Campus garbage picking robot

Abstract

The invention discloses a campus garbage picking robot, which integrates a motion prediction technology and a real-time obstacle avoidance strategy, carries a computer program, and can improve the autonomous navigation capability and cleaning efficiency of the robot in a complex campus environment. The method comprises the following steps: acquiring environmental data by using a laser radar and an RGB-D camera, generating a static grid map through an SLAM technology, and identifying the motion state of a dynamic obstacle in real time; a time convolution network and a bidirectional GRU model are adopted to predict a dynamic obstacle trajectory, and a reachable area is generated; constructing a static situation field, a dynamic situation field and a gravitational field based on an improved artificial potential field method, and dynamically adjusting an obstacle avoidance strategy; global path planning is realized through an A * algorithm, a local path is optimized in combination with a dynamic window method, and real-time updating is carried out; high-precision path tracking is realized through combination of stepping motor driving and a PID control algorithm, and an obstacle avoidance mode is triggered in case of emergency obstacles. The garbage sweeper is suitable for garbage sweeping operation in complex environments such as campuses, shopping malls and factories.

Description

Technical Field

[0001] It relates to the field of robotics, specifically a campus garbage collection robot. Background Art

[0002] With the rapid development of artificial intelligence, robotics, and automation, mobile robots have become indispensable application tools in many scenarios. For example, in industrial manufacturing, mobile robots are often used for material handling, precision assembly, and other tasks; in medical scenarios, mobile robots are used to transport medical devices and drugs; in the agricultural field, plant protection robots are used for field spraying pesticides or sowing operations. These application fields demonstrate the important value of mobile robots in improving work efficiency and saving labor costs.

[0003] In the campus environment, garbage collection robots, as auxiliary tools for environmental cleaning and management, have gradually received attention. In the prior art, there have been some studies and applications for garbage cleaning robots. For example:

[0004] Fixed-path garbage cleaning robots: Such robots clean the road surface garbage along a preset fixed cleaning path. Although they can achieve efficient operation in a closed environment (such as an industrial plant), due to their inability to dynamically respond to environmental changes, in an open campus environment, they are prone to stagnation or even collision when encountering sudden obstacles (such as pedestrians, vehicles, etc.).

[0005] Sweeping robots based on real-time obstacle avoidance: These robots are equipped with sensors (such as ultrasonic or lidar) to sense obstacles in real time and avoid them. However, due to the lack of global path planning ability, their cleaning process often has the problem of local optimization, which easily leads to circuitous paths, increased energy consumption, and low task efficiency.

[0006] Dynamic environment cleaning robots: Some robots attempt to use dynamic trajectory prediction technology to predict the movement trend of dynamic obstacles in advance, so as to adjust the traveling route. Although this method can theoretically improve the cleaning efficiency, due to the high algorithm complexity, large computational resource requirements, and imperfect modeling of dynamic obstacles, its application in actual scenarios still faces many challenges.

[0007] Prior art research shows that traditional garbage cleaning robots still have the following main problems in complex environments:

[0008] Lack of the ability to combine global and local path planning: Fixed path planning cannot cope with the complexity of dynamic environments, while a single real-time obstacle avoidance strategy is prone to lead to unreasonable local paths and low overall cleaning efficiency.

[0009] Insufficient handling of dynamic obstacles: Existing robots have limited perception of dynamic obstacles, unable to predict the movement trajectories of obstacles in advance, and the obstacle avoidance behavior is often lagged, which may cause safety problems.

[0010] Insufficient energy consumption optimization: Due to the lack of path optimization algorithms, existing robots are prone to redundant movements, resulting in unnecessary increased energy consumption. Summary of the Invention

[0011] To solve the technical problems existing in the prior art, namely, the existing dynamic environment cleaning robots lack the path planning ability that combines the global and local aspects and insufficient handling of dynamic obstacles, the technical solutions provided by the present invention include:

[0012] A method for planning the cleaning route of a campus garbage picking robot, including:

[0013] Steps of constructing a time series input of dynamic obstacles based on sensor data and generating a predicted reachable area of dynamic obstacles;

[0014] Steps of constructing a synthetic potential field including a static potential field, a dynamic potential field, and a gravitational field according to the predicted reachable area of the dynamic obstacles;

[0015] Steps of generating a global path according to the synthetic potential field and dynamically updating the global path based on real-time sensor data.

[0016] Furthermore, a preferred embodiment is provided. In the predicted reachable area of the dynamic obstacles, the future trajectories of the dynamic obstacles are predicted through a temporal convolutional network and a bidirectional GRU model.

[0017] Furthermore, a preferred embodiment is provided. The static potential field calculates the repulsive force value through the geometric features of the obstacles.

[0018] Furthermore, a preferred embodiment is provided. The dynamic potential field generates a repulsive area using the predicted trajectories of the dynamic obstacles and dynamically adjusts the weights.

[0019] Furthermore, a preferred embodiment is provided. The gravitational field generates a smooth gravitational gradient through the control of a Lyapunov function.

[0020] Furthermore, a preferred embodiment is provided. A global path is generated through the A* algorithm, and the local path is optimized in combination with the dynamic window method, and the path is dynamically updated based on real-time sensor data.

