A rescue robot

Abstract

This application provides a rescue robot, relating to the field of robotics technology, comprising a chassis, a main track unit mounted on the chassis, four secondary track units, and four secondary actuators. The four secondary track units are respectively connected around the main track unit. The main track unit includes a main drive wheel, a main driven wheel, and a main track. The main drive wheel is driven to the main driven wheel via the main track. Two secondary track units are driven to both sides of the main drive wheel along the width direction of the rescue robot, and the remaining two secondary track units are driven to both sides of the main driven wheel along the width direction of the rescue robot. The main drive wheel is driven to move synchronously, causing the four secondary track units to move synchronously. The four secondary actuators are driven to different secondary track units, causing the corresponding secondary track unit to swing relative to the main track unit in a vertical plane, improving the adaptability and stability of the rescue robot in varying terrains.

Description

Technical Field

[0001] This application relates to the field of robotics, and more specifically, to a rescue robot. Background Technology

[0002] With the continuous development of science and technology and the advancement of industrial automation, the application of robotics technology is gradually becoming more widespread in various fields, especially in special environments (such as disaster sites, extreme climate zones, and unstable terrain), where the demand for robotics technology is increasing. These environments often pose significant risks to human operation, and traditional wheeled or legged robots, due to their limited adaptability in complex terrain, struggle to complete tasks efficiently and stably. Therefore, intelligent tracked vehicles with strong off-road capabilities and the ability to adapt to complex terrain are gradually becoming a research hotspot.

[0003] Tracked robots, due to their excellent off-road mobility, especially in tasks such as climbing slopes and crossing ditches, have a significant advantage over wheeled robots and are suitable for many unstructured environments. This makes them more widely applicable in complex tasks such as disaster relief. However, when faced with extremely complex ruin terrain (such as scenarios with large elevation differences), tracked robots are still less efficient at overcoming obstacles.

[0004] To address this issue, the published document CN223224432U proposes an intelligent fire rescue robot capable of navigating stairs and traversing complex terrain by swinging its forearm. This design enhances the robot's terrain adaptability through the swinging mechanism of its forearm, effectively overcoming certain obstacles. However, this design suffers from a complex transmission method between the tracked robot body and the swinging mechanism. Furthermore, the swing arm can only be mounted on the front of the robot, lacking space for a rear swing arm. This results in insufficient robot stability, making it difficult to maintain good operability and stability in diverse and complex environments. Utility Model Content

[0005] The purpose of this application is to provide a rescue robot to address the shortcomings of the aforementioned technologies.

[0006] To achieve the above objectives, the technical solution adopted in this application is as follows:

[0007] This application provides a rescue robot, including a chassis, a main track unit, four auxiliary track units, and four auxiliary drivers. The main track unit is mounted on the chassis, and the four auxiliary track units are respectively connected around the main track unit. The main track unit includes a main drive wheel, a main driven wheel, and a main track. The main drive wheel is driven to the main driven wheel via the main track. Two auxiliary track units are driven to both sides of the main drive wheel along the width direction of the rescue robot, and the remaining two auxiliary track units are driven to both sides of the main driven wheel along the width direction of the rescue robot. The main drive wheel is driven to move to drive the four auxiliary track units to move synchronously.

[0008] The four auxiliary drives are connected to different auxiliary track units to drive the corresponding auxiliary track units to swing relative to the main track unit in the vertical plane.

[0009] Furthermore, each subtrack unit includes a subtrack drive assembly and an inner support plate. The main drive wheel or the main driven wheel is connected to the subtrack drive assembly to drive the subtrack drive assembly to move synchronously. The subtrack drive assembly is rotatably connected to the inner support plate. The sub-drive unit is connected to the inner support plate to drive the subtrack unit to swing via the inner support plate.

[0010] Furthermore, each secondary track drive assembly includes a secondary drive wheel, a secondary driven wheel, and a secondary track. The main drive wheel or the main driven wheel is connected to the secondary drive wheel via a drive transmission. The secondary drive wheel is connected to the secondary driven wheel via the secondary track. The two ends of the inner support plate are rotatably connected to the secondary drive wheel and the secondary driven wheel, respectively.

[0011] Furthermore, each secondary track unit also includes a connecting shaft and a connecting sleeve with an opening at one end. One end of the connecting shaft passes through the open end of the connecting sleeve to be inserted into the connecting sleeve, and the other end of the connecting shaft is connected to the secondary drive wheel. The closed end of the connecting sleeve is connected to the main drive wheel or the main driven wheel.

[0012] Furthermore, the cross-section of the connecting shaft is polygonal, and the connecting sleeve is interference-fitted with the connecting shaft.

[0013] Furthermore, each auxiliary track unit also includes a small pulley and a large pulley connected by a belt drive. The small pulley is connected to the output shaft of the auxiliary drive, and the large pulley is rotatably sleeved outside the connecting sleeve and connected to the inner support plate.

[0014] Furthermore, the rescue robot also includes two side plates installed on both sides of the main track unit, with each side plate having its two ends movably connected to two auxiliary track units located on one side of the main track unit along the width direction of the rescue robot.

[0015] Furthermore, each sub-track unit also includes a first drive wheel and a second drive wheel. The first drive wheel is rotatably sleeved outside the connecting sleeve. The large pulley is connected to the inner support plate via the first drive wheel. The second drive wheel is rotatably sleeved outside the first drive wheel. The two ends of each side plate are respectively connected to two second drive wheels located on one side of the main track unit along the width direction of the rescue robot.

[0016] Furthermore, each secondary track unit also includes multiple tensioning wheels spaced apart between the secondary drive wheel and the secondary driven wheel. The outer wall surface of each tensioning wheel abuts against the tensioning surface of the secondary track to adjust the tension of the secondary track.

[0017] Furthermore, the rescue robot also includes a six-degree-of-freedom robotic arm mounted on top of the main tracked unit for grasping rescue items.

[0018] The beneficial effects of this application include:

[0019] This application provides a rescue robot, including a main track unit, four secondary track units, and four secondary actuators. The four secondary track units are connected around the main track unit. The main track unit includes a main drive wheel, a driven wheel, and a main track, with the main drive wheel connected to the driven wheel via the main track. Two secondary track units are connected to the main drive wheel on both sides along the width of the rescue robot, and the remaining two secondary track units are connected to the driven wheel on both sides along the width of the rescue robot. The main drive wheel is driven to move synchronously, causing the four secondary track units to move synchronously. The four secondary actuators are connected to different secondary track units to drive the corresponding secondary track unit to swing relative to the main track unit in a vertical plane. With this design, the rescue robot can exhibit excellent off-road performance in different types of environments. Compared with traditional tracked robots, this solution, by setting secondary track units around the robot and combining them with precise control of the secondary actuators, achieves flexible vertical movement of the robot. This innovative design effectively enhances the robot's mobility in extreme terrains, especially in complex scenarios such as ruins and terrains with significant elevation differences, enabling it to complete tasks more efficiently and stably. The implementation of this technical solution greatly improves the adaptability and stability of rescue robots in varied terrains, providing stronger support for the application of robots in disaster relief and hazardous environments. Attached Figure Description

[0020] Figure 1 This application provides a structural schematic diagram of a rescue robot.

