Speed planning method for low-speed unmanned vehicle

Abstract

The invention relates to the technical field of unmanned vehicle speed planning, in particular to a speed planning method for a low-speed unmanned vehicle. The method comprises the steps that the maximum reference speed output by trajectory planning, the expected trajectory of a vehicle and the position and speed of an obstacle are obtained; dividing a front area and a rear area based on a vehicle coordinate system, determining a rear obstacle through a perpendicular foot ray, and calculating w through a safe distance dynamic formula; assuming that the obstacle moves linearly at a constant speed and predicting the position; calculating collision time t; correcting the collision time through a formula; and performing progressive decrease search from the maximum reference speed, and selecting and outputting the maximum reference speed which enables the minimum correction collision time tm to be greater than or equal to a preset threshold value T. According to the method, through front and rear risk grading correction and safety threshold-driven decreasing search, 'front priority and rear-end collision risk weakening 'is realized, the driving efficiency is kept to the maximum extent while the safety is ensured, and the problems of insufficient traffic rule mapping and high probability of failure in a complex scene in an existing method are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of speed planning for unmanned vehicles, and more specifically, to a speed planning method for low-speed driverless vehicles. Background Art

[0002] In the dynamic operation scenarios of low-speed driverless vehicles (such as cleaning vehicles, logistics vehicles, and sightseeing vehicles), speed planning is one of the core technologies to ensure safety and efficiency. In the prior art, common speed planning methods include: 1. The time-to-collision model, which calculates the time to collision with obstacles and adjusts the speed to avoid collisions, but has poor handling effects on lateral moving obstacles and is prone to conflicts in multi-obstacle scenarios; 2. The speed corridor method, which screens the feasible speed by calculating the infeasible speed intervals of obstacles, but may fail in complex scenarios due to an empty feasible speed space; 3. The model prediction method: predicting the positional relationship between the vehicle and obstacles through an optimization algorithm, but there are problems such as complex parameters, high engineering implementation difficulty, and possible no solution.

[0003] The above methods do not effectively distinguish the risk levels of front and rear obstacles, and have insufficient engineering mapping of traffic rules such as full liability for rear-end collisions, resulting in reduced safety and traffic efficiency of low-speed driverless vehicles in practical applications due to excessive deceleration (affected by rear obstacles) or untimely handling of collision risks (insufficient priority of front obstacles).

[0004] Therefore, a speed planning method for low-speed driverless vehicles is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a speed planning method for low-speed driverless vehicles to solve the problem of reduced safety and traffic efficiency due to excessive deceleration or untimely handling of collision risks.

[0006] To solve the above technical problems, the purpose of the present invention is to provide a speed planning method for low-speed driverless vehicles, including the following steps: S1. Obtain the maximum reference speed, the desired trajectory of the vehicle itself, and the positions and speeds of all obstacles in the scene; the maximum reference speed is output by trajectory planning, comprehensively considering road speed limits, traffic efficiency, and road curvature; S2. Calculate the attribute w for each obstacle; S3. Assume that all obstacles move in a uniform straight line, and predict the positions of each obstacle in the future for a period of time; S4. Set the vehicle speed as the current reference speed, calculate the collision time t between each obstacle and the vehicle itself. If there is no collision within the prediction time, set t = ∞; S5. Calculate the corrected collision time \(t_0 = t + w\); S6. Select the minimum value \(t_m\) among all \(t_0\). If \(t_m\geq T\) (\(T\) is the preset collision time threshold), go to S8; otherwise, go to S7; S7. Decrease the current reference speed. If the reference speed is less than 0, go to S8; otherwise, return to S4; S8. Output the reference speed that maximizes \(t_m\) as the planned speed.