[0021] Based on the same inventive concept, the present invention also provides a campus garbage picking robot, which includes a mobile platform and a manipulator. The mobile platform is used to carry the manipulator and enable the robot to move freely in the campus environment. The manipulator is used to pick up garbage and place it into the garbage collection device on the mobile platform. The robot is used to implement the method described above, including:

[0022] The mobile platform includes a rear-wheel drive system and a front-wheel steering system. The rear-wheel drive system is driven by a stepping motor through a chain drive to rotate the rear wheels. The front-wheel steering system is connected to the stepping motor through a coupling based on the Ackerman steering geometry principle to achieve precise steering;

[0023] The manipulator is installed on the top of the mobile platform and includes a large arm, a small arm, and a gripper. The large arm and the small arm are rotated through gear transmission. The gripper is driven to lift through a lead screw mechanism.

[0024] Based on the same inventive concept, the present invention also provides a computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes the method described above.

[0025] Based on the same inventive concept, the present invention also provides a computer, which includes a processor and a storage medium. When the processor reads the computer program stored in the storage medium, the computer executes the method described above.

[0026] Based on the same inventive concept, the present invention also provides a computer program product. As a computer program, when the computer program is executed, the method described above is implemented.

[0027] Compared with the prior art, the beneficial effects of the technical solution provided by the present invention are as follows:

[0028] By combining motion prediction technology and real-time obstacle avoidance strategies, the path planning ability of the garbage picking robot in a dynamic environment is significantly improved. Compared with traditional robots that only rely on real-time obstacle avoidance, this solution uses a temporal convolutional network (TCN) and a bidirectional GRU model to predict the future trajectories of dynamic obstacles, making the obstacle avoidance behavior more forward-looking and intelligent, reducing the lag problem in path adjustment, and greatly improving the cleaning efficiency and safety of the robot in a dynamic campus environment.

[0029] Based on the improved artificial potential field method, a dynamic potential field is generated for the predicted area of dynamic obstacles and weighted synthesized with the static potential field and the gravitational field, realizing the organic combination of global path planning and local path optimization. Compared with traditional fixed path planning methods, this method can dynamically adjust the cleaning path, avoiding stagnation or detouring caused by environmental changes, and ensuring the continuity and stability of the cleaning task.

[0030] The A* algorithm is used for global path planning, combined with the Dynamic Window Approach (DWA) to achieve local path optimization, enabling the robot to balance global goals and local environmental adaptability. Compared with robots that solely rely on global path planning, this solution can better adapt to sudden obstacles in complex environments, effectively reducing redundant movements, improving the rationality of path planning, and reducing energy consumption.

[0031] Through the integration of the sensing module and environmental modeling technology, high-precision static and dynamic obstacle perception is achieved. Compared with solutions that rely on a single sensor, this technology uses multi-modal data fusion of RGB-D cameras, lidar, and ultrasonic sensors to significantly improve the comprehensiveness and accuracy of environmental information collection, providing reliable basic data for subsequent path planning and obstacle avoidance strategies.

[0032] Using a stepper motor combined with a PID control algorithm ensures the accuracy and stability of path tracking during the robot's execution. Compared with traditional simple speed control methods, this method significantly reduces yaw or repetitive movements caused by error accumulation during the cleaning process, further improving the operation efficiency.

[0033] It is applicable to garbage cleaning and collection work in complex environments. Brief Description of the Drawings

[0034] Figure 1 It is a flowchart of the path planning method.

[0035] Figure 2 It is a schematic diagram of the mechanism of a campus garbage picking robot.

[0036] Figure 3 It is a three-dimensional schematic diagram of the manipulator. Detailed Embodiments

[0037] To make the advantages and benefits of the technical solution provided by the present invention more clearly demonstrated, the technical solution provided by the present invention will be further described in detail below in conjunction with the accompanying drawings. Specifically:

[0038] Embodiment 1. This embodiment provides a cleaning route planning method for a campus garbage picking robot, including:

[0039] Steps of constructing a time series input of dynamic obstacles based on sensor data and generating a predicted reachable area of dynamic obstacles;

[0040] Steps of constructing a synthetic potential field including a static potential field, a dynamic potential field, and a gravitational field according to the predicted reachable area of the dynamic obstacles;

[0041] Steps of generating a global path according to the synthetic potential field and dynamically updating the global path based on real-time sensor data.

[0042] In the predicted reachable area of the dynamic obstacle, the future trajectory of the dynamic obstacle is predicted through a temporal convolutional network and a bidirectional GRU model.

[0043] The static potential field calculates the repulsive force value through the geometric features of the obstacle.

[0044] The dynamic potential field generates a repulsive region using the predicted trajectory of the dynamic obstacle and dynamically adjusts the weight.

[0045] The gravitational field controls the generation of a smooth gravitational gradient through a Lyapunov function.

[0046] A global path is generated through the A* algorithm, and the local path is optimized in combination with the dynamic window method, and the path is dynamically updated based on real-time sensor data.