[0021] Figure 2 This is one of the partial exploded views of a rescue robot provided in this application;

[0022] Figure 3 This is the second partially exploded view of a rescue robot provided in this application;

[0023] Figure 4 A partial sectional view of a rescue robot provided in this application;

[0024] Figure 5 This application provides a design drawing of the electronic control system for a rescue robot.

[0025] Figure 6 This is a diagram illustrating the actual operation of SLAM for mapping and navigation.

[0026] Figure 7 Maps created using SLAM and deep learning.

[0027] Icons: 1-Main track unit; 101-Main drive wheel; 2-Secondary track unit; 201-Outer support plate; 202-Secondary track; 203-Tensioner wheel; 204-Secondary drive wheel; 205-Secondary driven wheel; 206-Inner support plate; 207-Fixed wheel; 208-Connecting shaft; 209-Flange; 210-Bolt; 211-First transmission wheel; 212-Second transmission wheel; 213-Crossed roller bearing; 214-Connecting sleeve; 215-Large pulley; 216-Secondary driver; 3-Side plate; 4-Mechanical arm. Detailed Implementation

[0028] This application provides a rescue robot that, through the coordinated transmission of the main track unit 1 and the auxiliary track unit 2, can effectively improve its obstacle-crossing and adaptability in complex terrain and environments. Figures 1 to 4 As shown, the rescue robot includes a chassis, a main track unit 1, four auxiliary track units 2, and four auxiliary actuators 216. The main track unit 1 is mounted on the chassis, and the four auxiliary track units 2 are respectively mounted around the main track unit 1 to enhance the robot's stability and mobility. The main track unit 1 includes a main drive wheel 101, a main driven wheel, and a main track. The main drive wheel 101 is connected to the main driven wheel via the main track. When the main drive wheel 101 is driven to rotate, it drives the main track, which in turn drives the main driven wheel to rotate synchronously, thereby moving the robot and realizing the driving function of the main track unit 1. The coordinated operation of the main drive wheel 101 and the main driven wheel enables the robot to maintain continuous and stable mobility in complex terrain.

[0029] Two auxiliary track units 2 are connected to both sides of the main drive wheel 101 along the robot's width direction, and two other auxiliary track units 2 are connected to both sides of the main and driven wheels along the robot's width direction. The movement of the main drive wheel 101 directly drives the two auxiliary track units 2 connected to both sides of the main drive wheel 101 to move synchronously. At the same time, it also drives the two auxiliary track units 2 connected to both sides of the main and driven wheels to move synchronously, thereby achieving synchronous operation of four auxiliary track units 2. This ensures that the robot has better balance and maneuverability during movement, especially in uneven or obstacle-ridden environments, effectively distributing the load and improving the adaptability of the track system. The auxiliary track units 2 make the robot more flexible during movement, enabling it to adapt to various terrain changes, such as crossing ditches or traversing complex obstacles.

[0030] Furthermore, four secondary actuators 216 are connected to different secondary track units 2, enabling each secondary track unit 2 to swing in the vertical plane. This swinging mechanism allows the robot to effectively handle complex obstacles such as elevation differences and stairs. The swinging function of the secondary track units 2 relative to the main track unit 1 not only enhances the robot's obstacle-crossing ability but also improves its adaptability in complex environments. For example, when the robot encounters obstacles with significant elevation differences, the swinging of the secondary track units 2 can quickly adjust their contact angle, guiding the robot to cross obstacles more smoothly, reducing dependence on terrain, and improving obstacle-crossing efficiency. It should be noted that the movement of each secondary track unit 2 is controlled by its own independent secondary actuator 216, thus achieving the independence and flexibility of the secondary track units 2. This design allows each secondary track unit 2 to freely adjust its swing angle according to the actual terrain requirements. The swing angle of the secondary track units 2 can reach 360°, but under normal circumstances, the swing angle does not exceed 90° to ensure the stability and operability of the robot in complex environments. Furthermore, the two front subtrack units 2 typically have the same swing angle, as do the two rear subtrack units 2; however, the swing angles of the front and rear subtrack units 2 can differ. This configuration not only ensures the robot's balance during movement but also optimizes its adaptability to different terrains. When encountering obstacles, slopes, or terrain with significant elevation differences, the subtrack units 2 can achieve better obstacle-crossing ability and maneuverability by using different swing angles. The swing angle of each subtrack unit 2 can be flexibly adjusted according to changes in terrain, ensuring the robot can efficiently traverse various complex environments.

[0031] This design enables the rescue robot to exhibit excellent off-road performance in various environments. Compared to traditional tracked robots, this solution achieves flexible vertical movement by incorporating secondary track units 2 around the robot and combining them with the precise control of secondary actuators 216. This innovative design effectively improves the robot's mobility in extreme terrains, especially in complex scenarios such as ruins and terrains with significant elevation differences, enabling it to complete tasks more efficiently and stably. The implementation of this technical solution greatly enhances the adaptability and stability of the rescue robot in varied terrains, providing stronger support for its application in disaster relief and hazardous environments.

[0032] Furthermore, the rescue robot includes two main track units 1 arranged side-by-side along its width. This configuration significantly enhances the robot's propulsion power while improving its balance and stability in rough and unstable terrain. The two main track units 1 can be controlled synchronously by a single main drive or separately by two independent main drives. The movement of each main track unit 1 can be precisely adjusted as needed while maintaining synchronous operation. The main tracks must possess excellent wear resistance and traction to adapt to diverse and harsh terrains such as mud and gravel. The main drives utilize surface-mount motors paired with reducers to provide ample and stable power for the robot's movement, ensuring stable progress even under heavy loads or when climbing slopes, preventing terrain from hindering search and rescue operations. Moreover, this layout reduces the space occupied by the equipment, allowing for a more rational chassis layout.

[0033] Furthermore, each secondary track unit 2 includes a secondary track transmission assembly and an inner support plate 206. The secondary track transmission assembly includes a secondary drive wheel 204, a secondary driven wheel 205, and a secondary track 202. The main drive wheel 101 is connected to the secondary drive wheels 204 of the two secondary track units 2 via a transmission mechanism, while the main and driven wheels are connected to the secondary drive wheels 204 of the other two secondary track units 2. This configuration ensures that when the main drive unit drives the main drive wheel 101, it can not only drive the two secondary drive wheels 204 directly connected to both sides of the main drive wheel 101 to move synchronously, but also further drive the other two secondary drive wheels 204 connected to both sides of the main and driven wheels to move synchronously through the main and driven wheels. Therefore, the movement of the main drive wheel 101 drives the four secondary drive wheels 204 to work together through the main track unit 1, ensuring the synchronous movement of the four secondary drive wheels 204, thereby improving the robot's driving stability and off-road performance.

[0034] Each secondary drive wheel 204 is connected to a secondary driven wheel 205 via a secondary track 202. Driven by the secondary drive wheels 204, the secondary track 202 operates, thereby driving the secondary driven wheels 205 to rotate synchronously. In this way, the secondary track unit 2 can assist the main track unit 1, jointly propelling the robot to move in complex environments. The design of the transmission system of the secondary track 202 makes the transmission relationship between the secondary drive wheels 204 and the secondary driven wheels 205 more stable, enabling efficient cooperation in complex terrain and ensuring that the robot can successfully overcome various obstacles and continue moving.