[0007] As a further improvement of this technical solution, in S2, the specific steps for calculating the attribute \(w\) are as follows: S2.1: Establish a coordinate system with the vehicle center as the origin, the forward direction as the positive x-axis direction, and the positive y-axis direction to the left. Define the first and fourth quadrants as the front, and the second and third quadrants as the rear; S2.2: If the obstacle is in front of the vehicle, set \(w = 0\); S2.3: If the obstacle is behind the vehicle, calculate the foot of the perpendicular from the vehicle's position to the ray determined by the obstacle's position and its speed. If the foot of the perpendicular is not on the ray, set \(w = 0\); S2.4: If the foot of the perpendicular is on the ray, calculate the distance \(d\) between the vehicle's position and the foot of the perpendicular. Calculate the safety distance \(d_s=k_s\times v + d_{s0}\), where \(k_s\) is a positive coefficient, \(v\) is the vehicle's speed, and \(d_{s0}\) is the lower limit of the safety distance and is greater than the radius of the vehicle's circumcircle; S2.5: If \(d_s - d\geq0\), set \(w=(d_s - d)\times k\), where \(k\) is the distance weight; otherwise, set \(w = 0\).

[0008] As a further improvement of this technical solution, in S2, the risks of front and rear obstacles are distinguished by the obstacle attribute \(w\): the collision time of the front obstacle is not corrected, and the collision time of the rear obstacle is delayed by \(w\), and the delay amount is inversely proportional to the distance of the obstacle to the rear path of the vehicle.

[0009] As a further improvement of this technical solution, in S4, the calculation method of the collision time \(t\) is: within the prediction time, if the distance between the obstacle and the vehicle at a certain moment is less than or equal to \(d_s\), then \(t\) is that moment; if the distance is always greater than \(d_s\) within the prediction time, then \(t=\infty\).

[0010] As a further improvement of this technical solution, in S6, the preset collision time threshold \(T\) is set according to the braking performance and safety requirements of the vehicle.

[0011] As a further improvement of this technical solution, in S7, the reduction step of the reference speed is a preset value, and the preset value is set according to the vehicle dynamics performance and scene requirements.

[0012] As a further improvement of this technical solution, in S8, if there are multiple reference speeds that make tm equal, the maximum reference speed among them is selected as the output.

[0013] As a further improvement of this technical solution, the speed planning method for the low-speed driverless vehicle is applicable to low-speed driverless vehicles in dynamic scenarios, including sweeping vehicles, logistics vehicles, and sightseeing vehicles.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. In the speed planning method for the low-speed driverless vehicle, by starting the search from the maximum reference speed and decreasing it, and taking "maximizing the corrected minimum collision time" as the optimization goal, while ensuring driving safety, the driving efficiency is maximally retained, achieving the balance between the safety and passing efficiency of the low-speed driverless vehicle in the dynamic scenario.

[0015] 2. In the speed planning method for the low-speed driverless vehicle, through the division of the vehicle coordinate system quadrants, the original collision time is directly adopted for the front obstacle to ensure that its risk is preferentially processed; for the rear obstacle, the collision time is delayed by weight to reflect the "rear-end collision is fully responsible" rule, avoiding the vehicle from decelerating excessively due to the rear risk and reducing the rear-end collision risk caused by sudden braking.