[0047] As Figure 2 and 3 shown, a campus garbage picking robot is also provided, including a mobile platform and a manipulator. The mobile platform is used to carry the manipulator and realize the free movement of the robot in the campus environment. The manipulator is used to pick up garbage and put it into the garbage collection device on the mobile platform. The robot is used to implement the method described in claim 1, including:

[0048] The mobile platform includes a rear-wheel drive system and a front-wheel steering system. The rear-wheel drive system drives the rear wheels to rotate through a chain drive by a stepper motor. The front-wheel steering system is connected to the stepper motor through a coupling based on the Ackermann steering geometry principle to achieve precise steering;

[0049] The manipulator is installed on the top of the mobile platform and includes a large arm, a small arm and a gripper. The large arm and the small arm realize rotation through gear transmission, and the gripper is driven to lift through a lead screw mechanism.

[0050] Specifically, it includes:

[0051] 1. Overall shape

[0052] The overall robot consists of a mobile platform and a manipulator, mainly designed for campus garbage picking. The mobile platform carries the manipulator and realizes free movement through rear-wheel drive and front-wheel steering, while the manipulator has flexible grasping and dropping capabilities for picking up target garbage.

[0053] 2. Mobile platform

[0054] Wheel structure: The mobile platform adopts a four-wheel design, where the rear wheels are driving wheels driven by a stepper motor through a chain drive; the front wheels are steering wheels designed with Ackermann steering geometry to ensure flexible steering ability.

[0055] Vehicle body frame: The main frame of the platform uses cast aluminum material, which not only ensures structural strength but also reduces weight. The vehicle body is designed compactly and is suitable for complex campus environments.

[0056] Transmission system: The rear-wheel drive system transmits the rotation of the motor to the rear axle through a sprocket mechanism and drives the wheels to rotate. The sprocket structure design effectively disperses the burden on the axle. The center distance and the number of teeth are reasonably configured to ensure smooth transmission.

[0057] 3. Manipulator structure

[0058] Overall layout: The manipulator is installed on the top of the mobile platform. The structure is columnar and has three degrees of freedom, including the rotation of the upper arm joint, the lower arm joint, and the lifting function of the gripper. The entire manipulator structure is controlled by a stepper motor, and the driving gear completes the movement of each component.

[0059] Upper arm design: The upper arm is connected to the base through a hinge and can rotate within a range of 240°. It is made of aluminum alloy material, combining strength and lightness.

[0060] Lower arm design: The lower arm is a hollow columnar structure, and the joint rotation is realized by gear transmission. The hollow design improves the bending resistance performance and reduces the structural weight.

[0061] Gripper structure: The gripper part adopts a clamping design, consisting of a two-finger rotary type. The maximum clamping range is 80mm, which is suitable for picking up common garbage (such as bottles and cans, etc.). The shape of the gripper is V-shaped to adapt to the shapes of various objects.

[0062] Drive system: Each joint of the manipulator is driven by a stepper motor, and the gripper lifting is realized in cooperation with a ball screw, featuring high precision and low error.

[0063] 4. Detail features

[0064] Base: The base of the manipulator adopts a cast aluminum structure and is fixedly connected to the mobile platform, providing stable support for the manipulator.

[0065] Steering mechanism: The steering mechanism of the front wheels is connected to the stepper motor through a coupling and realizes precise steering according to the Ackermann geometry principle.

[0066] Garbage collection function: There is a trash can on the side of the mobile platform, and the manipulator accurately puts the picked-up garbage into the trash can.

[0067] Structure optimization: The overall design focuses on simplifying the structure, reducing weight, and at the same time improving the modularization degree, which is convenient for later maintenance and upgrading.

[0068] Among them, the manipulator specifically is:

[0069] Overall Structure: The manipulator adopts a clamping design and realizes the functions of grasping and releasing objects through the linkage of two grippers. The overall design is compact and suitable for installation at the end of the robot to pick up target objects of various shapes.

[0070] Drive Mechanism: The manipulator is driven by a gear transmission system. The main gear is driven by a motor and meshes with the secondary gear to achieve the synchronous opening and closing movement of the grippers. The gear transmission ensures the accuracy and stability of the gripper movements.

[0071] Linkage System: The gripper part of the manipulator is connected to the gear system through multiple linkages. The linkage mechanism transmits power to enable the grippers to achieve smooth opening and closing movements, while enhancing the flexibility during the grasping process.

[0072] Gripper Design: The gripper has a two-finger structure, and the end is designed to adapt to the grasping of objects of various shapes. The distance between the grippers can be adjusted according to the size of the target object, making it suitable for grasping lightweight garbage such as bottles, cans, and papers in the campus environment.

[0073] Materials and Installation: The manipulator is made of lightweight and high-strength materials, which is convenient for the robotic arm to carry and reduces the total weight at the same time. The manipulator is connected to the robotic arm of the robot through a fixed installation interface at the bottom to ensure overall stability and grasping accuracy.

[0074] Embodiment 2. Combination Figures 1-3 To describe this embodiment, this embodiment provides a method for planning the cleaning route of a campus garbage collection robot that combines motion prediction technology and real-time obstacle avoidance strategy. Through the static and dynamic modeling of the environment, the combination of global path planning and local path optimization is achieved, improving the cleaning efficiency of the robot in the dynamic campus environment. The entire implementation process includes five main steps: environmental modeling and data processing, motion prediction, real-time obstacle avoidance strategy, cleaning path planning and real-time update, and execution control.