[0035] Furthermore, the output shaft of the secondary driver 216 is connected to the inner support plate 206, and both ends of the inner support plate 206 are rotatably connected to the secondary drive wheel 204 and the secondary driven wheel 205, respectively. By driving the rotation of the inner support plate 206 through the secondary driver 216, the secondary track unit 2 can swing in the vertical plane, thereby further enhancing the robot's obstacle-crossing ability. The secondary driver 216 employs a high-performance harmonic reduction motor to output sufficient torque, effectively supporting the secondary track unit 2 in achieving its multiple functions.

[0036] It is worth noting that the rotation of the secondary drive wheel 204 and the secondary driven wheel 205 is limited to driving the rotation of the secondary track 202 and the secondary driven wheel 205 themselves, and will not drive the rotation of the inner support plate 206. The inner support plate 206 rotates only under the drive of the secondary driver 216, thereby driving the secondary track unit 2 to swing independently in the vertical plane without interfering with the rotation of the transmission assembly of the secondary track 202 itself. This design ensures that the transmission and swing functions of the secondary track unit 2 can operate independently without interference, improving the robot's flexibility and accuracy. For example, during normal travel, the secondary track unit 2 can rotate precisely and synchronously with the main track unit 1, allowing the robot to advance quickly on flat or gentle slopes with a stable and efficient posture; when encountering obstacles such as ditches, steps, and protruding rocks, it can swing independently through the secondary driver 216, flexibly adjusting the posture of the secondary track unit 2, like a climbing arm, assisting the main track unit 1 in crossing obstacles, greatly improving the obstacle-crossing ability of the search and rescue robot, and helping it to overcome terrain limitations and successfully reach the target area in complex scenarios such as disaster relief and search and rescue.

[0037] Each secondary track unit 2 also includes an outer support plate 201 and two fixed wheels 207. The two ends of the outer support plate 201 are respectively fixed to different fixed wheels 207 by screws. The two fixed wheels 207 are rotatably connected to the secondary drive wheel 204 and the secondary driven wheel 205 via ball bearings. The inner support plate 206 and the outer support plate 201 are located on both sides of the transmission assembly of the secondary track 202 along the width direction of the robot. With this design, both sides of the secondary track unit 2 are protected by the inner and outer support plates 201, and the periphery of the secondary track unit 2 is covered by the secondary track 202, forming a stable and well-protected structure. This structure not only improves the stability of the secondary track unit 2 but also effectively reduces interference from the external environment, ensuring that the secondary track unit 2 can work reliably in harsh environments.

[0038] Furthermore, each secondary track unit 2 includes a connecting shaft 208 and a connecting sleeve 214 with an opening. One end of the connecting shaft 208 passes through the open end of the connecting sleeve 214 via a plug-in connection, thus achieving a stable connection between the two. The other end of the connecting shaft 208 is connected to the secondary drive wheel 204, ensuring that the secondary drive wheel 204 can receive power from the primary drive wheel 101 or the driven wheel. The closed end of the connecting sleeve 214 is connected to the primary drive wheel 101 or the driven wheel, ensuring that the power of the primary drive wheel 101 or the driven wheel can be accurately transmitted to the secondary drive wheel 204. The key to this structural design lies in the interference fit between the connecting shaft 208 and the connecting sleeve 214. This fit eliminates any gaps at the connection point, avoiding possible relative movement or loosening. Through this high-precision coordination, the power of the main drive wheel 101 or the main driven wheel can be efficiently and stably transmitted to the secondary drive wheel 204, ensuring that the robot maintains a consistent motion performance during movement. Especially when facing complex terrain, it can ensure the continuous transmission of power without any power loss or instability.

[0039] It should be noted that the cross-section of the connecting shaft 208 is polygonal, such as triangular, square, hexagonal, or octagonal. This design ensures a non-circular fit between the connecting shaft 208 and the connecting sleeve 214, further enhancing the connection's robustness and avoiding slippage or loosening issues that may occur between a circular shaft and a circular hole. The polygonal cross-section provides a larger contact area and higher torsional resistance, making the connection more robust and ensuring that the secondary drive wheel 204 can accurately follow the movement of the primary drive wheel 101, thereby improving the driving force and stability of the track system.

[0040] Furthermore, to enhance the stability and durability of the rescue robot in complex terrain, the rescue robot also includes two side plates 3 respectively installed on both sides of the main track unit 1 along the width direction of the rescue robot. Each side plate 3 has its two ends movably connected to two auxiliary track units 2 located on one side of the main track unit 1 along the width direction of the rescue robot. This design ensures that the auxiliary track units 2 can move independently during operation, while the side plates 3 remain stable and unaffected by the movement of the auxiliary track units 2, thereby effectively improving the overall stability of the robot.

[0041] The primary function of side plates 3 is to protect the robot's chassis mechanical structure from external environmental interference. This is especially crucial during rescue missions, where robots often need to move through complex environments such as ruins, areas with significant elevation changes, and numerous obstacles. Dust, debris, or other obstacles commonly found in the external environment can cause wear and tear on the chassis's mechanical components, severely impacting the robot's normal operation. By installing side plates 3 on both sides of the robot, the chassis is effectively protected, reducing the impact of external interference on the robot's internal structure and extending its service life.

[0042] Furthermore, to optimize the power transmission system of the rescue robot and improve its motion accuracy and stability, each sub-track unit 2 also includes a first transmission wheel 211, a second transmission wheel 212, a small pulley, a belt, and a large pulley 215. The small pulley and the large pulley 215 are connected by belt drive. The small pulley is connected to the output shaft of the sub-drive 216, realizing the power output of the sub-drive 216. The large pulley 215 is sleeved on the outside of the connecting sleeve 214 or the connecting shaft 208 through ball bearings, the first transmission wheel 211 is rotatably sleeved on the outside of the connecting sleeve 214 or the connecting shaft 208 through ball bearings, and the inner support plate 206 is rotatably sleeved on the outside of the connecting sleeve 214 or the connecting shaft 208 through ball bearings. Furthermore, the large pulley 215, the first transmission wheel 211, and the inner support plate 206 are sequentially fixedly connected by bolts 210. A flange 209 is also provided on the other side of the inner support plate 206 to ensure stable and precise power transmission among the three components. That is, the large pulley 215 is fixedly connected to the inner support plate 206 via the first transmission wheel 211, and the large pulley 215 can drive the rotation of the inner support plate 206 through the first transmission wheel 211. Thus, when the auxiliary drive 216 drives the small pulley to rotate, it drives the large pulley 215 to rotate via the belt. The rotation of the large pulley 215 drives the first transmission wheel 211 to rotate, and the rotation of the first transmission wheel 211 ultimately drives the inner support plate 206 to rotate.

[0043] Because the inner support plate 206 is rotatably connected to the connecting sleeve 214 or the connecting shaft 208, it maintains an independent motion relationship with the connecting sleeve 214. The rotation of the inner support plate 206 will not cause the rotation of the connecting sleeve 214, and vice versa. When the main track unit 1 drives the connecting sleeve 214 to rotate, it will not affect the movement of the inner support plate 206. This design effectively avoids mutual interference between various transmission components, ensures the efficient operation of the power transmission system, and can precisely control the motion state of the secondary track unit 2, improving the robot's mobility and obstacle-crossing ability.