[0016] 3. In the speed planning method for the low-speed driverless vehicle, different from the problem that the traditional speed corridor and model prediction method may have no solution due to complex scenarios, this method weakens the influence of the rear obstacle and retains more feasible speed intervals, ensuring that an effective speed planning result can still be output in the multi-obstacle scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is the overall flow block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] Currently, in the dynamic operation scenarios of low-speed driverless vehicles (such as cleaning vehicles, logistics vehicles, and sightseeing vehicles), speed planning is one of the core technologies to ensure safety and efficiency. In the existing technologies, common speed planning methods such as the time-to-collision model, the speed corridor method, and the model predictive method do not effectively distinguish the risk levels of obstacles in the front and rear, and the engineering mapping of traffic rules such as full liability for rear-end collisions is insufficient. As a result, low-speed driverless vehicles may reduce safety and traffic efficiency in actual applications due to excessive deceleration (affected by rear obstacles) or untimely handling of collision risks (insufficient priority for front obstacles). In view of this, as Figure 1 shown, the purpose of the present invention is to provide a speed planning method for low-speed driverless vehicles, including the following steps: S1. Obtain the maximum reference speed, the expected trajectory of the vehicle itself, and the positions and speeds of all obstacles in the scenario; the maximum reference speed is output by trajectory planning, comprehensively considering road speed limits, traffic efficiency, and road curvature; S2. Calculate the attribute w for each obstacle; S3. Assume that all obstacles move in a uniform straight line, and predict the positions of each obstacle within a future period of time; S4. Set the vehicle speed as the current reference speed, calculate the time-to-collision t between each obstacle and the vehicle itself. If there is no collision within the prediction time, set t = ∞; S5. Calculate the corrected time-to-collision t0 = t + w; S6. Select the minimum value tm among all t0. If tm ≥ T, where T is a preset time-to-collision threshold, then proceed to S8; otherwise, proceed to S7; S7. Decrease the current reference speed. If the reference speed is less than 0, proceed to S8; otherwise, return to S4; S8. Output the reference speed that maximizes tm as the planned speed.

[0020] By starting from the maximum reference speed and performing a decreasing search, with the optimization goal of "maximizing the corrected minimum time-to-collision", while ensuring driving safety, the driving efficiency is maximally retained, achieving the balance between the safety and traffic efficiency of low-speed driverless vehicles in dynamic scenarios.

[0021] Considering that low-speed driverless vehicles need to take into account regulatory restrictions (such as road speed limits), task efficiency (such as the transportation timeliness of logistics vehicles), and vehicle dynamics characteristics (such as the need to reduce speed due to centrifugal force limitations on curves) during actual operation, if a fixed speed is directly adopted or speed planning only relies on local obstacle information, it is likely to result in unreasonable speeds (such as speeding, skidding on curves, or inefficient driving). Therefore, in step S1, a technology that decouples speed planning from trajectory planning is adopted. The trajectory planning module outputs the maximum reference speed by integrating global environmental information such as road speed limits, traffic efficiency, and road curvature. This technology enables the initial value of speed planning to be pre-adapted to the basic constraints of the road and task objectives, providing a reasonable benchmark starting point for subsequent dynamic speed adjustment based on obstacles. Its beneficial effects are as follows: First, compliance and safety. By incorporating road speed limits and curvature constraints, it ensures that the driving speed of the vehicle without obstacles does not exceed the physical limit or regulatory restrictions, avoiding the risk of loss of control caused by excessive speed. Second, task efficiency. The reference speed is set in combination with the traffic efficiency requirements (such as time optimization of the coverage path of a sweeper), avoiding the reduction of operation efficiency due to overly conservative fixed speeds. Third, modular flexibility. The separation of speed planning and trajectory planning allows for independent adaptation to different types of low-speed vehicles (such as logistics vehicles with different load capacities, sightseeing vehicles with different steering characteristics). Only by adjusting the trajectory planning algorithm can the speed planning logic be reused, reducing the system development cost.

