transmission sleeve 4-6, and a deep groove ball bearing 4-7. The composite bushing 4-1 is embedded in the ear plate mounting hole extending radially on the outer periphery of the second sliding coupling 4-2; the second sliding coupling 4-2 is sleeved on the deep groove ball bearing 4-7; the first sliding coupling 4-3 is sleeved on the transmission sleeve 4-6; the hole retaining ring 4-4 ​​is embedded in the retaining ring groove of the second sliding coupling 4-2; the shaft retaining ring 4-5 is embedded in the retaining ring groove of the transmission sleeve 4-6; and the transmission sleeve 4-6 is sleeved on the calf motor drive shaft 2-8 in the calf motor assembly 2. Threaded holes are tapped on both sides of the circumference of shaft 4-6, and through holes with corresponding alignment are tapped on both sides of the circumference of shaft 2-8 of the lower leg motor drive shaft. The two are fixedly connected by a convex end set screw. Deep groove ball bearing 4-7 is sleeved on transmission sleeve 4-6. That is, the composite bushing 4-1, the hole snap ring 4-4, and the shaft snap ring 4-5 are all installed by inlay. The second sliding coupling 4-2, the first sliding coupling 4-3, the transmission sleeve 4-6, and the deep groove ball bearing 4-7 are all installed by sleeve. The composite bushing 4-1 is connected to the lower leg tie rod 10-1 in the lower leg assembly 10 by a pin. The first sliding coupling 4-3 is connected to the shift fork 6-1 in the shift fork assembly 6 by inlay with a small clearance fit in the inner groove.

[0010] Specifically, the rotating wheel assembly 5 includes a rotating wheel 5-1, a hexagonal nut 5-2, a fifth spacer 5-3, a rotating wheel shaft 5-4, and a second rotating wheel timing pulley 5-5; two rotating wheels 5-1 are sleeved on both ends of the rotating wheel shaft 5-4; two hexagonal nuts 5-2 are installed on the threads at both ends of the rotating wheel shaft 5-4, with the two rotating wheels 5-1 located between the two hexagonal nuts 5-2; two fifth spacers 5-3 arranged at intervals are sleeved on the rotating wheel shaft 5-4, with the two fifth spacers 5-3 located between the two hexagonal nuts 5-2, and the two hexagonal nuts 5-2 located between the two rotating wheels 5-1; the second rotating wheel timing pulley 5-5 is sleeved on the rotating wheel shaft. The rotating wheel 5-4 is located on the side of the corresponding rotating wheel 5-1 away from the hexagonal nut 5-2; that is, the rotating wheel 5-1, the fifth spacer 5-3, and the second rotating wheel synchronous pulley 5-5 are all sleeved on the rotating wheel shaft 5-4 by means of sleeve, and the hexagonal nut 5-2 is screwed on the rotating wheel shaft 5-4 by means of thread; the rotating wheel shaft 5-4 is respectively connected to the two second circular flange single-row bearings 3-1 below the thigh plate assembly 3 and the circular flange double-row bearings 10-5 in the calf assembly 10 by means of small clearance fit through connection, and the second rotating wheel synchronous pulley 5-5 is connected to the first rotating wheel synchronous pulley 2-6 in the calf motor assembly 2 by means of synchronous belt.

[0011] Specifically, the shift fork assembly 6 includes a shift fork 6-1, a guide rod 6-2, a trajectory bearing 6-3, and a linear bearing 6-4; the shift fork 6-1 and the linear bearing 6-4 are sleeved on the guide rod 6-2; the guide rod 6-2 is installed between the left side plate 1-2 and the right side plate 1-5 in the frame assembly 1 by means of hexagonal flat head screws; the trajectory bearing 6-3 is connected to the shift fork 6-1 by threads; that is, both the shift fork 6-1 and the linear bearing 6-4 are sleeved on the guide rod 6-2, and the trajectory bearing 6-3 is screwed onto the upper end of the shift fork 6-1 by means of threads; the shift fork 6-1 is embedded in the first sliding coupling 4-3 in the clutch assembly 4 by means of a small clearance fit with an inner groove, and the trajectory bearing 6-3 slides up and down in the cam plate 7-4 in the clutch drive assembly 7 according to the trajectory of the straight groove through a straight groove.

[0012] Specifically, the clutch drive assembly 7 includes an electric push rod washer 7-1, an electric push rod 7-2, a cam plate base 7-3, and a cam plate 7-4. The electric push rod washer 7-1 and the electric push rod 7-2 are fixedly connected to the upper side plate 1-3 of the frame assembly 1 by hexagonal head screws, and the electric push rod 7-2 is connected to the cam plate 7-4 by hexagonal head screws. The cam plate base 7-3 is fixedly connected to the upper side plate 1-3 by hexagonal head screws. The cam plate 7-4 is embedded in the cam plate base 7-3 through a groove in the cam plate base 7-3. That is, the electric push rod washer 7-1, the electric push rod 7-2, and the cam plate base 7-3 are all fixedly connected by screws, and the cam plate 7-4 is installed by being embedded in the cam plate base 7-3.

[0013] Specifically, the suspended floating assembly 8 includes a fixing block 8-1, threaded pins 8-2, a limiting block housing 8-3, a tower spring 8-4, and a limiting block 8-5. The fixing block 8-1 is installed on the left side plate 1-2 of the frame assembly 1 using hexagon socket head cap screws, and its installation position is lower than that of the lower leg motor assembly 2 and the upper leg motor assembly 9. One side of each of the three threaded pins 8-2 is threaded onto the left, right, and upper sides of the limiting block 8-5, respectively, and the other side of each threaded pin 8-2 extends into the straight slots on the left, right, and upper sides of the fixing block 8-1, respectively. The limiting block housing 8-3 is installed using hexagon socket head cap screws. On the limiting block 8-5, three tower springs 8-4 are fixed in the cylindrical groove of the limiting block 8-5 by means of their minor diameter and major diameter. The minor diameter is embedded in the cylindrical groove of the limiting block 8-5, and the major diameter is pushed against the inside of the fixing block 8-1. The limiting block 8-5 is suspended inside the fixing block 8-1 by means of threaded pin 8-2 and tower springs 8-4. That is, the fixing block 8-1 and the limiting block shell 8-3 are both fixedly connected by screws. The threaded pin 8-2 is screwed into the limiting block 8-5 by means of threads. The tower springs 8-4 are installed by means of embedding. The limiting block 8-5 is installed by means of threaded pin 8-2 extending into the straight groove of the fixing block 8-1.