[0075] 1. Environmental Modeling and Data Processing (Input: Static and Dynamic Obstacles in the Environment)

[0076] Collect environmental information through the sensing module to establish an environmental model of static and dynamic obstacles. First, use lidar to obtain the point cloud data of static obstacles and generate a static grid map of the campus through SLAM technology. Then, use an RGB-D camera to identify the speed and direction information of dynamic obstacles and record their positions and motion states. Finally, filter the data collected by the sensor, remove noise, normalize it to a unified coordinate system, and convert it into a grid map format for use by subsequent modules.

[0077] 2. Motion Prediction (Input: Positions, Speeds, and Time Series of Dynamic Obstacles)

[0078] Predict the motion trajectory of dynamic obstacles to enhance the foresight of obstacle avoidance behavior. Input the time series data of dynamic obstacles into a Temporal Convolutional Network (TCN) to extract its temporal features. Subsequently, use a bidirectional GRU model to predict the future multi-step motion trajectory of the obstacle, generate the reachable area of the dynamic obstacle, and dynamically adjust the priority of the prediction area through the elliptical envelope method. The prediction results are used as the input for the real-time obstacle avoidance strategy.

[0079] 3. Real-time obstacle avoidance strategy (Input: environmental model and predicted trajectory of dynamic obstacles)

[0080] Based on the improved artificial potential field method, construct a synthetic potential field of static potential field, dynamic potential field, and gravitational field. The static potential field calculates the repulsive force value through the geometric features of the obstacle and the Sigmoid function; the dynamic potential field is generated using the predicted area of the dynamic obstacle and its weight decreases with the time step; the gravitational field generates a gradient gravitational value based on the position of the target point. After the three potential fields are weighted and synthesized, the obstacle avoidance strategy of the robot is adjusted in real time to ensure the safety and continuity of the cleaning task.

[0081] 4. Cleaning path planning and real-time update

[0082] Achieve efficient planning of the robot's cleaning path through the combination of global path planning and local path optimization. First, use the A* algorithm to generate the global optimal path on the static map; subsequently, optimize the local path in combination with the Dynamic Window Approach (DWA) and adjust the obstacle avoidance area according to the real-time environmental information. The planning result is dynamically corrected under the update of the sensor 10 times per second. If the trajectory deviates from the predicted value, a new dynamic potential field is generated and the path is updated.

[0083] 5. Execution control

[0084] The robot performs the cleaning task according to the planned path. Control the speed and direction of the mobile platform through the stepper motor and servo system, and achieve high-precision path tracking in combination with the PID control algorithm. The manipulator is responsible for picking up garbage and completes garbage grasping and throwing using a three-degree-of-freedom joint design. If an emergency obstacle is detected, trigger the emergency obstacle avoidance mode, bypass the obstacle first and then continue to execute the task.

[0085] Among them: Environmental modeling and data processing: The modeling of environmental information is the basis of the entire system. Through the cooperation of lidar and camera, not only can a static map be generated, but also the shape, speed, and direction of dynamic obstacles can be identified. The grid map generated in combination with SLAM technology is updated in real time to ensure that the environmental model can reflect the changes in the current scene. The data filtering step uses the Kalman filter algorithm to enhance the reliability of the data.

[0086] Motion Prediction: The motion trajectory of dynamic obstacles is predicted using time series analysis methods to ensure the accuracy of future trajectory points. The TCN and GRU models are trained using the Adam optimizer, and the mean squared error (MSE) is used as the loss function. The predicted elliptical region gives priority to recent trajectories to avoid redundant calculations introduced by long-term information.

[0087] Real-time Obstacle Avoidance Strategy: The design of the gravitational field in the potential field function is based on the Lyapunov function to ensure path smoothness. The dynamic potential field uses a custom decreasing weight function to gradually reduce the priority of the predicted trajectory over time. The conflict resolution mechanism is used to prioritize safety when synthesizing the potential field.

[0088] Sweeping Path Planning and Real-time Update: The path update mechanism introduces a real-time weight adjustment function to make the fusion result of the global path and the local path adapt to the dynamic changes of the environment. The multi-threaded architecture ensures the synchronous operation of path planning and sensor data update, improving the real-time performance and response speed of the overall system.

[0089] Execution Control: The stepping motor used in the execution module is combined with a chain drive structure, which has high control accuracy and stability. The grasping action of the manipulator is achieved through point position control, and the lifting mechanism uses a ball screw design, which can not only improve the lifting accuracy but also reduce the volume and weight of the system.

[0090] Embodiment 3. Combination Figure 1 Describe this embodiment. This embodiment further describes the above-provided technical solution in detail through specific embodiments. Specifically:

[0091] 1. Technical Overview

[0092] The future trajectory of dynamic obstacles is predicted through motion prediction technology, and the sweeping path is dynamically adjusted in combination with a real-time obstacle avoidance strategy to form a path planning system based on the combination of global and local. In the sweeping task, the robot can not only autonomously plan an efficient global sweeping path but also avoid dynamic obstacles in real time to ensure the continuity and safety of the task.

[0093] 2. System Composition

[0094] (1) Sensing Module:

[0095] Implementation Method: Configure cameras (such as RGB-D cameras), lidar, and ultrasonic sensors, and connect them to the robot controller using the CAN bus.

[0096] Function: Real-time collection of environmental information, including the positions of static obstacles and data such as the speed and direction of dynamic obstacles.

