Next Article in Journal
RS-Xception: A Lightweight Network for Facial Expression Recognition
Previous Article in Journal
S-ZAC: Hardening Access Control of Service Mesh Using Intel SGX for Zero Trust in Cloud
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Shared-Road-Rights Driving Strategy Based on Resolution Guidance for Right-of-Way Conflicts

1
School of Mechanical and Electrical Engineering, Hainan University, Haikou 570228, China
2
Suzhou Automotive Research Institute, Tsinghua University, Suzhou 215000, China
3
Unit 61578 of the Chinese People’s Liberation Army, Shiyan 442000, China
*
Author to whom correspondence should be addressed.
Electronics 2024, 13(16), 3214; https://doi.org/10.3390/electronics13163214
Submission received: 27 June 2024 / Revised: 4 August 2024 / Accepted: 8 August 2024 / Published: 14 August 2024

Abstract

:
In addressing the critical issue of right-of-way conflicts in mixed-traffic environments, this paper introduces a novel shared right-of-way driving strategy that encompasses two guiding frameworks for resolution. The first framework applies to active lane changing. Before lane changing occurs, this framework allocates the right of way for autonomous vehicles (AVs). Based on the allocated right of way, the AVs decide whether to send a request for a shared right of way to relevant vehicles. To enhance lane-changing comfort, the vehicle assesses whether the variance of roll and lateral acceleration exceeds a preset threshold, ultimately deciding whether to proceed with the lane change. The second framework pertains to passive lane changing. After detecting an obstacle, this framework allocates the right of way. The AVs calculate acceleration based on their speed and distance from the obstacle, using this information to determine whether to change lanes or decelerate in order to avoid the obstacle. If lane changing is chosen, further evaluation is necessary. Based on the allocated right of way, the AVs decide whether to request a shared right of way from relevant vehicles. To improve lane-changing comfort, the AVs compare the variance of roll and lateral acceleration with that of pitch and longitudinal acceleration, and then they decide whether to proceed with the lane change. The proposed strategy has been validated in various scenarios, including high-speed (105 km/h), low speed (13 km/h), and general scenarios with AVs and obstacles at a distance of 125 m. The results show that the strategy effectively functions in both high-speed and low-speed scenarios.