[0022] In a dynamic traffic scenario, the risk of frontal collision may be caused by obstacles ahead (such as pedestrians, slow vehicles in the same direction), while the risk of rear-end collision of obstacles behind (such as vehicles following too closely) is usually borne by the following vehicle in traffic rules. If the collision time of the two types of obstacles is treated equally, it is likely to cause the vehicle to decelerate excessively due to the risk behind (reducing efficiency) or ignore the emergency risk ahead (increasing the collision probability). Therefore, in step S2, an obstacle risk grading technology based on the vehicle coordinate system is adopted, and differential treatment is achieved through the following core means: First, the spatial quadrant division technology is adopted. A coordinate system is established with the vehicle center as the origin. The first and fourth quadrants are defined as "front" (with the highest risk priority), and the second and third quadrants are defined as "rear" (with a lower risk priority), directly mapping the physical characteristic that "the risk of frontal collision is higher than the risk of rear-end collision" in the traffic scenario; Secondly, the weight dynamic calculation technology is adopted. For the obstacle in front, w is directly set to 0, so that its collision time participates in the decision-making without delay, ensuring that the vehicle responds to the frontal threat first. For the obstacle behind, the weight w is calculated through "foot perpendicular ray determination + dynamic safety distance formula (ds = ks × v + ds0, where ks is a positive coefficient, v is the vehicle speed, and ds0 is the lower limit of the safety distance and is greater than the radius of the vehicle's circumscribed circle)". Only when the trajectory of the obstacle may cut into the safe area behind the vehicle (the foot perpendicular is on the ray and d ≤ ds), the collision time is corrected with a delay amount inversely proportional to the distance (w = (ds - d) × k, where k is the distance weight), reflecting the rule that "the following vehicle should maintain a safe distance by itself". Through the design of "zero weight in the front and delayed correction in the rear", the legal principle of "priority in the front and full liability for rear-end collision" is transformed into a computable algorithm logic, avoiding unreasonable deceleration of the vehicle for the obstacle behind, conforming to the actual traffic liability division, and enabling the engineering implementation of traffic rules. At the same time, the obstacle behind is not "completely ignored" but "weakened as needed". Only when there is a real possibility of rear-end collision (the trajectories intersect and the distance is too close) is a limited impact exerted, which not only avoids the loss of feasible solutions caused by the "one-size-fits-all" method of traditional methods but also prevents the omission of extreme risks behind, achieving refined risk classification. Moreover, when facing the obstacle in front, the vehicle decelerates in time (because the collision time has no delay due to w = 0), ensuring that the safety threshold is quickly reached. When facing the obstacle behind, the vehicle gives priority to maintaining a high reference speed (only decelerates when tm does not meet the standard), avoiding sudden braking due to excessive avoidance of risks behind, which instead increases the probability of rear-end collision by the following vehicle, achieving a balance between safety and efficiency. Furthermore, by calculating the foot perpendicular and ray through vector velocity (instead of only the longitudinal distance), the risk area of obstacles in complex motion patterns such as diagonal insertion and approaching from the side rear is accurately identified, solving the problem of the ineffectiveness of traditional methods in dealing with laterally moving obstacles and improving the robustness in complex scenarios.

[0023] Considering the operation scenarios of low-speed driverless vehicles (such as cleaning vehicles, logistics vehicles), the moving speeds of obstacles (such as pedestrians, other vehicles moving at low speeds) are relatively low and the trajectories are relatively stable, with extremely few cases of drastic movements such as sudden acceleration and sharp turning. At the same time, limited by the computing power resources of the in-vehicle computing platform, if a complex motion prediction model (such as considering acceleration and curvature changes) is adopted, it will lead to an increase in calculation time and it is difficult to meet the real-time requirement. Therefore, the uniform linear motion assumption technology is adopted in step S3, that is, it is assumed that all obstacles maintain the current speed and direction unchanged within the prediction time period, and the future position is directly extrapolated based on the initial position and speed. By adopting the motion model simplification technology, the acceleration and turning behavior of the obstacle are ignored, and linear extrapolation is performed only through the initial velocity vector (magnitude and direction of velocity), converting the complex non-linear motion prediction into a simple coordinate transformation calculation; by adopting the time discretization technology, the position of the obstacle within a future period of time (2s in this embodiment) is discretized at a fixed time interval (0.01s in this embodiment), forming a position sequence at discrete time points, which is convenient for collision detection with the position of the host vehicle. Through the above technologies, the calculation efficiency is greatly improved. There is no need to solve complex dynamic equations or optimization problems. The position of the obstacle can be predicted only through basic vector operations (velocity × time + initial position), and the computational complexity is linear, meeting the real-time requirements of the embedded controller (such as a control frequency above 100Hz); at the same time, in the target application scenarios, the motion patterns of obstacles (such as pedestrians in the park, warehouse logistics robots) conform to the "uniform linear motion" assumption, and the prediction error is within the safety threshold (such as the speed change does not exceed 0.5m / s within 2 seconds), which can not only meet the accuracy requirements of collision detection, but also avoid the problem of error amplification caused by over-modeling; for occasional non-uniform motion (such as pedestrians suddenly stopping), the subsequent calculation of the time to collision (step S4) will be updated in real time, compensating for the limitations of the prediction model by dynamically adjusting the speed, thereby improving the adaptability to low-speed scenarios; in summary, through "reasonable simplification of the motion model + discrete time prediction", under the specific constraints of low-speed dynamic scenarios, the best balance point between prediction accuracy and calculation efficiency is found, enabling the speed planning algorithm to not only meet the real-time requirements, but also achieve reliable obstacle motion prediction at a low cost in engineering practice, providing a solid foundation for subsequent calculation of the time to collision and speed adjustment.