[0044] Furthermore, the second drive wheel 212 is rotatably mounted on the outside of the first drive wheel 211 via a crossed roller bearing 213. An annular groove is formed on the outer wall of the second drive wheel 212. Each side plate 3 has its two ends respectively fitted into the annular grooves of the two second drive wheels 212 located on one side of the main track unit 1, and is fixedly connected by screws, thus providing reliable support for the side plate 3. This configuration ensures that the second drive wheel 212 remains stationary while the first drive wheel 211 rotates, thereby guaranteeing the stability of the side plate 3. During the movement of the auxiliary track unit 2, the side plate 3 remains fixed, preventing interference from external movement or environmental factors, ensuring the robot's smooth movement and efficient operation in complex terrain.

[0045] This series of transmission structures features a clear design logic, well-defined transmission paths, and independent operation of each component, avoiding mutual interference or unnecessary power loss within the transmission system. Furthermore, the rational configuration and compact design of all transmission elements significantly save space, making the entire system not only highly efficient but also space-saving, adapting to the robot's maneuverability needs in confined or complex environments.

[0046] Through these designs, the robot maintains excellent off-road capability and precise control in diverse terrain conditions. The independent operation of each transmission component ensures uninterrupted movement between the robot's parts, providing reliable power output and stability. This technical solution excels in enhancing the robot's mobility and stability, providing a solid guarantee for the robot to successfully perform tasks in rubble, extreme weather, or hazardous environments during rescue missions.

[0047] Furthermore, each secondary track unit 2 also includes multiple tensioning rollers 203 spaced apart between the secondary drive wheel 204 and the secondary driven wheel 205 to enhance the tension of the secondary track 202 and ensure that the secondary track 202 maintains appropriate tension during operation, thereby avoiding slippage caused by uneven tension or slack. The outer wall surface of each tensioning roller 203 is in close contact with the tensioning surface of the secondary track 202, which can effectively adjust the tension of the secondary track 202 and ensure that the track system maintains optimal working condition when performing tasks. For example, five tensioning rollers 203 are arranged on each of the upper and lower sides of the secondary track 202. This configuration allows for effective distribution of the tension of the secondary track 202 and enhances track stability through multi-point tensioning. Each tensioning wheel 203, through contact with the tensioning surface of the sub-track 202, achieves uniform support for the sub-track 202, ensuring that the sub-track 202 maintains appropriate tension during movement, thus avoiding the problem of decreased transmission efficiency or track slippage caused by track slack.

[0048] Furthermore, to enhance the functionality and adaptability of the rescue robot in complex environments, especially in tasks such as grasping, detecting, and transporting rescue items, the rescue robot also includes a robotic arm 4 mounted on top of the main tracked unit 1. This robotic arm 4 performs grasping operations and can flexibly manipulate various objects. The robotic arm 4 has a rated maximum load of 5 kg, six degrees of freedom, and an articulated design, enabling it to cover any point in space. The robotic arm 4 has a relatively simple structure and is easy to control, thereby improving the robot's working efficiency in complex task environments.

[0049] To ensure the reliability and operational flexibility of robotic arm 4, considering factors such as torque, weight, and cost, high-torque servo motors and robot joint motors were selected as the power source. The first degree of freedom is responsible for the horizontal rotation of robotic arm 4, while the second and third degrees of freedom are used for rotation around the X-axis. These joint motors can withstand stresses in all directions, ensuring sufficient power and stability when performing grasping tasks to meet the high load requirements of actual use. The high torque output of the servo motors provides more stable power support for robotic arm 4 during delicate operations, especially in operations requiring precise control, effectively balancing output torque and control response.

[0050] To enhance the stability and gripping ability of the robotic arm 4, its gripper section adopts a three-jaw structure. The opening and closing motion of the gripper is transmitted through three ball joints, achieving multi-degree-of-freedom movement and ensuring the precision and efficiency of the gripping action. This design ensures that the robotic arm 4 provides a more stable gripping force when grasping objects, especially bottle caps and water cups. Furthermore, the material treatment at the claw tips increases the friction of the claws, further enhancing the stability of the robotic arm 4 during gripping operations and preventing objects from slipping or failing to be gripped.

[0051] In the structural design of robotic arm 4, both the upper and lower arms utilize carbon fiber tubing. To further enhance the torsional resistance and overall structural rigidity of robotic arm 4, carbon fiber plates are used to reinforce the connection points between the upper and lower arms, thereby improving the load-bearing capacity and reliability required during operation. This design not only enhances the strength of robotic arm 4 but also maintains a lightweight structure. To further enhance the strength and stability of robotic arm 4, aluminum profile tubing clamps, milled parts, and copper and aluminum pillars are used at the connection points of the upper arm, lower arm, and robotic gripper to maintain connection strength. This ensures the strength and durability of robotic arm 4 when performing high-load tasks, while avoiding the adverse effects of an excessively high center of gravity on the robot's stability.

[0052] It should be noted that the entire robot chassis can be assembled using SolidWorks 3D modeling software. From the track layout to the drive unit mounting positions, every structural element is digitally and precisely designed. Some non-load-bearing structural components, such as lightweight parts used for decoration and connection, are rapidly prototyped using 3D printing technology, which can meet customization needs and shorten the development cycle. Meanwhile, key load-bearing components, such as the main and auxiliary track 202 connection brackets and the chassis load-bearing frame, are made of hard aluminum alloy and carbon fiber. Hard aluminum alloy ensures structural strength and rigidity, while carbon fiber effectively reduces the chassis weight. The combination of the two achieves the design goal of being "strong and lightweight," allowing the robot to reduce its energy consumption and improve its endurance and operational efficiency while maintaining mobility.

[0053] In addition, to ensure chassis reliability, SolidWorks Simulation was used to conduct simulation verification of the chassis and overall vehicle structural strength. Simulations were performed under complex working conditions, such as extreme obstacle crossing and heavy-load driving, to identify structural weaknesses in advance, optimize design details, and ultimately ensure that the chassis meets the strength requirements for search and rescue operations in complex terrain. This provides a solid hardware foundation for the search and rescue robot to perform its tasks efficiently and stably.

[0054] Furthermore, rescue robots also include electronic control systems, such as... Figure 5As shown, the electronic control system includes a power supply unit, a motor drive unit, a communication network unit, a self-stabilizing gimbal unit, and a sensor unit. The system uses an STM32F405RGT6 microcontroller based on a Cortex-M4 core as the main control chip, with various integrated modules as secondary cores for joint control, and is equipped with a comprehensive anti-interference system, ensuring product stability and reliability. Furthermore, the various units communicate via automotive-grade protocols such as CAN and CANopen, greatly improving the independence, portability, and maintainability of each unit. The power supply unit uses multiple high-capacity batteries in parallel to provide long battery life and uninterrupted power.

[0055] Furthermore, the power supply unit includes a host computer power supply unit and a bottom circuit power supply unit to ensure that each control structure operates relatively independently and that the operation of the main body is not affected when a certain part fails.