[0014] Specifically, the thigh motor assembly 9 includes a keyless synchronous pulley 9-1, a thigh motor adapter plate 9-2, a thigh motor drive shaft 9-3, and a thigh motor 9-4. The keyless synchronous pulley 9-1 is sleeved on the thigh motor drive shaft 9-3. The thigh motor adapter plate 9-2 is connected to the thigh motor 9-4 by a countersunk hexagonal screw. The thigh motor drive shaft 9-3 is connected to the thigh motor 9-4 by a countersunk Phillips head screw. That is, the keyless synchronous pulley 9-1 is sleeved on the thigh motor drive shaft 9-3, and the thigh motor adapter plate 9-2, thigh motor drive shaft 9-3, and thigh motor 9-4 are all fixedly connected by screws. The keyless synchronous pulley 9-1 is connected to the thigh synchronous pulley 3-3 in the thigh plate assembly 3 by a synchronous belt, and the thigh motor adapter plate 9-2 is connected to the left side plate 1-2 in the frame assembly 1 by a countersunk hexagonal head screw.

[0015] Specifically, the lower leg assembly 10 includes a lower leg pull rod 10-1, a lower leg 10-2, a ball bearing 10-3, a foot end rubber sleeve 10-4, and a circular flange double row bearing 10-5; one end of the lower leg pull rod 10-1 is connected to the lower leg 10-2 via a pin; the ball bearing 10-3 is connected to the lower leg 10-2 via threads; the two foot end rubber sleeves 10-4 are respectively installed on both sides of the foot end of the lower leg 10-2 using hexagon countersunk screws; the circular flange double row bearing 10-5 is installed on the lower leg 10-2 using hexagon socket head cap screws. On 0-2; that is, the lower leg pull rod 10-1 is connected to the lower leg 10-2 by a pin, the ball bearing 10-3 is screwed onto the lower leg 10-2 by a thread, and the foot end rubber sleeve 10-4 and the circular flange double row bearing 10-5 are both fixedly connected by screws; the other end of the lower leg pull rod 10-1 is connected to the composite bushing 4-1 in the clutch assembly 4 by a pin, and the circular flange double row bearing 10-5 is connected to the rotating wheel shaft 5-4 in the rotating wheel assembly 5 by a small clearance fit.

[0016] The beneficial effects of this invention are:

[0017] This invention employs a design scheme combining standard parts, purchased parts, and modified parts, resulting in low overall manufacturing costs and good engineering feasibility. The wheel-legged robot's leg device has a compact overall structure, small size, and weighs only 4 kg, meeting the needs of use in complex terrain environments. It adopts a modular design, with the entire machine including a frame assembly, lower leg motor assembly, upper leg plate assembly, clutch assembly, rotating wheel assembly, shift fork assembly, clutch drive assembly, suspension and floating assembly, upper leg motor assembly, and lower leg assembly. Each module can be independently processed and installed, and the entire machine can be assembled after the unit assembly is completed, significantly improving manufacturing efficiency and facilitating later maintenance and replacement. This invention, based on reinforcement learning algorithms, achieves the function of adaptively selecting the best motion mode according to the current terrain features under complex terrain conditions through the following steps: environmental and state information acquisition → action space construction → reward function design → policy learning and updating → motion mode decision and execution → feedback and closed-loop optimization. This effectively improves the environmental adaptability and motion stability of the device. The coordinated cooperation of the clutch component, shift fork component, and clutch drive component of this invention enables selective switching between foot-based and wheel-based motion modes using only a single drive motor, reducing the complexity of the drive system and minimizing power redundancy. The suspension floating component of this invention also ensures that even if the lower leg drive fails when the device is in wheel-based motion mode, the lower leg can still be stably suspended within the suspension floating component, and the lower leg has elastic buffering capacity in multiple dimensions, thereby effectively reducing the impact force on the body of the device during movement and further improving the stability and reliability of the overall operation. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0019] Figure 2 These are the left and right isometric projections of the present invention;

[0020] Figure 3 This is the front view of the present invention;

[0021] Figure 4 This is the left view of the present invention;

[0022] Figure 5 This is the right view of the present invention;

[0023] Figure 6 This is a top view of the present invention;

[0024] Figure 7 This is a rear view of the present invention;

[0025] Figure 8 This is a structural diagram of the framework components of the present invention;

[0026] Figure 9This is a structural diagram of the calf motor assembly of the present invention;

[0027] Figure 10 This is a structural diagram of the thigh plate assembly of the present invention;

[0028] Figure 11 This is a structural diagram of the clutch assembly of the present invention;

[0029] Figure 12 This is an exploded view of the clutch assembly of the present invention;

[0030] Figure 13 This is a structural diagram of the rotating wheel assembly of the present invention;

[0031] Figure 14 This is an exploded view of the rotating wheel assembly of the present invention;

[0032] Figure 15 This is a structural diagram of the shift fork assembly of the present invention;

[0033] Figure 16 This is a diagram of the clutch drive structure of the present invention;

[0034] Figure 17 This is a structural diagram of the suspended floating component of the present invention;

[0035] Figure 18 This is a structural diagram of the thigh motor assembly of the present invention;

[0036] Figure 19 This is a structural diagram of the lower leg assembly of the present invention.

[0037] The components in the diagram are labeled as follows: 1. Frame assembly; 2. Lower leg motor assembly; 3. Upper leg plate assembly; 4. Clutch assembly; 5. Rotating wheel assembly; 6. Shift fork assembly; 7. Clutch drive assembly; 8. Suspension floating assembly; 9. Upper leg motor assembly; 10. Lower leg assembly; 1-1. First circular flange single-row bearing; 1-2. Left side plate; 1-3. Upper side plate; 1-4. Limiting support plate; 1-5. Right side plate; 1-6. Support rod; 2-1. Lower leg motor adapter plate; 2-2. Lower leg motor; 2-3. First spacer; 2-4. Second spacer; 2-5. Third spacer; 2-6. First rotating wheel synchronous belt pulley; 2-7. Fourth spacer; 2-8. Lower leg motor drive shaft; 3-1. Second circular flange single-row bearing; 3-2. First upper leg plate; 3-3. Upper leg synchronous belt pulley; 3-4. Second upper leg plate; 4-1. Composite bushing; 4-2. Second sliding coupling; 4-3. 1. Sliding coupling, 4-4 circlip for hole, 4-5 circlip for shaft, 4-6. Transmission sleeve, 4-7. Deep groove ball bearing, 5-1. Rotating wheel, 5-2. Hexagonal nut, 5-3. Fifth spacer, 5-4. Rotating wheel shaft, 5-5. Second rotating wheel synchronous belt pulley, 6-1. Shift fork, 6-2. Guide rod, 6-3. Track bearing, 6-4. Linear bearing, 7-1. Electric push rod washer, 7-2. Electric push rod, 7-3. Cam plate base, 7-4. Cam plate, 8-1. Fixing block, 8-2. Threaded pin, 8-3. Limiting block housing, 8-4. Tower spring, 8-5. Limiting block, 9-1. Keyless synchronous belt pulley, 9-2. Thigh motor adapter plate, 9-3. Thigh motor drive shaft, 9-4. Thigh motor, 10-1. Lower leg tie rod, 10-2. Lower leg, 10-3. Ball bearing, 10-4. Foot end rubber sleeve, 10-5. Circular flange double row bearing. Detailed Implementation