[0097] (2) Motion Prediction Module:

[0098] Implementation method: Build a Temporal Convolutional Network (TCN) and a bidirectional GRU model based on a deep learning framework (such as TensorFlow or PyTorch).

[0099] Training data: Collect the motion trajectory data of pedestrians, vehicles, etc. in the campus environment to construct a time series training set.

[0100] Function: Input the motion state of dynamic obstacles and output their future reachable areas.

[0101] (3) Real-time obstacle avoidance module:

[0102] Implementation method: Based on the improved artificial potential field method, construct the repulsive force fields of dynamic and static obstacles, and calculate the total potential field by combining the gravitational field.

[0103] Algorithm: Implement a custom dynamic adjustment function to perceive and update the repulsive weights of obstacles in real time.

[0104] (4) Path planning module:

[0105] Implementation method: Use the A* algorithm based on graph search for global path planning, and combine it with DWA (Dynamic Window Approach) to achieve local path optimization.

[0106] Function: Generate a globally optimal path and dynamically update the obstacle avoidance area.

[0107] (5) Execution control module:

[0108] Implementation method: Adopt a stepper motor and a servo system to achieve path tracking through an embedded controller (such as STM32).

[0109] Function: Adjust the speed and direction of the robot according to the planned path.

[0110] 3. Specific implementation methods

[0111] (1) Environment modeling and data processing

[0112] Steps for environment modeling:

[0113] Use a lidar to obtain static environment point cloud data.

[0114] Generate a campus static grid map through SLAM (Simultaneous Localization and Mapping) technology.

[0115] Cooperate with a camera to identify dynamic targets and record their positions and speeds.

[0116] Tools: Use an open-source SLAM framework (such as Cartographer or RTAB-Map).

[0117] Steps for data processing:

[0118] Filter the raw data collected by the sensor to remove noise.

[0119] Use the normalization algorithm to map the sensor data to a unified coordinate system.

[0120] Convert it into the format of an environmental grid map for use by subsequent modules.

[0121] (2) Motion prediction

[0122] Steps for predicting the trajectory of dynamic obstacles:

[0123] Input the time series of dynamic obstacles into the Temporal Convolutional Network (TCN).

[0124] After extracting its temporal features, use a bidirectional GRU to predict the future multi-step motion trajectory.

[0125] Implementation: Use the Adam optimizer during model training, and the loss function is the mean squared error (MSE). Specifically:

[0126] Prediction of the trajectory of dynamic obstacles

[0127] Input of time series

[0128] Let the historical trajectory of the dynamic obstacle be:

[0129] S = {(x i , y i , υ xi , υ yi , t i )}, i = 1, 2,..., T

[0130] where (x i , y i ) represents the position, (υ xi , υ yi ) represents the velocity components, and t i is the timestamp.

[0131] TCN feature extraction: Input the time series into the Temporal Convolutional Network (TCN) to extract temporal features:

[0132] H t = f TCN (S)

[0133] where H t represents the feature vector at time t.

[0134] GRU prediction of the future trajectory

[0135] Input the extracted features into a bidirectional GRU to predict the future multi-step trajectory points:

[0136]

[0137] where represents the predicted point at the k-th step in the future.

[0138] Elliptical dynamic region generation: Generate the elliptical region of the dynamic obstacle based on the predicted point:

[0139]

[0140] where a k and b k are the major axis and minor axis of the ellipse respectively, and are dynamically adjusted according to the speed and direction of the obstacle. Steps for dynamic region generation:

[0141] Generate a multi-step elliptical envelope potential field according to the predicted trajectory.

[0142] Dynamically adjust the weights to make the recent region have a higher priority.

[0143] Tool: Use NumPy and Matplotlib for real-time ellipse drawing.

[0144] (3) Real-time obstacle avoidance strategy

[0145] Steps for constructing the static potential field:

[0146] Calculate the repulsive force value of the static obstacle through the Sigmoid function.

[0147] Assign different weights to the geometric features of the obstacle.

[0148] Implementation: Use a custom function to implement the dynamically adjusted Sigmoid.

[0149] Steps for constructing the dynamic potential field:

[0150] Define the repulsive force field of the dynamic obstacle using the predicted elliptical region.

[0151] Reduce the repulsive weight step by step with the time step to reduce the influence of the long-term trajectory on the current plan.

[0152] Steps for constructing the gravitational field:

[0153] Generate a gravitational field for the target point.

[0154] Use the logarithmic Lyapunov function to control the gravitational gradient.

[0155] Implementation: Define a dynamically adjusted function for the target point to ensure a smooth path.

[0156] Steps for synthesizing the total potential field:

[0157] Weighted synthesis of the dynamic potential field, static potential field, and gravitational field.

[0158] Define a conflict resolution mechanism with safety as the top priority.

[0159] Specifically:

[0160] Dynamic potential field generation: Utilize the prediction area of dynamic obstacles to construct a dynamic potential field:

[0161]

[0162] where: (x k , y k ) is the center point of the dynamic obstacle; w k is the weight of the obstacle, which is dynamically adjusted: t k is the prediction time step, α controls the weight decay rate, t is the current time; ∈ is a small value to avoid a zero denominator.

[0163] Static potential field generation

[0164] Construct a repulsive potential field for static obstacles:

[0165]

[0166] where: represents the distance between the robot and the i-th static obstacle.