[0056] Each module in the host computer power supply unit is a standardized industrial-grade product, requiring relatively low current and rarely experiencing circuit failures such as overheating, short circuits, or open circuits. Therefore, a centralized power supply scheme is adopted, with each module directly powered by the power supply. Each unit is relatively independent and does not affect others, facilitating unified control and management of the power supply. It can use a single DJI battery for power supply or two DJI batteries connected in parallel (allowing for uninterrupted battery replacement). In the lower-level circuit power supply unit, the current required by each unit and motor drive is relatively high. To reduce the power supply current while ensuring the normal operation of each module, the power supply unit adopts a distributed power supply (connected tier by tier, with lower tiers powered by upper tiers).

[0057] Furthermore, the main track unit 1 uses an ESCON-Module-50-8 drive board, and the main driver uses a MAXON motor. The drive board's input voltage is 48V, provided by a dedicated power module, ensuring a stable power input for the main track motor. Specifically, the drive board receives the pulse-width modulation output signal from the main control system, converts this signal into a speed command, and transmits it to the motor. Simultaneously, the motor's actual speed is fed back to the drive board via an encoder, forming a closed-loop control that ensures the accuracy of the main driver's movement. This control method offers high precision and can meet the robot's stable movement requirements in various terrains.

[0058] The secondary actuator 216 uses an EuPH20 planetary series motor with a 24V power supply, powered through a corresponding branch in the underlying circuit power supply unit to ensure the torque output requirements of the secondary actuator 216. The control of the secondary actuator 216 employs an incremental PID algorithm. In practical applications, the parameters KP, Ki, and Kd need to be tuned, i.e., determining the specific values ​​of the proportional coefficient, integral time, derivative time, and sampling period of the regulator. This improves the dynamic and static indicators of the system, achieving optimal control performance. This allows the secondary actuator 216 to flexibly adjust its rotation or swing arm according to the motion state of the main track unit 1 or obstacle conditions, enhancing the robot's obstacle-crossing ability.

[0059] All joint motors of robotic arm 4, including the motors used in the primary arm and other high-torque servos, are powered by 24V, supplied by the robotic arm power supply branch in the underlying power supply unit, meeting the power requirements for the multi-degree-of-freedom movement of robotic arm 4. Various sensors (such as the JY61 digital electronic sensor, MPU9250 inertial measurement unit, etc.) and control boards (such as the chassis control board, robotic arm control board, gimbal control board, etc.) are powered by 22.2V, supplied by DJI batteries in the upper-level power supply unit through a centralized power supply scheme, ensuring the stability of sensor data acquisition and control command transmission.

[0060] Furthermore, the robot's underlying communication network adopts the CAN bus as the communication standard and employs a two-tier distributed structure with upper and lower level computers. The upper-level computer (industrial control computer) connects to the CAN network via a self-designed USB-to-CAN unit, communicating with various modules at the lower level. Multiple modules in the lower-level electronic control layer (such as the chassis main controller, robotic arm main controller, gimbal main controller, and carbon dioxide acquisition board) exchange information via the CAN bus. The CAN bus is a serial communication network that effectively supports distributed and real-time control.

[0061] The characteristics of CAN in robot communication include: low cost; extremely high bus utilization; long data transmission distance (up to 10 km); high data transmission rate (up to 1 Mbps); the ability to decide whether to receive or block a message based on its ID; reliable error handling and detection mechanisms; automatic retransmission after the transmitted information is corrupted; automatic disconnection of nodes from the bus in case of serious errors; and messages do not contain source or destination addresses, but only use identifiers to indicate functional and priority information.

[0062] The secondary driver 216 section adopts the CANopen network transmission application layer protocol based on the CAN bus, which improves the isolation and security of master-slave communication while retaining the advantages of the CAN protocol. Furthermore, the motor operation status is monitored in real time via heartbeat messages in the NMT.

[0063] Furthermore, the rescue robot uses standard-length IDs and a 32-bit filter identifier mask pattern to filter out non-compliant ID information into the FIFO. Each frame of information consists of 8 bytes, with the first and last 4 bytes separated. Every 4 bytes transmit an integer or single-precision floating-point number, ensuring the accuracy and efficiency of data transmission.

[0064] Furthermore, the self-stabilizing gimbal unit includes an attitude sensor, a control unit, and an actuator. The attitude sensor integrates a JY61 digital electronic sensor and an MPU9250 inertial measurement unit, where the MPU9250 includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer for real-time acquisition of the gimbal's attitude data. The control unit uses an STM32F10RCT6 microcontroller as an independent computing unit, responsible for processing the attitude sensor data and generating control commands. The actuator includes two AX-12 digital servo platforms for achieving pitch and roll rotation of the gimbal to compensate for attitude changes during robot movement.

[0065] Based on accelerometer, gyroscope, and magnetometer data acquired by the MPU9250, attitude calculation is performed using the quaternion method. Compared to the traditional Euler angle calculation method, the quaternion method can more effectively avoid the gimbal lock problem in the attitude calculation process, improving the stability and accuracy of attitude calculation. Kalman filtering and complementary filtering are applied to the attitude sensor data to eliminate sensor data drift. Kalman filtering is used to fuse multi-sensor data to predict and update the attitude state; complementary filtering combines the advantages of accelerometers and gyroscopes to suppress high-frequency and low-frequency noise, improving the reliability of attitude data. Based on the calculated and filtered attitude data, the angle and direction that the gimbal needs to adjust are calculated. By controlling the AX-12 digital servo platform to dynamically respond in real time, the LiDAR is kept in a horizontal state, providing a stable data source for SLAM mapping and improving mapping quality.

[0066] Furthermore, the sensor unit includes a lidar unit, specifically the RPLIDAR A2, which connects to the main control system via a dedicated interface circuit. The interface circuit primarily provides power to the lidar (typically 5V or 12V, depending on the lidar model) and handles data transmission. Data transmission utilizes either a USB or UART interface, ensuring that the 360° laser ranging data acquired by the lidar is transmitted to the main control system in real-time and accurately, providing environmental contour information for SLAM mapping and autonomous navigation.

[0067] Furthermore, the sensor unit includes a thermal imager, which uses a FLIR One camera and connects to the main control system via a USB interface. The interface circuit needs to provide a stable 5V power supply and ensure the high speed and stability of USB data transmission, so that the thermal image data captured by the thermal imager can be transmitted to the system in a timely manner for locating trapped persons or dangerous targets such as fire sources in adverse conditions such as darkness and smoke.

[0068] Furthermore, the sensor unit includes a fisheye wide-angle camera, specifically a custom-designed 130° ultra-wide-angle fisheye camera, which is also connected to the main control system via a USB interface. This allows for a clear observation of the rescue robot's surroundings, facilitating specific decision-making by the operator.

[0069] Furthermore, the sensor unit includes a carbon dioxide sensor, which monitors and displays the carbon dioxide concentration in the environment in real time, with a detection range of 300-5000 ppm, meeting the gas concentration monitoring needs at disaster sites (such as enclosed spaces after earthquakes and smoke-filled environments after fires). The fusion of carbon dioxide sensor data with data from thermal imagers, sound sensors, and other sources can improve the accuracy of locating trapped individuals. For example, in enclosed ruins, areas with high carbon dioxide concentrations may be associated with the accumulation of gases produced by human respiration; combining this with thermal imaging signals can more accurately pinpoint the location of trapped individuals.