[0024] Considering that in dynamic scenarios, low-speed autonomous vehicles need to judge the collision risk with obstacles in real time and accurately, and traditional methods for calculating the time to collision (such as complex integration based on relative acceleration, probability density estimation) may lead to decision-making lag or misjudgment due to computational time consumption or model uncertainty, especially difficult to meet the real-time requirements on embedded platforms with limited computing power. Therefore, in step S4, a safety distance threshold comparison technology based on discrete time points is adopted to quickly determine the time to collision through the following core means: Firstly, the discrete time sampling technology is adopted to discretize the prediction time period (such as the future 2s) at a fixed time interval (such as 0.01s), generating an equally spaced time series t1, t2,..., tn, and calculating the distance between the host vehicle and the obstacle at each moment in turn. Secondly, the safety distance threshold determination technology is adopted. The dynamic safety distance ds = ks × v + ds0 (related to the speed of the host vehicle) is predefined. If the distance between the obstacle and the host vehicle at a certain moment ≤ ds, then that moment is determined as the time to collision t; if the distance at all moments is > ds, then the time to collision is set to infinity (t = ∞). Collision determination depends on the dynamic safety distance ds (which increases with the vehicle speed), rather than a fixed threshold, in line with the physical law that "the higher the vehicle speed, the greater the required safety buffer distance". For example, when the vehicle speed is 3 m / s, ds = 0.1×3 + 2 = 2.3 m, ensuring more safety margins during high-speed driving; ds0 (the lower limit of the safety distance) is associated with the vehicle size (it needs to be greater than the circumradius), avoiding the "blind zone" problem where collision detection is not triggered when the obstacle is close to the vehicle body. For example, when a sweeper is driving on a narrow path, ds0 can ensure that the safety distance around the vehicle body is effectively monitored, thus achieving accurate mapping of dynamic safety constraints; moreover, through the lightweight design of "discrete sampling + threshold comparison", while ensuring the accuracy of collision risk judgment, the computational complexity is significantly reduced, enabling it to run in real time on a low-cost hardware platform. Combined with the collaborative design of the dynamic safety distance, it not only meets the safety constraints in low-speed scenarios but also provides clear quantitative inputs for subsequent risk correction based on traffic rules (Steps S5 - S8), forming an efficient closed-loop of "detection - evaluation - decision".