1. Introduction

The problem of the harmonious and safe coexistence between AVs and other vehicles in complex multi-lane traffic environments has been the driving strategy’s focus. The main types of autonomous driving strategies include defensive driving, competitive driving, and cooperative driving; all aim at minimizing the potential risks associated with the complexity of the traffic environment [1,2,3].
Defensive driving strategies aim to minimize the risk of traffic accidents and enhance safety by making independent decisions. A typical example of defensive driving strategies is the responsibility-sensitive safety strategy (RSS strategy) [4]. The RSS strategy is primarily devised to guarantee that AVs do not merely react to their surroundings but also formulate decisions that mirror the safe and prudent conduct of experienced human drivers. This innovative approach endeavors to significantly diminish accident risks by closely emulating the driving habits and judgment of skilled human operators. To fulfill this objective, the RSS strategy integrates the real-time monitoring of crucial parameters, including the vehicle’s relative speed and distance vis-à-vis its immediate environment. Utilizing these dynamically refreshed metrics, the strategy offers precise acceleration, deceleration, and steering recommendations tailored to uphold safe distances and comply with speed limits. By adhering to formalized rules that steer these decisions, the RSS strategy ensures that the decision-making process is both interpretable and readily comprehensible, fostering greater trust and assurance in the roadworthiness of AVs [5,6].
The RSS strategy excels in safety control. However, the decisions made by AVs to avoid low-probability accidents often result in suboptimal road traffic efficiency. To address the efficiency issue, the concept of learning has been applied to AVs [7,8,9,10,11,12]. A typical example is the competitive driving strategy, which can complete driving tasks quickly, thus reducing travel time. However, the competitive driving strategy is highly reliant on training data obtained from real-time environments, and this approach confronts two major challenges. Firstly, there is the unpredictability of actual driving environments. When the model struggles to fully adapt to constantly evolving new scenarios, it is prone to the issue of “overfitting”, where it performs excessively well using known data but poorly on unknown or novel situations. This phenomenon has been extensively documented in prior research [13,14,15]. Secondly, there is the issue of the representativeness of training data. If the training data contain significantly more samples of rational driving behaviors than irrational or abnormal ones, AVs may overlearn and tend to adopt frequently occurring, sometimes even aggressive, driving strategies. While such strategies may temporarily enhance driving speed and efficiency, due to the lack of interpretability in the model’s decisions, they often compromise passenger comfort in pursuit of ultimate efficiency. As a result, this delicate balance between efficiency and comfort inadvertently increases the risk of accidents involving AVs.
The emergence of cooperative driving strategies aims to dynamically allocate weights to safety, efficiency, and comfort based on different driving scenarios. The cooperative driving strategy, achieved through intelligent route planning and coordinated vehicle operation, aims to mitigate congestion and road construction issues, thereby enhancing traffic efficiency [16,17,18,19]. Through cooperative vehicle control and energy management, it also aims to achieve efficient energy utilization. Under such a comprehensive driving strategy, which prioritizes overall driving safety and efficiency, AVs no longer need to engage in risky driving behaviors to fulfill tasks, significantly reducing the occurrence of accidents. However, cooperative driving technology is still in its developmental stage, and it faces several technical limitations, such as communication delays and sensor errors, which may lower driving safety and reliability. Cooperative AVs require coordination among multiple vehicles and information sharing, leading to increased system complexity, development difficulty, and reliance on substantial data support, including inter-vehicle communication data and sensor data. This necessitates frequent data transmission and processing among vehicles, adding to system burden and latency. Moreover, real-world traffic environments lack comprehensive intelligent infrastructure. Therefore, the full realization of cooperative driving strategies requires long-term development and construction. In the future, people will undoubtedly encounter challenges such as mixed-traffic scenarios with AVs and traditional vehicles [20,21,22,23].
In response to the shortcomings of defensive driving, competitive driving, and cooperative driving strategies, scholars have innovatively proposed a negotiated driving strategy in their research [24]. The negotiated driving strategy achieves harmonious coexistence between AVs and traditional vehicles by clearly defining the concept of road rights and its mathematical expression. During the negotiation process, AVs utilize signals recognizable for traditional vehicles, such as turn signals and sounds, to effectively negotiate and allocate road use rights. The negotiated driving strategy significantly enhances traffic flow capacity and facilitates smooth interaction between autonomous and traditional vehicles. However, it is noteworthy that, in extreme traffic conditions, AVs may still encounter collisions, and the current strategy has yet to fully address potential issues such as passive lane changes and passenger comfort.
In summary, researchers from various fields have completed numerous explorations and practical works regarding driving strategies for AVs, encompassing defensive driving, competitive driving, cooperative driving, and negotiated driving. Nevertheless, these strategies are still confronted with notable deficiencies: the defensive strategy is overly conservative, compromising traffic efficiency; the competitive strategy relies heavily on data training, resulting in poor scenario adaptability, intensifying competition among vehicles, and increasing accident risks; the cooperative strategy is highly dependent on sophisticated technologies and communication, facing challenges such as network delays and infrastructure construction issues; the negotiated strategy, meanwhile, overlooks crucial scenarios like passive lane changes in strategy formulation, deviating from the original intention of enhancing driving experience and optimizing passenger comfort through technological advancements. Consequently, when designing driving strategies, given that the variation in vehicle acceleration is directly linked to passenger comfort, it is imperative to flexibly control vehicle acceleration based on specific scenario characteristics, ensuring that, while guaranteeing driving safety, passenger comfort is also maintained within an acceptable range [25,26,27,28,29]. The research findings of McConnell’s curve [30] have shed light on passengers’ sensitivity and thresholds to changes in acceleration with different degrees of freedom, providing a crucial theoretical basis for further optimizing driving strategies.
In view of the above problems, this paper proposes a shared-road-rights driving strategy based on resolution guidance for right-of-way conflicts that balances traffic efficiency and comfort on the basis of safety. The main contribution of this paper is reflected in three aspects. Firstly, the dual framework scheme of the strategy intervenes in the division of road rights through an easily understood mathematical model, making the decision-making process of AVs more transparent and predictable. Secondly, in the strategy, the variance of roll acceleration and lateral acceleration, as well as the variance in pitch acceleration and longitudinal acceleration, are introduced as indicators for evaluating comfort. This design makes the ride comfort of AVs adjustable. Finally, a passive lane-changing framework is constructed for obstacle-avoidance scenarios. The passive lane-changing framework consists of simple and easily understood mathematical formulas, ensuring that rapid and reasonable decisions can be made in emergency situations. This mechanism can improve the adaptability and safety of autonomous driving systems, providing more solid protection for passengers and other road users. This paper represents a further advancement in driving strategies, offering a novel path towards the harmonious coexistence of AVs and traditional vehicles. However, to ensure the smooth conduct of validation, the recognition process of traditional vehicles’ motion states and intentions using AVs has been simplified. The focus of future research in this area should lie in two main directions: firstly, enhancing the accuracy of AVs in recognizing the motion states and intentions of traditional vehicles, as well as optimizing the precision of the strategy itself; secondly, categorizing driving styles based on comfort and safety thresholds to accommodate the diverse preferences and scenario requirements of different passengers.
This article is structured as follows: Section 2 describes road rights in traffic scenarios. Section 3 provides a detailed design of a driving strategy that considers comfortably shared road rights. Section 4 validates the proposed driving strategy through simulation and actual vehicle testing. Finally, Section 5 presents the discussion and conclusion.