[0070] Furthermore, filter capacitors and inductors are added to the power input to suppress power supply noise and ripple, ensuring that each module receives clean power. The drive power supply and signal power supply are isolated, for example, by adding optocoupler isolation between the motor drive board and the control board to prevent interference from the drive power supply to the signals.

[0071] Furthermore, CAN bus communication uses shielded twisted-pair cable to reduce the impact of electromagnetic interference on signal transmission. Sensor signal lines use shielded cable and are routed strategically to avoid running parallel to high-voltage lines, thus reducing crosstalk.

[0072] Furthermore, a single-point grounding principle is adopted to separate the digital ground and analog ground, and finally converge them at a single point to reduce ground loop interference. The grounding planes of the control board and motor drive board are rationally designed to ensure minimum grounding impedance and improve the system's anti-interference capability.

[0073] Furthermore, the electronic control system adopts an industrial computer and uses the Ubuntu operating system to coordinate and control various functions of the platform, including simultaneous localization and mapping (SLAM) technology, vehicle body and nozzle control based on the ROS operating system, and a remote transmission system built on wireless bridge technology for transmitting camera images, thermal sensor images and voice information.

[0074] The specific workflow is as follows:

[0075] When the operator manually controls the robot, they can directly observe it and monitor its status in real time using data transmitted from onboard sensors (including laser sensors, attitude sensors, cameras, etc.). The operator can also send motion commands to the industrial control computer via a remote terminal using input devices such as a joystick, utilizing a 5.8GHz wireless network. The industrial control computer processes the commands and then sends them to the chassis, driving the robot to complete the actual movements.

[0076] When switched to autonomous navigation mode, the industrial control computer receives radar data and various sensor data transmitted from the underlying layer. It integrates SLAM technology and autonomous navigation technology based on A* heuristic algorithm and Dijkstra algorithm to draw and correct the site map in real time, while sending motion commands to the chassis to enable the robot to move autonomously and achieve the goal of autonomous navigation mapping and control.

[0077] The target recognition function works in both manual control and autonomous navigation modes. By pre-inputting target features or acquiring them using deep learning methods, the target recognition algorithm can accurately identify predetermined targets. Taking an area experiencing a sudden outbreak of epidemic as an example, the system can identify residents, monitor their body temperature using thermal imaging technology, and mark residents with abnormal temperatures on a map generated by the autonomous navigation system.

[0078] Furthermore, SLAM includes the following steps: feature extraction, data association, state estimation, state update, and feature update. All these steps are organically combined, with the ultimate goal of accurately correcting the robot's position estimation information. In practical applications, estimating position solely based on the robot's own motion often results in significant errors; therefore, motion data alone cannot determine the robot's position. The common approach is to first perform a preliminary position estimation based on the robot's motion equations, and then fully utilize the rich environmental information acquired by sensors such as ranging units for position correction. During the correction process, feature information in the environment is first extracted. After the robot moves, the positional changes of these features are observed again. Through comparative analysis, the robot's position is ultimately accurately corrected. This continuous iterative optimization process enables SLAM to provide reliable position and map information for the robot in unknown environments, greatly improving the robot's autonomous navigation capabilities.

[0079] Among numerous positioning technologies, visual positioning and lidar positioning are relative positioning, while only GPS (Global Positioning System) positioning is absolute positioning. Relative positioning requires extracting features from a structured environment; once in an area with sparse features, the system is prone to failure. For example, lidar positioning is prone to errors when environmental features are insufficient. Therefore, GPS is a crucial supplement. However, in urban environments, tall buildings can cause GPS multipath effects, leading to measurement errors of 5-10 meters or even greater.

[0080] Therefore, a solution combining laser SLAM and GPS is adopted. First, the GPS module determines the approximate location. Then, the robot's onboard LiDAR acquires location data and features, which are matched with the GPS map for precise local correction, resulting in an accurate map and positioning. Simultaneously, the real-time updated surrounding map provided by SLAM assists the robot in obstacle avoidance, improving driving safety and stability. The rescue robot is equipped with a GPS module, a self-stabilizing gimbal, and LiDAR. Utilizing autonomous navigation technology based on the A* algorithm and Dijkstra's algorithm, it integrates high-precision GPS maps with real-time SLAM positioning for path planning, achieving autonomous navigation and efficiently completing the task.

[0081] Figure 6 This diagram demonstrates the practical operation of a two-dimensional map and autonomous navigation technology, where the autonomous navigation is based on the A* heuristic algorithm and Dijkstra's algorithm. The red line in the diagram represents the route planned by the robot as it explores its surroundings during autonomous navigation.

[0082] For the sensor unit, the RPLIDAR A2 lidar was selected. Based on the principle of triangulation, this lidar can perform 360° laser ranging of obstacles within an 18-meter range in place, and transmit the ranging data to the algorithm layer to provide basic information for subsequent processing.

[0083] For the autonomous mapping function, the mature and open-source Cartographer algorithm is used. This graph optimization-based algorithm has lower performance overhead and faster data convergence speed when dealing with large maps and complex terrain. The core structure of the Cartographer algorithm consists of three main modules: front-end matching, loop closure detection, and back-end optimization.

[0084] Front-end matching: This module is responsible for scanning and matching new radar scan data with existing stock data. Through least squares optimization, the newly acquired submap is inserted into the existing map, and branch localization and pre-calculated grids are used to achieve local and global loop closures, thereby ensuring map coherence and accuracy.

[0085] Loop closure detection: In the Cartographer algorithm, loop closure detection is used to optimize the pose of all submaps. Specifically, when the pose of the current scan data is sufficiently close to the pose of a laser scan data in an already created submap, the system identifies and confirms the loop using a specific scanmatch strategy. Cartographer employs a branch-and-bound optimization method for efficient search; when a sufficiently good match is obtained, the existence of the loop is confirmed.

[0086] Backend optimization: This stage primarily focuses on optimizing camera pose. Based on the pose of the current scan data and the matching result with a pose in the closest submap, the poses in all submaps are optimized. This ensures the accuracy and consistency of the entire map, providing the robot with reliable map information to support its autonomous navigation and task execution.

[0087] In practical applications, robots are required to have fully autonomous motion mapping and scanning capabilities. To this end, an exploration algorithm is adopted, utilizing multi-layer grid maps, enabling the robot to efficiently and maximize the exploration of unknown areas in unfamiliar environments while ensuring the safety of itself and its surroundings.

[0088] Meanwhile, to improve positioning accuracy, the robot is equipped with GNSS positioning hardware, which can acquire coordinates in real time based on triangulation and then integrate multi-source data such as base station positioning and WiFi information, greatly improving positioning accuracy. After loading the school's coordinate map, the robot can achieve efficient and accurate point-to-point navigation, perfectly meeting the practical needs of campus patrols and other tasks.

[0089] Furthermore, the rescue robot utilizes the ROS control package to complete control tasks. Specifically, the rescue robot uses an industrial computer based on an Intel Core 10th generation i5 platform with an x64 architecture as the host computer, and combines it with the ROS Melodic Morenia robot operating system based on Ubuntu 18.04 to control the robot chassis and perform low-level communication. Wireless communication between the robot and the control console uses a 5.8GHz bridge communication scheme.