[0038] The invention will be further described below with reference to the accompanying drawings and embodiments, but the scope of the invention is not limited to the description.

[0039] Example 1: As Figure 1-19As shown, this invention provides a leg device for a wheeled robot based on a reinforcement learning algorithm, comprising a frame assembly 1, a lower leg motor assembly 2, a thigh plate assembly 3, a clutch assembly 4, a rotating wheel assembly 5, a fork assembly 6, a clutch drive assembly 7, a suspension and floating assembly 8, a thigh motor assembly 9, and a lower leg assembly 10. The lower leg motor assembly 2 is mounted on the left side of the frame assembly 1 using hexagonal flathead screws and passes through the frame assembly 1, thigh plate assembly 3, and clutch assembly 4. The lower leg motor assembly 2, in cooperation with the clutch assembly 4 and the fork assembly 6, selectively transmits power from the lower leg motor 2-2 to the lower leg 10-2 or the rotating wheel 5-1. The rotating wheel assembly 5 is installed between the lower leg assemblies 10, providing stable support for the lower leg assemblies 10. The fork assembly 6 is installed on the first sliding... On coupling 4-3, the first sliding coupling 4-3 can move left and right, that is, whether it engages with the second sliding coupling 4-2, thereby realizing the engagement and disengagement of the foot movement power; the clutch drive assembly 7 is installed on the upper side plate 1-3 by hexagonal flat head screws, providing the power for the shift fork assembly 6 to move left and right; the suspension floating assembly 8 is installed on the right rear of the left side plate 1-2 by hexagonal cylindrical head screws, providing the lower leg 10-2 with suspension and multi-directional elasticity in the wheel-like movement state; the thigh motor assembly 9 is installed on the left side of the frame assembly 1 by hexagonal flat head screws, providing power to the thigh plate 3-2 through the keyless synchronous pulley 9-1, the thigh synchronous pulley 3-3 and the synchronous belt; the lower leg assembly 10 and the clutch assembly 4 are connected by a pin shaft, forming a crank rocker mechanism.

[0040] Furthermore, the frame assembly 1 includes a first circular flange single-row bearing 1-1, a left side plate 1-2, an upper side plate 1-3, a limiting support plate 1-4, a right side plate 1-5, and a support rod 1-6; the two first circular flange single-row bearings 1-1 are respectively installed on the left side plate 1-2 and the right side plate 1-5 using hexagon socket head cap screws. The use of the first circular flange single-row bearings 1-1 is to facilitate the fixing of the bearings and the frame assembly 1; the limiting support plate 1-4 is installed between the left side plate 1-2 and the right side plate 1-5 using hexagon socket head cap screws. The use of the limiting support plate 1-4 is primarily to increase... The working space for the thigh and calf is also to increase the overall strength of the frame assembly 1; the left side plate 1-2, the upper side plate 1-3, and the right side plate 1-5 are spliced ​​and fixed together by hexagonal head screws; the support rod 1-6 is installed between the left side plate 1-2 and the right side plate 1-5 by hexagonal head screws, and the use of the support rod 1-6 is also to increase the overall strength of the frame assembly 1; that is, the first circular flange single-row bearing 1-1, the left side plate 1-2, the upper side plate 1-3, the limiting support plate 1-4, the right side plate 1-5, and the support rod 1-6 are all fixedly connected by screws.

[0041] Further, the calf motor assembly 2 includes a calf motor adapter plate 2-1, a calf motor 2-2, a first spacer 2-3, a second spacer 2-4, a third spacer 2-5, a first rotating wheel synchronous pulley 2-6, a fourth spacer 2-7, and a calf motor drive shaft 2-8. The calf motor adapter plate 2-1 is connected to the calf motor 2-2 by countersunk hexagonal screws. The motor adapter plate 2-1 is used to facilitate the installation and removal of the motor. The first spacer 2-3 is connected to the side of the calf motor adapter plate 2-1 away from the calf motor 2-2 by countersunk hexagonal screws. The second spacer 2-4, the third spacer 2-5, the first rotating wheel synchronous pulley 2-6, the fourth spacer 2-7, and the calf motor drive shaft 2-8. -5 and the fourth spacer 2-7 are coaxially arranged with the calf motor drive shaft 2-8 and are sleeved on the calf motor drive shaft 2-8; the calf motor drive shaft 2-8 is connected to the calf motor 2-2 by a cross-head countersunk screw; the first rotating wheel synchronous pulley 2-6 is sleeved on the calf motor drive shaft 2-8; that is, the calf motor adapter plate 2-1, the calf motor 2-2, and the calf motor drive shaft 2-8 are all fixedly connected by screws, and the second spacer 2-4, the third spacer 2-5, the first rotating wheel synchronous pulley 2-6, and the fourth spacer 2-7 are sleeved on the calf motor drive shaft 2-8. The calf motor assembly 2 is the power source for either foot movement or wheel movement, and its main purpose is to reduce power redundancy so that only one motor can be used to selectively achieve both movement states.

[0042] Furthermore, the thigh plate assembly 3 includes a second circular flange single-row bearing 3-1, a first thigh plate 3-2, a thigh timing pulley 3-3, and a second thigh plate 3-4. The four second circular flange single-row bearings 3-1 are respectively installed on the first thigh plate 3-2 and the second thigh plate 3-4 by hexagon socket head cap screws. The thigh timing pulley 3-3 is installed on the side of the first thigh plate 3-2 near the left side plate 1-2 in the frame assembly 1 by hexagon socket head cap screws. The thigh timing pulley 3-3 is used to drive the thigh plate to reduce the complexity of the structure and to transmit the thigh power more directly. That is, the second circular flange single-row bearing 3-1, the thigh timing pulley 3-3, and the first thigh plate 3-2 and the second thigh plate 3-4 are all fixedly connected by screws.