[0025] Considering that in the scenario of dynamic obstacles, low-speed autonomous vehicles need to maximize driving efficiency on the premise of ensuring safety (avoiding collisions), and directly adopting a fixed speed or a random search speed may lead to non-compliance with safety thresholds or low efficiency, and traditional optimization algorithms (such as MPC) are difficult to converge quickly due to complex parameters. Therefore, in Steps S5 - S8, an iterative speed search technique based on safety thresholds is adopted, and speed planning is achieved through a closed-loop mechanism of "modified collision time screening + decreasing speed optimization". The specific technology and effects are as follows: Firstly, a decision-making technique driven by safety thresholds is adopted. By presetting the collision time threshold T (set based on the vehicle braking performance, such as the time corresponding to the emergency braking distance), the "safety or not" is transformed into a clear mathematical judgment condition (tm ≥ T). Only when the minimum value of the modified collision time of all obstacles meets the safety threshold, the current speed is accepted; Secondly, a decreasing speed search strategy is adopted: starting from the maximum reference speed, the speed is gradually reduced (the step size is adjustable) until the highest feasible speed that satisfies tm ≥ T is found, forming a decision-making logic of "efficiency first, safety guaranteed"; Finally, a multi-objective balance optimization technique is adopted. During the speed adjustment process, the optimization objective is to "maximize the modified minimum collision time tm" to ensure the maximum safety margin is retained after deceleration; when multiple speeds correspond to the same tm, the maximum speed among them is selected (S8), reflecting the principle of "giving priority to efficiency under the same safety level" and avoiding unnecessary speed reduction; By directly correlating the threshold T with the vehicle's braking performance (e.g., the greater the braking acceleration, the smaller T can be), it is ensured that the planned speed always meets the physical constraint of "being able to stop safely within T time", avoiding the risk of insufficient braking distance due to excessive speed. For example, when the braking distance is 3m and the speed is 3m / s, T is set to 2s (reserving a 0.5s reaction time) to ensure safe parking in case of an emergency; the most urgent collision risk is screened through tm = min(t0), and the obstacle with the greatest threat to safety (such as a pedestrian in the front at a short distance) is given priority, avoiding interference from secondary risks (such as a vehicle following at a long distance behind) to the core decision-making; thus, safety is strongly bound to the braking performance; at the same time, by starting the search from the maximum reference speed, the highest speed is directly output when there are no obstacles, ensuring the operation efficiency of road sweepers, logistics vehicles, etc. When there are obstacles, deceleration is only performed when the current speed does not meet the safety threshold (tm < T), and the collision time is recalculated after each deceleration, avoiding the excessive deceleration of "going straight to the end" in traditional methods (such as directly reducing from 3m / s to 1m / s), achieving a smooth speed adjustment of "being as fast as possible when possible and being slow when necessary", reducing energy loss and mechanical shock; and, different from the problem that the traditional speed corridor method may result in an empty feasible interval due to multiple obstacles, this technology retains more feasible speeds through the delay of the collision time of the rear obstacle (w mechanism), combined with a decreasing search strategy, ensuring that a feasible speed not lower than 0 can be found even in complex scenarios (such as triggering parking when the final output is 0), avoiding algorithm failure; the discrete speed search (such as a step size of 1m / s) reduces the computational complexity. Compared with continuous optimization problems, there is no need to handle gradient solving and boundary conditions, and the engineering implementation difficulty is greatly reduced, making it suitable for real-time operation of embedded controllers; in summary, through the design of "safety threshold drive + decreasing speed optimization", a decision-making framework of "efficiency first, safety guarantee" is constructed at the algorithm level, which not only meets the strict requirements of low-speed driverless vehicles for real-time performance and reliability, but also realizes a smooth transition from the theoretical model to engineering implementation through parameterized thresholds (T) and modular search strategies. This speed planning mechanism of "taking safety as the bottom line and efficiency as the guide" provides key technical support for the large-scale application of unmanned road sweepers, logistics vehicles, etc. in open parks, warehousing scenarios.

[0026] In summary, take the following scenario as an example: There is a pedestrian crossing at 1m / s 5m ahead (belonging to the front area), and a vehicle is approaching at 4m / s 10m behind (belonging to the rear area); First, perform the calculation in step S2: For the pedestrian, w = 0. The foot of the perpendicular of the vehicle behind is on the ray, and d = 8m > ds = 0.1×3 + 2 = 2.3m, so w = 0 (because ds - d < 0).

[0027] Second, perform the collision time calculation in step S4: The collision time of the pedestrian t = 5 / (3 - 1) = 2.5s (relative speed 2m / s), and there is no collision with the vehicle behind (t = ∞).