[0090] The rescue robot uses ROS as its core operating system and leverages related software packages such as `base_controller` to achieve precise control of the robot's main body. This effectively facilitates interaction between the robot and its hardware, enabling the robot to achieve flexible and maneuverable control in complex and ever-changing environments. `ros_control`, a key middleware provided by ROS, encompasses a rich set of functional components, including various controller interfaces, transmission interfaces, hardware interfaces, and controller toolkits. It greatly simplifies the robot application development process, facilitates rapid implementation, and significantly improves development efficiency. Furthermore, with the help of the ROS framework, the system can efficiently collect various relevant data, providing comprehensive and accurate data support for SLAM (Simultaneous Localization and Mapping) technology, thereby achieving autonomous mapping and navigation. In unknown environments, the robot can rely on this function to achieve autonomous control and complete a series of complex tasks such as patrolling residential areas and fire prevention.

[0091] From the control flowchart, the entire control architecture exhibits a high degree of systematicity and coordination. The various functional modules work together to ensure smooth operation of the robot in a series of stages, including receiving commands, data processing, path planning, environmental perception, and motion control. For example, the host computer sends control commands through the ROS system, which are then transmitted to the robot via the wireless communication module. After receiving the commands, the robot uses the ros_control framework to precisely control the chassis, joints, and other actuators, while continuously collecting sensor data for feedback, enabling real-time monitoring and adjustment of the robot's motion state. During autonomous navigation, the robot uses SLAM technology to build a map and determine its own position. Combined with path planning algorithms in ROS, such as the A* heuristic and Dijkstra's algorithm, it plans the optimal path, thereby achieving efficient and autonomous logistics delivery and patrol tasks.

[0092] In summary, by selecting an advanced hardware platform, a mature ROS operating system, and a reasonable wireless communication scheme, the rescue robot is able to achieve fully autonomous motion mapping and scanning and efficient task execution in complex real-world environments, meeting the design requirements and application scenario needs of a multi-functional autonomous inspection robot.

[0093] Furthermore, a wireless bridge is used to establish multi-segment network communication, including that of the rescue robot, to enable multi-party data transmission.

[0094] The communication system of the rescue robot enables two-way information exchange between the front and rear, encompassing data communication, video signal transmission, and audio signal transmission. Video and audio signals are transmitted wirelessly using microwave equipment, while control commands are implemented via wired or wireless remote control systems. Currently, the control modes of multi-functional autonomous inspection robots are mainly divided into wired control, wireless remote control, and autonomous control.

[0095] The rescue robot is equipped with a wireless bridge terminal, enabling efficient and stable information transmission with the control or information receiving end. Specifically, it can transmit images captured by cameras in real time, providing first-hand image data of the target environment to the rear; it can also transmit thermal data collected by thermal sensors to accurately measure the target's body temperature characteristics; and even more significantly, it can transmit voice signals, allowing personnel at both the front and rear lines to conduct direct voice communication, greatly improving the timeliness and accuracy of information exchange.

[0096] Furthermore, to enhance the efficiency of rescue robot missions, the rescue robot integrates deep learning-based object recognition technology to achieve sub-meter-level localization and semantic annotation of targets within a geographic information system. Specifically, it employs a YOLO V3 network based on the Darknet architecture, which achieves a good balance between computational efficiency and detection performance. As a single-stage detection model, this framework achieves a 57.9% mAP@0.5 value on the COCO-test-dev dataset, and with an input image resolution of 608×608, it achieves a throughput of 20 FPS on the Pascal Titan X platform. Its accuracy is essentially consistent with RetinaNet, derived from Focal Loss, while its inference latency is only one-quarter of the latter, demonstrating significant advantages over multi-stage architectures such as Faster R-CNN and SSD.

[0097] Furthermore, the rescue robot collects facial images of residents in the community using cameras, constructing a facial dataset, and trains it using a VGG-Face network built with PyTorch. After the collected image information is transmitted to the central information processing unit, it first undergoes simple sharpening. Then, OpenCV and Dlib libraries are used to perform face detection to determine if a face exists in the image. Next, OpenCV's affine transformations are used to complete face alignment, and the trained model is used for face matching. If the matching result matches the final recipient, the data is uploaded and the object is retrieved.

[0098] Then, the Faster R-CNN framework implemented in C++ was integrated and deployed to the Robot Operating System (ROS) to build a real-time target recognition module for mobile robots. Local computing environment testing showed that the system exhibited excellent recognition accuracy when using the pre-trained caffemodel model to perform the detection task. After migrating to the ROS platform, comparative verification showed that the framework maintained the same performance level as in the local environment. Through the robot's onboard vision sensors, the system can accurately identify targets within its field of view. The areas marked by the detection boxes in the figure demonstrate the algorithm's recognition results for targets in a single frame of video stream. It should be noted that the test target (a toy doll) presented a highly unconventional posture, the recognition difficulty of which even exceeded human visual discrimination ability, effectively simulating the extreme challenge of posture recognition for trapped personnel in rescue scenarios. To verify the system's robustness, a simulated toxic area marker detection experiment was added: six types of hazardous materials were printed on A4-sized paper pasted on the wall. The experimental results showed that the system successfully detected all markers (100% detection rate), and their category information was simultaneously displayed in the output image and on the terminal interface. The results show that, with sufficient training data, deep learning methods can effectively address the challenges of recognizing complex-shaped targets and maintain high-precision recognition capabilities in robotic application scenarios.

[0099] Furthermore, the rescue robot establishes a connection with the base station via an embedded 4G module, accesses a cloud server deployed on the public network via the internet, and ultimately achieves data routing with the main control computer. The advantages of this relay communication mode are: it supports millisecond-level end-to-end transmission of video streams captured by cameras, providing real-time visual data for environmental monitoring; it ensures low-latency transmission of infrared thermal imaging sensor data, enabling dynamic monitoring of target body temperature characteristics; it establishes a two-way voice communication channel to meet the real-time voice interaction needs of emergency command; and it provides remote control capabilities that overcome spatial limitations—within the 4G network coverage area, the logistics unit can maintain continuous data interaction with the control center.

[0100] Furthermore, the rescue robot rapidly locates the body temperature distribution of trapped individuals using non-contact infrared thermometry. When it detects abnormalities in the core body temperature (such as hypothermia / hyperthermia crisis), it automatically triggers a tiered alarm protocol, simultaneously transmitting vital signs data and spatial coordinates to the command center. This technology overcomes the limitations of visible light, enabling rapid identification of trapped individuals in dense smoke, darkness, or visually obstructed environments, providing crucial decision support for precise rescue within the golden rescue window.