[0167] Gravitational field generation, generate a gravitational field for the target point:

[0168]

[0169] where λ is the gravitational field strength coefficient.

[0170] Total potential field synthesis, synthesize the dynamic potential field, static potential field, and gravitational field through weighting:

[0171] F total (x,y) = β · F dynamic (x,y) + γ · F static (x,y) + F attract (x,y)

[0172] where: β and γ are the weight parameters of the dynamic potential field and static potential field respectively, which are adjusted according to the real-time environment.

[0173] Obstacle avoidance decision

[0174] Calculate the robot's movement direction based on the total potential field:

[0175]

[0176] where θ is the optimal direction for the robot to travel.

[0177] (4) Cleaning Path Planning and Real-time Update

[0178] Path planning steps:

[0179] Perform global A* path search on the static map.

[0180] Segment the path and optimize the local path in combination with DWA.

[0181] Implementation: Combine the ROS MoveBase framework to implement path planning.

[0182] Real-time update steps:

[0183] Obtain the obstacle status updated by the sensor every 10Hz.

[0184] If the trajectory deviates from the predicted value, regenerate the dynamic potential field.

[0185] Tool: Run the real-time controller node and use multi-threading to handle planning and sensing.

[0186] (5) Execution Control

[0187] Path execution steps:

[0188] Adjust the robot's speed and direction according to the planned path.

[0189] Use the PID control algorithm to ensure path tracking accuracy.

[0190] Tool: Embed the code for implementing PID control into STM32.

[0191] Exception handling steps:

[0192] Trigger the emergency obstacle avoidance mode when detecting an emergency obstacle.

[0193] Prioritize finding the shortest path to bypass the obstacle.

[0194] Specifically:

[0195] Global path planning (optimized by A* algorithm), the global path planning is based on the A* algorithm and introduces a dynamic cost adjustment mechanism:

[0196] f(n) = g(n) + h(n) + δ(n)

[0197] Where: f(n): The total cost of node n. g(n): The actual path cost from the starting point to node n. h(n): The heuristic cost from node n to the target point, calculated using the Euclidean distance: δ(n) is the dynamic cost, used to consider the influence of dynamic obstacles: Where d k is the distance between node n and dynamic obstacle k, ωk is the obstacle weight.

[0198] Local path optimization (Dynamic Window Approach, DWA), using the Dynamic Window Approach (DWA) to optimize the local path, with the goal of maximizing the comprehensive score of the robot's movement:

[0199]

[0200] where: W: the set of velocities within the dynamic window. heading(V): the degree of alignment between the current velocity direction and the target direction: where θ err is the angle between the velocity direction and the target direction; clearance(V): the minimum distance to the nearest obstacle in the velocity direction: D(V) is the set of obstacle distances in the velocity direction; υelocity(V): the magnitude of the current velocity, a negative cost term.

[0201] Real-time update mechanism, updating the path planning in real time during the robot's movement, calculating the update trigger conditions:

[0202]

[0203] where: (x curr , y curr ): the current robot position. (x pred , y pred ): the target position obtained based on motion prediction. If ΔP > ∈, trigger re-planning of the path, where ∈ is the allowable deviation threshold.

[0204] Fusion of dynamically updated paths, by weighted fusion of the global path and the locally optimized path:

[0205] P final (t) = λP global (t) + (1 - λ)P local (t)

[0206] where: P global (t): the result of global path planning. P local (t): the result of local path optimization. λ: the dynamic adjustment coefficient, adjusting the weights of the global and local paths according to the real-time environment.

[0207] Structural part: The campus garbage collection robot in this embodiment is composed of a mobile platform and a manipulator. Its specific working principle is that the mobile platform carries the manipulator and walks on the campus road surface. It identifies the garbage on the road surface through cameras and sensors. Then the mobile platform moves to the position where the garbage is located. After stopping, the manipulator picks up the garbage and puts it into the trash can on the side of the mobile platform.

[0208] The mobile platform adopts a rear-wheel drive and front-wheel steering mode. The rear wheels are driven by a stepping motor that drives a chain drive and then drives the rotation of the shaft. At the front wheels, there is a steering mechanism designed according to the Ackermann steering geometry principle, and the rotation of the stepping motor drives the rotation of the wheels.

[0209] To enable the garbage collection robot to move freely on the campus ground, the mobile platform part adopts rear-wheel drive and front-wheel steering. The structural schematic diagram is as follows. Rear-wheel drive is used to control the front-wheel steering. The wheel rotation structure can be specifically divided into the following four parts:

[0210] 1. Two deep groove ball bearings are selected, axially positioned by a retaining ring and the shoulder of the motor housing shaft, and radially positioned by the outer surface of the shaft.

[0211] 2. The motor is installed on the motor bracket, and then the motor and the motor bracket are fixed to the vehicle body together. The positioning method is to rely on the edge of the motor bracket for positioning.

[0212] 3. The wheels are positioned by bearings, and then connected to the coupling and the steering shaft through the wheels. This process requires adjusting the concentricity of the motor shaft, and then fastening the motor to the vehicle frame through the reserved mounting holes on the vehicle frame.

