RDA-based local path obstacle avoidance method

Abstract

The invention discloses an obstacle avoidance method of a local path based on RDA. Comprising the following steps: inputting sensor data, calculating an initial test path, predicting a dynamic obstacle, detecting the obstacle, adjusting a local path, checking a feasible path, executing the path, if # imgabs0 #, adjusting a path point and iteratively updating, and when the path meets safety and smoothness conditions, moving along the path # imgabs1 #. According to the invention, the local path can be detected and optimized in the real-time obstacle avoidance process, and correct response can be made in real time for obstacle avoidance.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and specifically relates to an obstacle avoidance method for a local path based on RDA. Background Art

[0002] The obstacle avoidance ability of autonomous vehicles is a key component for achieving safe and efficient travel. With the continuous development of intelligent technologies, the complexity and intelligence level of obstacle avoidance systems are also continuously improving. This ability not only concerns the autonomous safety of vehicles but also directly affects the acceptance of autonomous driving technology by society. High-precision obstacle avoidance algorithms usually require a large amount of computing resources and cannot meet the requirements of real-time processing, especially in high-speed driving or complex scenarios. In complex urban environments, such as in the case of multiple obstacles like traffic signals, pedestrians, and non-motor vehicles, they cannot make correct real-time responses. In the prior art, Chinese Patent Publication No.: CN113721637A, Publication Date: November 30, 2021, discloses an intelligent vehicle dynamic obstacle avoidance path continuous planning method, system, and storage medium. The method includes: using Frenet coordinate transformation to convert the local path planning problem when the vehicle is driving along a complex curved road into a double-shift line path planning problem under a straight road, greatly reducing the complexity of planning; and dividing the double-shift line path into an obstacle avoidance section and a return section, and performing path planning based on Bezier curves respectively; when the surrounding environment changes dynamically, especially when the vehicle is tracking the local path and an obstacle suddenly accelerates, decelerates, or undergoes a lateral displacement, the end point of the planned path will be updated accordingly, and at the same time, the local path will be replanned, and the curvature continuity at the path connection point will be ensured. The present invention can improve the real-time performance of the intelligent vehicle's planned path and help enhance the adaptability of the intelligent vehicle to complex traffic environments. This patent only smooths the path during path planning and cannot detect and optimize the local path during real-time obstacle avoidance. Summary of the Invention

[0003] The purpose of the present invention is to provide an obstacle avoidance method for a local path based on RDA that can detect and optimize the local path during real-time obstacle avoidance and make correct real-time responses for obstacle avoidance, overcoming the above-mentioned drawbacks.

[0004] An obstacle avoidance method for a local path based on RDA according to the present invention is characterized by including the following steps:

[0005] 1) Input sensor data;

[0006] The robot motion scenario is defined as a two-dimensional plane coordinate system, the starting point is S(x s , y s ), the target point is G(x g , y g ), and the obstacle set is O = {o1, o2,..., on}, where the obstacle o i is located at (x i , y i ) and has a speed of The goal is to generate a path P(t) = {(x t , y t )} from S to G while avoiding the dynamic obstacle 0 and satisfying the following conditions:

[0007] A Path smoothness:

[0008] The curvature |k(t)| satisfies the smoothness constraint |k(t)| < k max .

[0009] B Path safety:

[0010] The minimum distance d between the robot and the obstacle is > r s + r i , where r s is the radius of the robot and r i is the radius of the obstacle;

[0011] 2) Calculate the initial path;

[0012] Generate the initial path through the shortest path algorithm:

[0013] p0(t) = (1 - t)S + tG, t ∈ [0, 1]

[0014] 3) Dynamic obstacle prediction;

[0015] For each obstacle, predict its future position:

[0016]

[0017] where v is the velocity vector of the obstacle;

[0018] 4) Detect obstacles;

[0019] Real-time detect the distance between the path points and the obstacles;

[0020]

[0021] If d i (t) < r s + r i then there is a collision risk;

[0022] 5) Adjust the local path;

[0023] Use the potential field to optimize the path adjustment:

[0024] A Target attraction:

[0025]

[0026] where k att is the attraction coefficient.

[0027] B Obstacle Repulsion Force:

[0028]

[0029] where k rep is the repulsion coefficient;

[0030] C Resultant Force Path Adjustment:

[0031]

[0032] Update the path points using the gradient descent method:

[0033]

[0034] 6) Check the feasible path;

[0035] Calculate the path curvature

[0036]

[0037] Execute the path;

[0038] If |k(t)| > k max , then adjust the path points and update iteratively;

[0039] 7) Execute the adjusted path;

[0040] When the path meets the safety and smoothness conditions, move along the path p′(t).