[0028] Finally, perform the judgments in steps S6 to S8: tm = 2.5 s ≥ T = 2 s, directly output v = 3 m / s, and the vehicle passes through with the highest efficiency while ensuring the safety time with the pedestrian.

[0029] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only the preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A speed planning method for a low-speed driverless vehicle, characterized in that: It includes the following steps: S1. Obtain the maximum reference speed, the desired trajectory of the host vehicle, and the positions and speeds of all obstacles in the scenario; The maximum reference speed is output by trajectory planning, comprehensively considering road speed limits, traffic efficiency, and road curvature; S2. Calculate the attribute w for each obstacle; S3. Assume that all obstacles move in a uniform straight line, and predict the positions of each obstacle in the future for a period of time; S4. Set the speed of the host vehicle as the current reference speed, calculate the collision time t between each obstacle and the host vehicle. If there is no collision within the prediction time, set t = ∞; S5. Calculate the corrected collision time t0 = t + w; S6. Select the minimum value tm among all t0. If tm ≥ T, where T is a preset collision time threshold, go to S8; otherwise, go to S7; S7. Decrease the current reference speed. If the reference speed is less than 0, go to S8; Otherwise, return to S4; S8. Output the reference speed that maximizes tm as the planned speed.

2. The speed planning method for a low-speed driverless vehicle according to claim 1, wherein In S2, the specific steps for calculating the attribute w are as follows: S2.1: Establish a coordinate system with the center of the vehicle as the origin, the forward direction as the positive x-axis direction, and the positive y-axis direction to the left. Define the first and fourth quadrants as the front, and the second and third quadrants as the rear; S2.2: If the obstacle is in front of the host vehicle, set w = 0; S2.3: If the obstacle is behind the host vehicle, calculate the foot of the perpendicular from the position of the host vehicle to the ray determined by the position and speed of the obstacle. If the foot of the perpendicular is not on the ray, set w = 0; S2.4: If the foot of the perpendicular is on the ray, calculate the distance d between the position of the host vehicle and the foot of the perpendicular. Calculate the safety distance ds = ks × v + ds0, where ks is a positive coefficient, v is the speed of the host vehicle, and ds0 is the lower limit of the safety distance and is greater than the radius of the circumcircle of the host vehicle; S2.5: If ds - d ≥ 0, set w = (ds - d) × k, where k is the distance weight; otherwise, set w = 0.

3. The speed planning method for a low-speed driverless vehicle according to claim 2, wherein In S2, the risks of front and rear obstacles are distinguished by the obstacle attribute w: the collision time of front obstacles is not corrected, and the collision time of rear obstacles is delayed by w, and the delay amount is inversely proportional to the distance of the obstacle to the rear path of the host vehicle.

4. The speed planning method for a low-speed driverless vehicle according to claim 2, characterized in that, In S4, the calculation method of the collision time t is: within the prediction time, if the distance between the obstacle and the host vehicle at a certain moment is less than or equal to ds, then t is that moment; if the distance is always greater than ds within the prediction time, then t = ∞.

5. The speed planning method for a low-speed driverless vehicle according to claim 1, characterized in that: In S6, the preset collision time threshold T is set according to the braking performance and safety requirements of the vehicle.

6. The speed planning method for a low-speed driverless vehicle according to claim 1, characterized in that: In S7, the step size for decreasing the reference speed is a preset value, and the preset value is set according to the vehicle dynamics performance and scenario requirements.

7. The speed planning method for a low-speed driverless vehicle according to claim 1, characterized in that: In S8, if there are multiple reference speeds that make tm equal, select the maximum reference speed among them as the output.

8. The speed planning method for a low-speed driverless vehicle according to claim 1, characterized in that: The speed planning method for the low-speed driverless vehicle is applicable to low-speed driverless vehicles in dynamic scenarios, including cleaning vehicles, logistics vehicles, and sightseeing vehicles.

Images