[0043] Further, the clutch assembly 4 includes a composite bushing 4-1, a second sliding coupling 4-2, a first sliding coupling 4-3, a hole retaining ring 4-4, a shaft retaining ring 4-5, a transmission sleeve 4-6, and a deep groove ball bearing 4-7. The composite bushing 4-1 is embedded in the ear plate mounting hole extending radially on the outer periphery of the second sliding coupling 4-2. The composite bushing 4-1 is used to increase lubrication and reduce sliding friction. The second sliding coupling 4-2 is sleeved on the deep groove ball bearing 4-7. The first sliding coupling 4-3 is sleeved on the transmission sleeve 4-6. The first sliding coupling 4-3 and the transmission sleeve 4-6 are made into a splined internal and external meshing configuration. The power is transmitted in a certain way; the circlip 4-4 is embedded in the circlip groove of the second sliding coupling 4-2, and the circlip 4-4 is used to fix the deep groove ball bearing 4-7; the shaft circlip 4-5 is embedded in the circlip groove of the transmission sleeve 4-6, and the shaft circlip 4-5 is used to fix the deep groove ball bearing 4-7; the transmission sleeve 4-6 is sleeved on the calf motor drive shaft 2-8 in the calf motor assembly 2, and the transmission sleeve 4-6 has threaded holes on both sides of its circumference, and the calf motor drive shaft 2-8 has through holes on both sides of its circumference, and the two are fixedly connected by a convex end set screw; the deep groove ball bearing 4-7 is sleeved on the transmission sleeve 4-6. Specifically, the composite bushing 4-1, the hole circlip 4-4, and the shaft circlip 4-5 are all installed by inserting, while the second sliding coupling 4-2, the first sliding coupling 4-3, the transmission sleeve 4-6, and the deep groove ball bearing 4-7 are all installed by sleeve.

[0044] Further, the rotating wheel assembly 5 includes rotating wheels 5-1, hexagonal nuts 5-2, fifth spacers 5-3, rotating wheel shaft 5-4, and a second rotating wheel synchronous pulley 5-5; two rotating wheels 5-1 are sleeved on both ends of the rotating wheel shaft 5-4; two hexagonal nuts 5-2 are installed on the threads at both ends of the rotating wheel shaft 5-4, and the two rotating wheels 5-1 are located between the two hexagonal nuts 5-2. The hexagonal nuts 5-2 are used to prevent axial movement of the rotating wheel shaft 5-4; two fifth spacers 5-3 arranged at intervals are sleeved on the rotating wheel shaft 5-4. The rotating wheel shaft 5-4 has two fifth spacers 5-3 located between two hexagonal nuts 5-2, and the two hexagonal nuts 5-2 located between two rotating wheels 5-1; the second rotating wheel synchronous pulley 5-5 is sleeved on the rotating wheel shaft 5-4 and located on the side of the corresponding rotating wheel 5-1 away from the hexagonal nuts 5-2; that is, the rotating wheel 5-1, the fifth spacer 5-3, and the second rotating wheel synchronous pulley 5-5 are all sleeved on the rotating wheel shaft 5-4, and the hexagonal nuts 5-2 are screwed on the rotating wheel shaft 5-4 by means of threads.

[0045] Furthermore, the shift fork assembly 6 includes a shift fork 6-1, a guide rod 6-2, a track bearing 6-3, and a linear bearing 6-4; the shift fork 6-1 and the linear bearing 6-4 are sleeved on the guide rod 6-2; the guide rod 6-2 is installed between the left side plate 1-2 and the right side plate 1-5 in the frame assembly 1 by means of an internal hexagonal flat head screw. The guide rod 6-2 is used to increase the torque resistance of the shift fork 6-1; the track bearing 6-3 is connected to the shift fork 6-1 by means of threads; that is, both the shift fork 6-1 and the linear bearing 6-4 are sleeved on the guide rod 6-2 by means of sleeve, and the track bearing 6-3 is screwed onto the upper end of the shift fork 6-1 by means of threads.

[0046] Furthermore, the clutch drive assembly 7 includes an electric push rod washer 7-1, an electric push rod 7-2, a cam plate base 7-3, and a cam plate 7-4. The electric push rod washer 7-1 and the electric push rod 7-2 are fixedly connected to the upper side plate 1-3 of the frame assembly 1 by hexagonal head screws, and the electric push rod 7-2 is connected to the cam plate 7-4 by hexagonal head screws. The cam plate base 7-3 is fixedly connected to the upper side plate 1-3 by hexagonal head screws. The cam plate 7-4 is embedded in the cam plate base 7-3 through a groove in the cam plate base 7-3. The cam plate 7-4 is used to place the electric push rod 7-2 vertically, as horizontal placement would affect the installation space and structural complexity. That is, the electric push rod washer 7-1, the electric push rod 7-2, and the cam plate base 7-3 are all fixedly connected by screws, and the cam plate 7-4 is installed by being embedded in the cam plate base 7-3.

[0047] Further, the suspended floating assembly 8 includes a fixing block 8-1, threaded pins 8-2, a limiting block housing 8-3, tower springs 8-4, and limiting blocks 8-5; the fixing block 8-1 is installed to the right rear of the left side plate 1-2 in the frame assembly 1 by means of hexagon socket head cap screws; one side of the three threaded pins 8-2 is respectively installed on the left, right, and top sides of the limiting block 8-5 by means of threads, and the other side of the three threaded pins 8-2 extends into the straight slots on the left, right, and top sides of the fixing block 8-1 respectively; the limiting block housing 8-3 is installed on the limiting block 8-5 by means of hexagon socket head cap countersunk screws; the three tower springs 8-4, by means of their small diameter and large diameter characteristics, The smaller diameter is embedded and fixed in the cylindrical groove of the limiting block 8-5, while the larger diameter rests inside the fixing block 8-1. The tower spring 8-4 is used to ensure that the lower leg still has elasticity in multiple dimensions after being suspended in the suspension floating assembly 8. The limiting block 8-5 is suspended inside the fixing block 8-1 by the threaded pin 8-2 and the tower spring 8-4. That is, the fixing block 8-1 and the limiting block shell 8-3 are both fixedly connected by screws. The threaded pin 8-2 is screwed into the limiting block 8-5 by threads, and the tower spring 8-4 is installed by embedding. The limiting block 8-5 is installed by the threaded pin 8-2 extending into the straight groove of the fixing block 8-1.