[0213] 4. The whole wheel is divided into a wheel with a shaft diameter and a wheel without a shaft diameter. The two need to be used in combination to form a complete wheel. The wheel shaft diameter can be used as the axial positioning standard for the two wheels. The wheel is fixed by connecting the outer side screw with the tightening baffle.

[0214] Overall design scheme of the manipulator: Generally speaking, the main functions of the manipulator are as follows: improving the working conditions, reducing the possibility of personal accidents, ensuring the quality of the products produced, reducing the consumption of manpower and material resources, facilitating efficient and orderly production, and improving the degree of automation of the production process. In this embodiment, the upper part of the campus garbage collection robot is the manipulator. The function of the manipulator is to pick up the garbage and put it into the trash can on the side of the mobile platform after the mobile platform moves to a suitable position, completing the task of garbage collection. In order to complete the task, the designed manipulator has three degrees of freedom: (1) the rotation of the big arm joint; (2) the rotation of the small arm joint; (3) the lifting of the gripper.

[0215] Due to the large population flow on campus, in order to facilitate the work of the manipulator, a more flexible manipulator with a large working range needs to be designed. And since the volume and weight of the garbage are not too large, the designed manipulator is as simple, lightweight and compact as possible.

[0216] Therefore, the following basic parameters are selected:

[0217] 1. Gripping weight: maximum 1 kg;

[0218] 2. Number of degrees of freedom: 3;

[0219] 3.Structural type: articulated;

[0220] 4. Upper arm length: 400mm;

[0221] 5. Arm length: 320mm;

[0222] 6. Driving mode: stepper motor;

[0223] 7. Control mode: point program control;

[0224] 8. Clamping range of the claw: up to 80mm;

[0225] 9. Rotation angle of upper arm and middle arm joint: 240°.

[0226] Because the manipulator in this embodiment has three degrees of freedom, a variety of structural types of combinations can be used. This embodiment adopts a joint-type design, and each manipulator arm is hinged and can rotate freely around the shoulder and elbow. It is very dexterous, has a large working range, and is very practical.

[0227] The drive system is a key link in the design of the robot. According to the different power sources, the common robot drive systems are mainly divided into four categories: hydraulic drive, pneumatic drive, motor drive and mechanical drive. Commonly used drives are motor drive, hydraulic cylinder drive and pneumatic drive. Among them, motor drive is the most common drive method, which has the characteristics of high accuracy, strong controllability and large speed range. Therefore, this robot chooses an electric mechanism as the drive, and the commonly used motor drive methods are as follows:

[0228] 1. Stepper motor: high precision, strong controllability, can stop and start at any time, and has good positioning.

[0229] 2. DC servo motor: DC servo motor has good speed regulation characteristics, quick response, good stability, etc. It is also easy to maintain and disassemble with low cost.

[0230] 3. Direct drive: The motor is directly connected to the element it drives, without the use of a reduction device to slow down, which has a greater impact on the life of the motor.

[0231] After analysis, the manipulator of this embodiment is finally driven by a stepper motor because of its simple structure, good positioning, easy disassembly, low cost and high working efficiency.

[0232] The configuration form of the robotic arm is an important part of the robotic arm layout. According to the design requirements, to achieve the picking-up function of the robotic arm, its working range needs to be wide enough. Therefore, the column type is finally selected as the configuration scheme, which does not occupy too much space and can have a large working range.

[0233] Combined with the design requirements and the required technical requirements, the structural design scheme of the robotic arm is formulated as follows:

[0234] The robotic arm in this embodiment is required to have a simple structure and flexible movement, so the number of degrees of freedom is set to 4. They are respectively (1) the rotation of the upper arm joint; (2) the rotation of the forearm joint; (4) the lifting of the gripper part.

[0235] Working principle: The stepping motor at the upper arm joint drives the gear transmission to realize the rotation of the upper arm; the stepping motor at the forearm joint drives the gear transmission to realize the rotation of the forearm; the hand is lifted by a single-axis transmission for the gripper. The main structure of the robotic arm consists of the upper arm, forearm, single-axis driver, and hand clamping mechanism. The base is an integral casting made of cast aluminum, and the other joints are made of cast aluminum, and a main shaft for driving the joint rotation and a matching bearing are configured therein.

[0236] Selection of the robotic arm material and weight estimation: Since the function of the robotic arm is to pick up garbage, the grasping weight is very low. Also, because the garbage-picking robot works on campus where there is a large flow of people and complex road conditions, sufficient strength is required. And to increase the working speed, the material should be as light as possible. Moreover, when the robotic arm material bears a certain load, it should not undergo serious deformation and fracture. Therefore, the material must have a certain strength, so mechanical materials such as cast aluminum or aluminum alloy should be selected. Moreover, since the arm is working all the time, the arm structure must be easy to control. So according to the situation

[0237] Generally speaking, materials with sufficient strength and relatively low density should be preferred. By consulting relevant materials, finally, cast aluminum is selected as the material for the robotic arm. It not only has a certain strength but is also very light, which is convenient for future disassembly and maintenance. The weight estimates of the upper arm, middle arm, and forearm are 2 kg, 4 kg, and 1 kg respectively.

[0238] Motor selection: Select the 57BYGH56 model motor. The static torque is 1.2 N·m. This motor can still maintain sufficient torque stability when the speed rises.