[0101] Furthermore, the resolution of target spatial coordinates primarily relies on deep learning-based visual detection algorithms. By analyzing images within the field of view, the target's pose information relative to the robot is obtained. Simultaneously, thermal radiation features and acoustic signals are fused to construct a multi-source perception framework, enabling multi-feature collaborative identification of specific targets. The calculated target pose is mapped to the global coordinate system, and pose updates for all tracked objects are dynamically published. When a new target is detected, the system dynamically registers it to the target list and assigns an independent identifier to each entity. In a 2D grid map constructed based on Cartographer SLAM, the spatial distribution of targets is annotated in real time, ultimately outputting a topology map archive conforming to the TIFF standard. Figure 7 As shown, the constructed map closely matches the test site. Mapping accuracy verification results demonstrate that the constructed digital map of the rescue environment achieves centimeter-level spatial fidelity to the physical scene (RMSE = 0.18m). Through multi-source sensor fusion mapping technology, the system accurately reproduces the environmental geometric features and topological relationships under complex ruin structures. Its spatial error distribution meets the positioning accuracy threshold (<0.3m) for rescue robots specified in the BS ISO 19285:2017 standard, providing a reliable spatial benchmark for planning search paths for trapped individuals.

[0102] Furthermore, YOLOv3 employs an independent logistic classifier, and the classification task optimization uses a binary cross-entropy loss function.

[0103] Furthermore, the multi-scale prediction architecture implementation scheme includes:

[0104] Anchor box generation strategy: Obtain 9 prior box dimensions through K-means clustering, and distribute them to the three-level feature pyramid in ascending order of scale;

[0105] Scale 1 (13×13): The detection results are directly output by the convolutional module at the end of the basic network;

[0106] Scale 2 (26×26): The second-level features of Scale 1 are fused, upsampled by 2 times, and combined with a 16×16 feature map. After convolution and stacking, a detection box is generated.

[0107] Scale 3 (52×52): Reuses the Scale 2 processing flow, increasing the input feature map resolution to 32×32;

[0108] Backbone network Darknet-53 performance analysis:

[0109] In ImageNet benchmark tests, its classification accuracy is comparable to ResNet-101 / 152 (Top-5 error ±1%), but its inference throughput is improved by 1.5 times, providing computational efficiency for real-time detection.

[0110] Furthermore, the robotic arm 4 adopts a dual-channel control framework, which can seamlessly switch between "remote control" and "autonomous" modes to achieve safe and efficient inverse kinematics solving and collision-free trajectory generation.

[0111] The software runs on Ubuntu 18.04 + ROS Melodic. The host computer is remotely interconnected with the IPC via an Ethernet Bridge, and then connected to the CAN bus via a self-developed USB-CAN converter. The communication link uses a proprietary frame format to ensure low latency and high reliability. The overall architecture follows a top-down design principle and is divided into three layers:

[0112] (1) Motion control layer (implemented in C++17)

[0113] 1) Forward kinematics: Analyzes HID messages from mainstream controllers such as Xbox / PS, encapsulates them twice, maps them to the joint space, and drives the servos of each degree of freedom in real time;

[0114] 2) Inverse Kinematics: The OMPL library is called, and sampling-optimization hybrid algorithms such as RRT*, PRM*, and LazyLBKPIECE are integrated. The obstacle avoidance inverse kinematics planning is completed in combination with OctoMap 3D grid map.

[0115] 3) Smooth trajectory: B-spline post-processing is used to ensure the continuity of joint velocity and acceleration.

[0116] (2) Pose visualization layer (RViz plugin + Qt5)

[0117] It receives sensor_msgs / JointState and geometry_msgs / PoseStamped from the motion control layer, refreshes the ideal pose at 30 Hz within the local RViz window, and overlays obstacle point clouds to achieve WYSIWYG monitoring.

[0118] (3) Data transfer layer (C++17 + Boost.Asio)

[0119] The planned joint trajectory or handle control quantity is encoded into a private protocol frame and sent to the USB-CAN module via serial port; the frame format includes CRC-16 check and retransmission mechanism, packet loss rate <0.0001, end-to-end delay <5 ms.

[0120] It should be understood that, in addition to robotic arm 4, the dual-channel control frame can also be reused for various motion platforms such as tracked robots, gimbal cameras, and AGVs, covering scenarios such as industrial assembly, nuclear facility maintenance, and scientific research and education.

[0121] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A rescue robot, characterized in that, The system includes a chassis, a main track unit, four auxiliary track units, and four auxiliary drives. The main track unit is mounted on the chassis, and the four auxiliary track units are connected around the main track unit. The main track unit includes a main drive wheel, a main driven wheel, and a main track. The main drive wheel is connected to the main driven wheel via the main track. Two auxiliary track units are connected to the main drive wheel on both sides along the width of the rescue robot, and the remaining two auxiliary track units are connected to the main driven wheel on both sides along the width of the rescue robot. The main drive wheel is driven to move synchronously to drive the four auxiliary track units. The four auxiliary drives are connected to different auxiliary track units to drive the corresponding auxiliary track units to swing relative to the main track unit in the vertical plane; Each secondary track unit includes a secondary track drive assembly and an inner support plate. The main drive wheel or the main driven wheel is connected to the secondary track drive assembly to drive the secondary track drive assembly to move synchronously. The secondary track drive assembly is rotatably connected to the inner support plate. The secondary drive unit is connected to the inner support plate to drive the secondary track unit to swing via the inner support plate. Each secondary track drive assembly includes a secondary drive wheel, a secondary driven wheel, and a secondary track. The main drive wheel or the main driven wheel is driven by the secondary drive wheel. The secondary drive wheel is driven by the secondary track and the secondary driven wheel. The two ends of the inner support plate are rotatably connected to the secondary drive wheel and the secondary driven wheel, respectively. Each secondary track unit also includes a connecting shaft and a connecting sleeve with an opening at one end. One end of the connecting shaft passes through the open end of the connecting sleeve to be inserted into the connecting sleeve, and the other end of the connecting shaft is connected to the secondary drive wheel. The closed end of the connecting sleeve is connected to the main drive wheel or the main driven wheel.

2. The rescue robot according to claim 1, characterized in that, The cross-section of the connecting shaft is polygonal, and the connecting sleeve is interference-fitted with the connecting shaft.

3. The rescue robot according to claim 1 or 2, characterized in that, Each secondary track unit also includes a small pulley and a large pulley connected by a belt drive. The small pulley is connected to the output shaft of the secondary drive, and the large pulley is rotatably sleeved outside the connecting sleeve and connected to the inner support plate.

4. The rescue robot according to claim 3, characterized in that, The rescue robot also includes two side plates installed on both sides of the main track unit. The two ends of each side plate are movably connected to two auxiliary track units located on one side of the main track unit along the width direction of the rescue robot.

5. The rescue robot according to claim 4, characterized in that, Each secondary track unit also includes a first drive wheel and a second drive wheel. The first drive wheel is rotatably sleeved outside the connecting sleeve. The large pulley is connected to the inner support plate via the first drive wheel. The second drive wheel is rotatably sleeved outside the first drive wheel. The two ends of each side plate are respectively connected to two second drive wheels located on one side of the main track unit along the width direction of the rescue robot.

6. The rescue robot according to claim 1 or 2, characterized in that, Each secondary track unit also includes multiple tensioning wheels spaced apart between the secondary drive wheel and the secondary driven wheel. The outer wall surface of each tensioning wheel abuts against the tensioning surface of the secondary track to adjust the tension of the secondary track.

7. The rescue robot according to claim 1 or 2, characterized in that, The rescue robot also includes a six-degree-of-freedom robotic arm mounted on top of the main tracked unit for grasping rescue items.

Images