[0048] Furthermore, the thigh motor assembly 9 includes a keyless synchronous pulley 9-1, a thigh motor adapter plate 9-2, a thigh motor drive shaft 9-3, and a thigh motor 9-4. The keyless synchronous pulley 9-1 is sleeved on the thigh motor drive shaft 9-3. The thigh motor adapter plate 9-2 is connected to the thigh motor 9-4 by countersunk hexagonal screws. The thigh motor adapter plate 9-2 is used to facilitate the installation and removal of the thigh motor 9-4. The thigh motor drive shaft 9-3 is connected to the thigh motor 9-4 by countersunk Phillips head screws. That is, the keyless synchronous pulley 9-1 is sleeved on the thigh motor drive shaft 9-3, and the thigh motor adapter plate 9-2, thigh motor drive shaft 9-3, and thigh motor 9-4 are all fixedly connected by screws.

[0049] Furthermore, the lower leg assembly 10 includes a lower leg rod 10-1, a lower leg 10-2, a ball bearing 10-3, a foot end rubber sleeve 10-4, and a circular flange double-row bearing 10-5. One end of the lower leg rod 10-1 is connected to the lower leg 10-2 via a pin. The lower leg rod 10-1 is used to form a crank-rocker structure, which provides more uniform power transmission compared to synchronous belt and synchronous belt pulley drives. The ball bearing 10-3 is connected to the lower leg 10-2 via threads. Connection: Two foot end rubber sleeves 10-4 are respectively installed on both sides of the foot end of the lower leg 10-2 by internal hex socket countersunk screws; the round flange double row bearing 10-5 is installed on the lower leg 10-2 by internal hex socket cylindrical head screws; that is, the lower leg tie rod 10-1 is connected to the lower leg 10-2 by a pin, the ball bearing 10-3 is screwed onto the lower leg 10-2 by a thread, and both the foot end rubber sleeves 10-4 and the round flange double row bearing 10-5 are fixedly connected by screws.

[0050] Furthermore, the reinforcement learning algorithm is applied to the device of the present invention, and the specific application process is as follows:

[0051] 1) Constructing the state space: A current sensor, encoder, and temperature sensor are installed on the lower leg motor 2-2 in the lower leg motor assembly 2 to collect the current, speed, and temperature information of the lower leg motor 2-2, respectively. An angle sensor is installed at the joint of the lower leg 10-2 in the lower leg assembly 10 to collect the angle information of the lower leg 10-2 joint. A current sensor, encoder, and temperature sensor are installed on the thigh motor 9-4 in the thigh motor assembly 9 to collect the current, speed, and temperature information of the thigh motor 9-4, respectively. An angle sensor is installed at the joint of the first thigh plate 3-2 in the thigh plate assembly 3 to collect the thigh joint angle information. A proximity switch is installed on the clutch assembly 4 to collect the engagement and disengagement status information of the clutch assembly 4. A linear displacement sensor is installed on the suspension floating assembly 8 to collect the floating displacement information of the limit block 8-5 in the suspension floating assembly 8. An inertial measurement unit and a vision sensor are installed on the upper side plate 1-3 in the frame assembly 1 to collect the robot's posture information and terrain features, respectively. The above multi-source data are preprocessed to construct a state space to describe the robot's current motion state and environmental features.

[0052] 2) Defining the action space: Based on the clutch switching mechanism composed of clutch assembly 4, shift fork assembly 6, and clutch drive assembly 7 in the leg device, and combined with the coordinated driving capability of thigh motor assembly 9 and calf motor assembly 2, the action space of the reinforcement learning algorithm is divided into the following discrete action modes:

[0053] ① Wheel-type movement mode: The clutch assembly 4 is in the disengaged state, the power of the lower leg 10-2 is lost and it is elastically supported by the suspension floating assembly 8. The two rotating wheels 5-1 in the rotating wheel assembly 5 are driven by the lower leg motor drive shaft 2-8 in the lower leg motor assembly 2 to achieve rolling forward.

[0054] ② Foot-based movement mode: The clutch assembly 4 is engaged, the crank rocker mechanism formed by the lower leg assembly 10 and the clutch assembly 4 is unlocked, and foot-based walking is achieved by the coordinated drive of the thigh motor assembly 9 and the lower leg motor assembly 2.

[0055] ③ Wheel-foot switching mode: The clutch drive assembly 7 drives the shift fork assembly 6 to change the engagement state of the clutch assembly 4, realizing the dynamic switching between wheel-type and foot-type, and controlling the thigh motor assembly 9 and the calf motor assembly 2 to output corresponding coordinated movements during the switching process.

[0056] 3) Construct a multi-objective reward function The effectiveness of action execution is quantitatively evaluated, with each sub-reward function defined as follows:

[0057] ① Shock absorption and terrain adaptability bonus The floating displacement is based on the linear displacement sensor installed on the suspended floating component 8. calculate, ,in To achieve the desired equilibrium displacement, the damping effect and terrain adaptability are evaluated by penalizing displacement deviation.

[0058] ② Energy consumption level reward The current values ​​are collected by current sensors installed on the thigh motor assembly 9 and the calf motor assembly 2. , calculate, To punish high energy consumption;

[0059] ③ Wheel-type rolling efficiency reward Based on the travel speed of the rotating wheel assembly 5 in wheeled mode Total power of the motor calculate, ,in To avoid division by zero for small constants, used to evaluate rolling efficiency;

[0060] ④ Footwork stability reward : The rate of change of angles collected by thigh joint angle sensors and lower leg joint angle sensors in foot mode , calculate, To prevent drastic fluctuations in the angle of punishment, the smooth motion of the crank-rocker mechanism is ensured.

[0061] ⑤ Switch performance rewards The switching completion time is based on the proximity switch set on clutch assembly 4. And the impact acceleration collected by the inertial measurement unit (IMU) calculate, It is used to evaluate the response speed and impact during the switching process.

[0062] in These are the weighting coefficients. The adjustment factors for each sub-item are all determined through reinforcement learning training;

[0063] 4) Training and strategy optimization: Based on the current state, the selected action and the obtained reward value, the motion control strategy is trained and iteratively updated using a reinforcement learning algorithm, so that the algorithm gradually learns the best collaborative control strategy for each component in the leg device under different terrain conditions, including the output distribution between the thigh motor component 9 and the calf motor component 2, the action timing of the clutch drive component 7, and the stroke control of the fork component 6.