[0239] The main significance of this chain drive is to transfer the rotation of the motor to the shaft. At the same time, there is no need to place the motor on the shaft to reduce the bearing load of the shaft. Therefore, the rotational speed of the small sprocket is 180 r / min, and the rotational speed of the large sprocket is also 180 r / min. The center distance is not less than 250 mm. Large and small sprockets with 17 teeth each are selected. The chain pitch p = 12.7 mm, 08A single-row chain, 67 links, the number of sprocket teeth Z 1 = Z 2 = 17, the center distance a 1 = 292.1 mm, the axial pressure F Q = 93.75 N.

[0240] The cross-section of the forearm is designed as a hollow column, which can not only increase the bending resistance coefficient but also reduce the total cross-sectional area, thus reducing the overall weight and making it economical and lightweight.

[0241] Cast aluminum of ZL102 model is selected. The theoretical weight is 11.261 kg / m, W y = 9.72 cm 3 , h = 100 mm, d = 4.5 mm, and the forearm length is 320 mm. It is appropriate to choose ZL102 cast aluminum as the material.

[0242] The material of the upper arm is aluminum alloy. Because aluminum alloy is not only lightweight but also can bear a certain load, which simplifies the overall structure and reduces the weight. It is appropriate to choose ZL102 cast aluminum.

[0243] The driving force of the manipulator is not large and there are no excessive requirements for precision. Therefore, after consideration, a stepping motor with a simple structure and a small accumulated error of the output rotation angle is finally selected. The stepping motor is 45BF005.

[0244] The lifting mechanism uses a single-axis drive, that is, the principle of a ball screw inside to control the lifting.

[0245] The clamping type hand structure is divided into two parts: fingers and a force transmission mechanism. The structure of the manipulator used is a two-finger rotary type. Since the workpieces are generally irregularly shaped or cylindrical beverage bottles, the finger shape is designed as a V shape.

[0246] Design parameters of the gripper:

[0247] A finger-type gripper is used, and the actions are grasping → releasing

[0248] (1) The maximum diameter of the workpiece to be grasped is 60 mm;

[0249] (2) The maximum distance between the two claws when releasing is 80 mm;

[0250] (3) It takes 1 s to grasp, and the clamping speed is 10 - 20 mm / s;

[0251] (4) The maximum weight of the workpiece is 0.5 kg.

[0252] The technical solutions provided by the present invention are further described in detail through several specific embodiments to highlight the advantages and beneficial effects of the technical solutions provided by the present invention. However, the several specific embodiments described above are not used as limitations on the present invention. Any reasonable modifications and improvements to the present invention, combinations of implementation manners, equivalent replacements, etc. within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A cleaning route planning method for a campus garbage picking robot, characterized in that: include: The step of constructing a time series input of a dynamic obstacle based on the sensor data and generating a predicted reachable area of ​​the dynamic obstacle; A step of constructing a synthetic potential field including a static potential field, a dynamic potential field and a gravitational field according to the predicted reachable area of ​​the dynamic obstacle; The steps of generating a global path according to the synthetic potential field and dynamically updating the global path based on real-time sensor data.

2. The cleaning route planning method of a campus garbage picking robot according to claim 1 is characterized in that: In the predicted reachable area of ​​the dynamic obstacle, the future trajectory of the dynamic obstacle is predicted by using a temporal convolutional network and a bidirectional GRU model.

3. The cleaning route planning method of a campus garbage picking robot according to claim 1 is characterized in that: The static potential field calculates the repulsive force value through the geometric characteristics of the obstacle.

4. The cleaning route planning method of a campus garbage picking robot according to claim 1 is characterized in that: The dynamic potential field generates an exclusion zone using the predicted trajectory of the dynamic obstacle and dynamically adjusts the weight.

5. The cleaning route planning method of a campus garbage picking robot according to claim 1 is characterized in that: The gravitational field is controlled by a Lyapunov function to generate a smooth gravitational gradient.

6. The cleaning route planning method of a campus garbage picking robot according to claim 1, characterized in that: The global path is generated through the A* algorithm, and the local path is optimized in combination with the dynamic window method, and the path is dynamically updated based on real-time sensor data.

7. A campus garbage picking robot, comprising a mobile platform and a manipulator, wherein the mobile platform is used to carry the manipulator and realize the free movement of the robot in the campus environment, and the manipulator is used to pick up garbage and put it into the garbage collection device on the mobile platform, characterized in that: The robot is used to implement the method according to claim 1, comprising: The mobile platform includes a rear-wheel drive system and a front-wheel steering system. The rear-wheel drive system is driven by a stepper motor through a chain drive to drive the rear wheel to rotate. The front-wheel steering system is connected to the stepper motor through a coupling based on the Ackerman steering geometry principle to achieve precise steering. The manipulator is installed on the top of the mobile platform, and comprises an upper arm, a lower arm and a gripper. The upper arm and the lower arm are rotated by gear transmission, and the gripper is driven to rise and fall by a screw mechanism.

8. A computer storage medium for storing a computer program, characterized in that: When the computer program is read by a computer, the computer executes the method of claim 1 .

9. A computer, comprising a processor and a storage medium, characterized in that: When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 1 .

10. A computer program product, being a computer program, characterized in that When the computer program is executed, the method of claim 1 is implemented.

Images