[0041] Compared with the prior art, the present invention has obvious beneficial effects. From the above technical solutions, it can be seen that: the present invention calculates the trajectory from the current state to the target state, and this trajectory is defined as a series of control commands or states. Motion planners can be divided into four categories: graph-search-based, sampling-based, interpolation-based, and optimization-based. The graph-search-based technique is the state lattice and the A* algorithm, which regards the configuration space as a graph composed of vertices and edges and searches for the minimum-cost path. The sampling-based technique is the rapidly-exploring random tree (RRT), which probes the configuration space through a sampling scheme. The interpolation-based technique constructs a new set of data between known reference points through an interpolation model, thereby generating a feasible and smooth trajectory in a structured environment. Compared with other methods, the optimization-based technique can generate an optimized and robust trajectory in complex scenarios by minimizing a cost function under physical constraints such as dynamics, kinematics, and collision avoidance. The optimization-based navigation problem is formulated with a non-point-mass obstacle model. Through joint linearization and strong duality, the non-convex constraints are reconstructed into a bi-convex dual form. The safety distance in the collision avoidance constraint is automatically adjusted through regularization. A parallel local planner, called the regularized dual alternating direction multiplier method (RDA), is used for motion planning based on model predictive control (MPC) to solve the bi-convex problem in parallel while enhancing the robustness of algorithm convergence. The RDA algorithm has good adaptability and can effectively identify and handle different types of obstacles in various environments, improving the safety and efficiency of obstacle avoidance. It realizes the detection and optimization of the local path during real-time obstacle avoidance and makes a correct real-time reaction to avoid obstacles. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] Embodiment 1:

[0044] An obstacle avoidance method for a local path based on RDA of the present invention is characterized in that it includes the following steps:

[0045] 1) Input sensor data;

[0046] The robot motion scenario is defined as a two-dimensional plane coordinate system, the starting point is S(x s , y s ), the target point is G(x g , y g ), the obstacle set is 0 = {o1, o2,..., o n}, where the position of the obstacle 0 i is (x i , y i ), and the speed is The goal is to generate a path P(t) = {(xt , y t )}, while avoiding dynamic obstacle 0 and satisfying the following conditions:

[0047] A Path smoothness:

[0048] The curvature |k(t)| satisfies the smoothness constraint |k(t)| < k max .

[0049] B Path safety:

[0050] The minimum distance d between the robot and the obstacle is > r g + r i , where r a is the radius of the robot, and r i is the radius of the obstacle;

[0051] 2) Calculate the initial path;

[0052] Generate the initial path through the shortest path algorithm:

[0053] p0(t) = (1 - t)S + tG, t ∈ [0, 1]

[0054] 3) Dynamic obstacle prediction;

[0055] For each obstacle, predict its future position:

[0056]

[0057] where v is the velocity vector of the obstacle;

[0058] 4) Detect obstacles;

[0059] Real-time detect the distance between the path points and the obstacles;

[0060]

[0061] If, d i (t) < r a + r i Then there is a collision risk;

[0062] 5) Adjust the local path;

[0063] Use the potential field to optimize the path adjustment:

[0064] A Target attraction:

[0065]

[0066] where k att is the attraction coefficient.

[0067] B Obstacle repulsion:

[0068]

[0069] where k rep is the repulsive force coefficient;

[0070] C resultant force path adjustment:

[0071]

[0072] Update the path points using the gradient descent method:

[0073]

[0074] 6) Check the feasible path;

[0075] Calculate the path curvature

[0076]

[0077] Execute the path;

[0078] If |k(t)| > k max , then adjust the path points and iterate to update; where; Figure 1 in N is to adjust the path points and iterate to update;

[0079] 7) Execute the adjusted path;

[0080] When the path meets the safety and smoothness conditions, move along the path P′(t) to finish.

[0081] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for obstacle avoidance of a local path based on RDA, characterized in that; It includes the following steps: 1) Input sensor data; The robot motion scenario is defined as a two-dimensional plane coordinate system, with the starting point S(x s , y s ), the target point G(x g , y g ), the obstacle set O = {O1, O2,..., O n}, where the position of obstacle O i is (x i , y i ), and the speed is The goal is to generate a path P(t) = {(x t , y t )} from S to G, while avoiding dynamic obstacles O and satisfying the following conditions: A Path smoothness: The curvature |k(t)| satisfies the smoothing constraint |k(t)| < k max , B Path safety: The minimum distance d between the robot and the obstacle is d > r s + r i , where r s is the radius of the robot, and r i is the radius of the obstacle; 2) Calculate the initial path; Generate the initial path through the shortest path algorithm: p0(t) = (1 - t)S + tG, t ∈ [0, 1] 3) Dynamic obstacle prediction; For each obstacle, predict its future position: where v is the velocity vector of the obstacle; 4) Detect obstacles; Real-time detect the distance between the path points and the obstacles; If d i (t) < r s + r i then there is a risk of collision; 5) Adjust the local path; Utilize potential field to optimize path adjustment: A Target attraction: where k att is the attraction coefficient B Obstacle repulsion: where k rep is the repulsive force coefficient; C Resultant force path adjustment: Update the path points using the gradient descent method: 6) Check the feasible path; Calculate the path curvature Execute the path; If |k(t)| > k max , then adjust the path points and update iteratively; 7) Execute the adjusted path; When the path meets the safety and smoothness conditions, move along the path P′(t).

Images