[0064] 5) Online decision-making and execution: Based on the optimal strategy obtained from training, the action mode decision results and the target angle, target speed, target torque limit of thigh motor 9-4 and calf motor 2-2 and the clutch target state of clutch assembly 4 are output during real-time operation. Control commands are sent to the control board through the host computer to drive thigh motor assembly 9, calf motor assembly 2, clutch drive assembly 7 and shift fork assembly 6 to work together to complete the actual execution of wheel movement, foot movement or wheel-foot switching.

[0065] 6) Closed-loop feedback and continuous optimization: During motion execution, real-time operational data is continuously acquired through sensors installed on each component, including the current, speed, and temperature information of the lower leg motor 2-2, the angle information of the lower leg joint, the current, speed, and temperature information of the thigh motor 9-4, the engagement and disengagement status information of the clutch assembly 4, the floating displacement information of the limit block 8-5, robot posture information, and terrain features. This real-time operational data is then re-input into the reinforcement learning algorithm for policy correction and parameter fine-tuning, thereby continuously improving the leg device's adaptability and motion stability in complex terrain environments.

[0066] The wheel-legged robot leg device described in any of the above-mentioned embodiments can be applied to a mobile robot in complex terrain to enable the robot to adaptively switch between wheeled and legged motion in different terrain environments.

[0067] Initially, the device is in its initial motion mode. Driven by the control system, the thigh motor assembly 9 and the lower leg motor assembly 2 move the leg structure according to a preset trajectory, enabling the wheel-footed robot's leg device to move on flat ground in a wheeled motion mode. When the device enters a complex terrain area, the sensor assembly collects terrain information, posture information, and joint state information in real time, and inputs this information into a reinforcement learning algorithm for processing. The reinforcement learning algorithm outputs a motion mode decision result based on the current state information, driving the clutch assembly 4, the shift fork assembly 6, and the clutch drive assembly 7 to work together, completing the switch from wheeled motion mode to legged motion mode under the drive of a single motor. After the motion mode switch is completed, the thigh motor assembly 9 and the lower leg motor assembly 2 work together to drive the leg structure, enabling the device to support and traverse uneven terrain in a legged motion mode. During legged motion, the suspension floating assembly 8 provides multi-dimensional elastic buffering and constraint support to the lower leg assembly 10, effectively reducing the impact of terrain impact on the device body and improving motion stability. When the device re-enters a terrain area suitable for wheeled movement, the reinforcement learning algorithm re-determines the movement mode based on real-time feedback information, controlling the clutch assembly 4, fork assembly 6, and clutch drive assembly 7 to switch from legged movement mode to wheeled movement mode, enabling the device to resume high-speed rolling motion. In wheeled movement mode, even if the lower leg drive fails, the lower leg assembly 10 can still be stably suspended within the suspension floating assembly 8, ensuring the overall safety and reliability of the device's operation. Through the above process, the wheeled robot's leg device achieves continuous, autonomous movement and mode switching under different terrain conditions, fully demonstrating its excellent environmental adaptability and movement stability.

[0068] The device of this invention has a compact structure, simple configuration, and high integration. It can not only analyze the motion state and terrain information of the wheel-legged robot's leg device in real time through reinforcement learning algorithms and adaptively select the optimal motion mode, but also achieve rapid switching between wheeled and legged motion with a single drive motor through the coordinated action of the clutch component, shift fork component, and clutch drive component. This allows the device to adapt to the mobility and operation needs under different terrain conditions.

[0069] The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A wheel-legged robot leg device based on a reinforcement learning algorithm, characterized by: The frame assembly (1), the lower leg motor assembly (2), the thigh plate assembly (3), the clutch assembly (4), the rotating wheel assembly (5), the shift fork assembly (6), the clutch drive assembly (7), the suspension floating assembly (8), the thigh motor assembly (9), and the lower leg assembly (10) are provided. The lower leg motor assembly (2), the clutch drive assembly (7), the suspension floating assembly (8), and the thigh motor assembly (9) are installed on the frame assembly (1), and the lower leg motor assembly (2) penetrates the frame assembly (1), the thigh plate assembly (3), and the clutch assembly (4), one end of the lower leg assembly (10) is connected with the clutch assembly (4) through a pin shaft, and a crank rocker mechanism is formed; the rotating wheel assembly (5) is installed on both sides of the middle part of the lower leg assembly (10), the shift fork assembly (6) is installed on the first sliding coupling (4-3) in the clutch assembly (4), and the other end of the lower leg assembly (10) is matched with the suspension floating assembly (8).

2. The wheel-legged robot leg device based on a reinforcement learning algorithm according to claim 1, characterized in that: The frame assembly (1) comprises first circular flange single-row bearings (1-1), left side plates (1-2), upper side plates (1-3), limiting support plates (1-4), right side plates (1-5), and support rods (1-6); the two first circular flange single-row bearings (1-1) are installed on the left side plates (1-2) and the right side plates (1-5) arranged in parallel through inner hexagonal cylindrical head screws, the limiting support plates (1-4) are installed between the left side plates (1-2) and the right side plates (1-5) through inner hexagonal cylindrical head screws, the left side plates (1-2), the upper side plates (1-3), and the right side plates (1-5) are spliced and fixed together through inner hexagonal cylindrical head screws, and the support rods (1-6) are installed between the left side plates (1-2) and the right side plates (1-5) through inner hexagonal cylindrical head screws; the lower leg motor assembly (2) and the thigh motor assembly (9) are installed at the same height of the left side plates (1-2) through inner hexagonal flat round head screws, the guide rods (6-2) in the shift fork assembly (6) are installed between the left side plates (1-2) and the right side plates (1-5) through inner hexagonal flat round head screws, the clutch drive assembly (7) is installed on the upper side plates (1-3) through inner hexagonal flat round head screws, and the suspension floating assembly (8) is installed on the left side plates (1-2) through inner hexagonal cylindrical head screws and is installed below the lower leg motor assembly (2) and the thigh motor assembly (9).

3. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The calf motor assembly (2) comprises a calf motor adapter plate (2-1), a calf motor (2-2), a first spacer sleeve (2-3), a second spacer sleeve (2-4), a third spacer sleeve (2-5), a first rotating wheel synchronous pulley (2-6), a fourth spacer sleeve (2-7), and a calf motor drive shaft (2-8); the calf motor adapter plate (2-1) is connected with the calf motor (2-2) through an internal hexagon countersunk screw; the first spacer sleeve (2-3) is connected with the calf motor adapter plate (2-1) away from the calf motor (2-2) side through an internal hexagon flat round head screw; the second spacer sleeve (2-4), the third spacer sleeve (2-5), and the fourth spacer sleeve (2-7) are sleeved on the calf motor drive shaft (2-8) and coaxially arranged with the calf motor drive shaft (2-8); one end of the calf motor drive shaft (2-8) is connected with the output end of the calf motor (2-2) through a cross countersunk screw; and the first rotating wheel synchronous pulley (2-6) is sleeved on the other end of the calf motor drive shaft (2-8); the first spacer sleeve (2-3) is connected with the left side plate (1-2) in the frame assembly (1) through an internal hexagon flat round head screw; the first rotating wheel synchronous pulley (2-6) is connected with the second rotating wheel synchronous pulley (5-5) in the rotating wheel assembly (5) through a synchronous belt; the calf motor drive shaft (2-8) is respectively connected with the first circular flange single row bearing (1-1) in the frame assembly (1) and the second circular flange single row bearing (3-1) in the thigh plate (3) through small gap fitting penetration connection; and the transmission sleeve (4-6) in the clutch assembly (4) is sleeved on the calf motor drive shaft (2-8), and the two are fixedly connected through a convex end thimble; The thigh motor assembly (9) comprises a keyless synchronous pulley (9-1), a thigh motor adapter plate (9-2), a thigh motor drive shaft (9-3), and a thigh motor (9-4); the keyless synchronous pulley (9-1) is sleeved on the thigh motor drive shaft (9-3); the thigh motor adapter plate (9-2) is connected with the thigh motor (9-4) through an internal hexagon countersunk screw; and the thigh motor drive shaft (9-3) is connected with the thigh motor (9-4) through a cross countersunk screw; the keyless synchronous pulley (9-1) is connected with the thigh synchronous pulley (3-3) in the thigh plate assembly (3) through a synchronous belt; and the thigh motor adapter plate (9-2) is connected with the left side plate (1-2) in the frame assembly (1) through an internal hexagon flat round head screw.

4. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The thigh plate assembly (3) comprises second circular flange single row bearings (3-1), a first thigh plate (3-2), a thigh synchronous pulley (3-3), and a second thigh plate (3-4); the four second circular flange single row bearings (3-1) are respectively installed on both ends of the first thigh plate (3-2) and the second thigh plate (3-4) by means of hexagonal head screws; the thigh synchronous pulley (3-3) is installed on the first thigh plate (3-2) close to the left side plate (1-2) of the frame assembly (1) by means of a hexagonal head screw; the two second circular flange single row bearings (3-1) on one end of the first thigh plate (3-2) and the second thigh plate (3-4) close to the frame assembly (1) are connected with the calf motor drive shaft (2-8) in the calf motor assembly (2) through small gap fitting; the two second circular flange single row bearings (3-1) on the other end of the first thigh plate (3-2) and the second thigh plate (3-4) are connected with the rotating wheel shaft (5-4) in the rotating wheel assembly (5) through small gap fitting; and the thigh synchronous pulley (3-3) is connected with the keyless synchronous pulley (9-1) in the thigh motor assembly (9) through a synchronous belt.

5. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The clutch assembly (4) comprises a composite shaft sleeve (4-1), a second sliding coupling (4-2), a first sliding coupling (4-3), a hole clasp spring (4-4), a shaft clasp spring (4-5), a transmission sleeve (4-6), and a deep groove ball bearing (4-7); the composite shaft sleeve (4-1) is embedded in the ear plate mounting hole extending in the radial direction of the second sliding coupling (4-2); the second sliding coupling (4-2) is sleeved on the deep groove ball bearing (4-7); the first sliding coupling (4-3) is sleeved on the transmission sleeve (4-6); the hole clasp spring (4-4) is embedded in the clasp spring groove of the second sliding coupling (4-2); the shaft clasp spring (4-5) is embedded in the clasp spring groove of the transmission sleeve (4-6); the transmission sleeve (4-6) is sleeved on the calf motor drive shaft (2-8) in the calf motor assembly (2), and threaded holes are formed on both sides of the circumference of the transmission sleeve (4-6); the calf motor drive shaft (2-8) has aligned through holes on both sides of the circumference thereof; the transmission sleeve (4-6) is sleeved on the deep groove ball bearing (4-7); the composite shaft sleeve (4-1) is connected with the calf pull rod (10-1) in the calf assembly (10) through a pin shaft; and the first sliding coupling (4-3) is embeddedly connected with the shift fork (6-1) in the shift fork assembly (6) through a small gap fitting recess.

6. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The rotating wheel assembly (5) comprises rotating wheels (5-1), hexagonal nuts (5-2), fifth spacer sleeves (5-3), rotating wheel shafts (5-4), and second rotating wheel synchronous pulleys (5-5); two rotating wheels (5-1) are sleeved on both ends of the rotating wheel shaft (5-4), two hexagonal nuts (5-2) are installed on the threads at both ends of the rotating wheel shaft (5-4), and the two rotating wheels (5-1) are located between the two hexagonal nuts (5-2); two fifth spacer sleeves (5-3) are sleeved on the rotating wheel shaft (5-4) and located between the two hexagonal nuts (5-2) and the two rotating wheels (5-1); the second rotating wheel synchronous pulley (5-5) is sleeved on the rotating wheel shaft (5-4) and located on the side of the corresponding rotating wheel (5-1) away from the hexagonal nut (5-2); the rotating wheel shaft (5-4) is connected through small gap fitting with the two second circular flange single row bearings (3-1) below the thigh plate assembly (3) and the circular flange double row bearing (10-5) in the calf assembly (10), and the second rotating wheel synchronous pulley (5-5) is connected with the first rotating wheel synchronous pulley (2-6) in the calf motor assembly (2) through a synchronous belt.

7. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The shifting fork assembly (6) comprises a shifting fork (6-1), a guide rod (6-2), a track bearing (6-3), and a linear bearing (6-4); the shifting fork (6-1) and the linear bearing (6-4) are sleeved on the guide rod (6-2), the guide rod (6-2) is installed between the left side plate (1-2) and the right side plate (1-5) in the frame assembly (1) through an internal hexagonal flat round head screw, and the track bearing (6-3) is connected with the shifting fork (6-1) through threads; the shifting fork (6-1) is connected with the first sliding coupling (4-3) in the clutch assembly (4) through internal groove small gap fitting embedding, and the track bearing (6-3) slides up and down in the cam plate (7-4) in the clutch drive assembly (7) according to the track of the straight slot in the form of a straight slot; The clutch drive assembly (7) comprises an electric push rod gasket (7-1), an electric push rod (7-2), a cam plate base (7-3), and a cam plate (7-4); the electric push rod gasket (7-1) and the electric push rod (7-2) are fixedly connected to the upper side plate (1-3) in the frame assembly (1) through an internal hexagonal flat round head screw, the electric push rod (7-2) is connected with the cam plate (7-4) through an internal hexagonal flat round head screw, the cam plate base (7-3) is fixedly connected to the upper side plate (1-3) through an internal hexagonal flat round head screw, and the cam plate (7-4) is embedded in the cam plate base (7-3) through the groove in the cam plate base (7-3).

8. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The suspension floating assembly (8) comprises a fixed block (8-1), threaded pins (8-2), a limiting block shell (8-3), tower springs (8-4), and limiting blocks (8-5). The fixed block (8-1) is installed on the left side plate (1-2) of the frame assembly (1) by means of an internal hexagonal cylindrical head screw, and the installation position is lower than the calf motor assembly (2) and the thigh motor assembly (9). One side of the three threaded pins (8-2) is respectively installed on the left side, right side and upper side of the limiting block (8-5) by means of threads. The other side of the three threaded pins (8-2) respectively extends into the straight slot on the left side, right side and upper side of the fixed block (8-1). The limiting block shell (8-3) is installed on the limiting block (8-5) by means of an internal hexagonal countersunk screw. The three tower springs (8-4) are characterized by small and large diameters. The small diameter is embedded and fixed in the cylindrical groove of the limiting block (8-5), and the large diameter is on the inside of the fixed block (8-1). The limiting block (8-5) is suspended in the inside of the fixed block (8-1) by means of the threaded pins (8-2) and the tower springs (8-4).

9. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The calf assembly (10) comprises a calf pull rod (10-1), a calf (10-2), a ball bearing (10-3), a foot end rubber sleeve (10-4), and a circular flange double row bearing (10-5). One end of the calf pull rod (10-1) is connected to the calf (10-2) through a pin shaft. The ball bearing (10-3) is connected to the calf (10-2) through threads. Two foot end rubber sleeves (10-4) are respectively installed on both sides of the foot end of the calf (10-2) through internal hexagonal countersunk screws. The circular flange double row bearing (10-5) is installed on the calf (10-2) through an internal hexagonal cylindrical head screw. The other end of the calf pull rod (10-1) is connected to the composite shaft sleeve (4-1) in the clutch assembly (4) through a pin shaft. The circular flange double row bearing (10-5) is connected to the rotating wheel shaft (5-4) in the rotating wheel assembly (5) through small gap fitting.

10. The wheel-legged robot leg device based on reinforcement learning algorithm according to claim 1, characterized in that: The reinforcement learning algorithm is applied to the device, and the specific application process is as follows: 1) Constructing state space: Set current sensors, encoders and temperature sensors on the calf motor (2-2) in the calf motor assembly (2) to collect current, speed and temperature information of the calf motor (2-2), and set an angle sensor at the joint of the calf (10-2) in the calf assembly (10) to collect the joint angle information of the calf (10-2); Set current sensors, encoders and temperature sensors on the thigh motor (9-4) in the thigh motor assembly (9) to collect current, speed and temperature information of the thigh motor (9-4), and set an angle sensor at the joint of the first thigh plate (3-2) in the thigh plate assembly (3) to collect the joint angle information of the thigh; Set a proximity switch on the clutch assembly (4) to collect the combined and separated state information of the clutch assembly (4); Set a linear displacement sensor on the suspension floating assembly (8) to collect the floating displacement information of the limit block (8-5) in the suspension floating assembly (8); Set an inertial measurement unit and a vision sensor on the upper side plate (1-3) in the frame assembly (1) to collect robot posture information and terrain features; Preprocess the above multi-source data to construct a state space for describing the current motion state and environmental characteristics of the robot; 2) Define the action space: Based on the clutch switching mechanism composed of the clutch assembly (4), the shift fork assembly (6) and the clutch drive assembly (7) in the leg device, combined with the cooperative driving ability of the thigh motor assembly (9) and the calf motor assembly (2), the action space of the reinforcement learning algorithm is divided into the following discrete action modes: ① Wheeled motion mode: the clutch assembly (4) is in a separated state, the calf (10-2) power disappears and is supported by the suspension floating assembly (8) for elastic lifting, and the two rotating wheels (5-1) in the rotating wheel assembly (5) are driven to roll forward by the calf motor drive shaft (2-8) in the calf motor assembly (2); ② Legged motion mode: the clutch assembly (4) is in a connected state, the crank rocker mechanism composed of the calf assembly (10) and the clutch assembly (4) is unlocked, and the legged walking is realized by the cooperative driving of the thigh motor assembly (9) and the calf motor assembly (2); ③ Wheel-foot switching mode: the engagement state of the clutch assembly (4) is changed by driving the shift fork assembly (6) through the clutch drive assembly (7), realizing the dynamic switching between wheeled and legged modes, and controlling the thigh motor assembly (9) and the calf motor assembly (2) to output corresponding coordinated motion during the switching process; 3) Constructing multi-objective reward function; 4) Training and strategy optimization: based on the current state, the selected action and the obtained reward value, the reinforcement learning algorithm is used to train and iteratively update the motion control strategy, so that the algorithm gradually learns the best cooperative control strategy for each component in the leg device under different terrain conditions; 5) Online decision and execution: According to the optimal strategy obtained by training, the action mode decision result and the corresponding target angle, target speed, target torque limit value of the thigh motor (9-4) and the small leg motor (2-2), and the clutch target state of the clutch assembly (4) are output in real-time operation process. Control instructions are sent to the control board through the upper computer to drive the thigh motor assembly (9), the small leg motor assembly (2), the clutch driving assembly (7) and the shift fork assembly (6) to co-act, and the actual execution of wheeled motion, foot motion or wheel-foot switching is completed; 6) Closed-loop feedback and continuous optimization: During the motion execution process, real-time running data is continuously obtained through sensors arranged on each component, and the real-time running data is re-input into the reinforcement learning algorithm for strategy correction and parameter fine-tuning.

Images