2. The Description of Road Rights in Traffic Scenarios

The description [24] of road rights in vehicle-following scenarios is shown in Figure 1. V 1 legally owns a certain distance behind it, and V 2 should maintain an appropriate distance to avoid rear-end collisions with the preceding vehicle. J ( m , n ) represents the length of the exclusion zone between m and n , and the road rights belong to m . J ( V 1 , V 2 ) is the legal exclusion zone for V 1 . The length of the exclusion zone is when V 1 decelerates at the maximum acceleration, and V 2 can safely stop according to the distance maintained. The length of the exclusion zone is shown in Equation (1).
J ( V 1 , V 2 ) = v 2 θ + v 2 2 2 a V 2 , b r a k e v 1 2 2 a max , b r a k e a V 2 , b r a k e = a min , b r a k e + v 2 v max a max , b r a k e a min , b r a k e
where the following applies: θ represents the reaction time of AVs; v 1 represents the speed of the preceding vehicle; v 2 represents the speed of the following vehicle; v max represents the maximum speed limit of the line. The rule of road rights is that, within the scope of the road rights, if there is a sole owner, at the same time and without the owner’s consent, vehicles without the road rights are prohibited from entering the zone. When an accident occurs in this zone, the zone owner should not be held responsible or take on minor responsibilities. The scope of road rights can be further divided into the exclusion zone and the shared zone, and the zone owner has the right to share the shared zone with the applicant. For zones without owners, the road rights adhere to the following rule: first come, first served.
The description of road rights in mixed-flow scenarios is shown in Figure 2. V 1 is an AVs, and V 3 is a shared car. V 1 needs to divide the surrounding road rights into four traditional vehicles, V 2 , V 3 , V 4 , and V 5 , before changing lanes. X ( m , n ) represents the length of the shared zone between m and n , and the road rights belongs to m . J ( V 5 , V 1 ) is the legal exclusion zone for V 5 , J ( V 4 , V 1 ) is the legal exclusion zone for V 4 , J ( V 3 , V 1 ) is the legal exclusion zone for V 3 , and X ( V 3 , V 1 ) is the legal shared zone for V 3 . V 3 can decide whether to share X ( V 3 , V 1 ) with V 1 , according to preference.

3. Design of Driving Strategy

3.1. Two Frameworks

The driving strategy considering comfort-based, shared road rights consists of two frameworks. Figure 3 shows Framework 1: active lane changing for shared road rights. And Figure 4 shows Framework 2: passive lane changing for shared road rights when encountering obstacles. Each framework is divided into three modules with the following assumptions:
  • The frameworks are only applicable to lane-changing scenarios in the same direction, and no interference from other traffic participants during AVs driving;
  • AVs enable the perception of other vehicles or obstacles, including speed, acceleration, distance, etc.;
  • In traffic scenarios, only one vehicle can exist in the same place on a lane.
Module 1 of Framework 1 is a stage of longitudinal adjustment for AVs, and it determines whether the remaining lanes can be merged directly. Suppose the conditions for merging the remaining lanes directly are met. In that case, it is still necessary to determine whether Decision Model 1 (Equation (3)) in Module 2 exceeds the pre-set comfort and safety thresholds. If the pre-set comfort and safety threshold is not exceeded, AVs can change lanes; otherwise, importing stops, and AVs go ahead and wait for the next opportunity. If the conditions for merging the remaining lanes directly are not met, it is still necessary to determine, through Module 2, whether the spacing of the safe import into the remaining lanes is exceeded by Decision Model 1 and whether the pre-set comfort and safety thresholds are exceeded. If it is at a safe pitch and does not exceed pre-set comfort and safety thresholds, a sharing request can be made to the shared car. If the shared car agrees to the request, AVs can perform an interchange; otherwise, they stop merging, and the AVs move on and wait for the next opportunity. Framework 2 adds Decision Model 2 (Equation (4)) compared to active lane changing, which determines deceleration or turning to avoid obstacles based on comfort.

3.2. Driving Strategy Applied to Transportation Scenarios

3.2.1. AVs Can Merge Directly without Sending a Sharing Request

The scenario where AVs meet the conditions for direct lane changing and do not need to send a sharing request is shown in Figure 5. e is an AVs, and V 3 is a shared car. V 1 is not in the exclusion zone of the two vehicles in front of it. At the same time, the distance between V 1 and V 3 is greater than the sum of the lengths of V 3 ’s exclusion zone and V 3 ’s shared zone, i.e., V 1 is in the unrestricted zone, which is judged using Equation (2). V 3 does not accelerate or decelerate, and the vehicle in front does not decelerate or accelerate. V 1 does not need to send a sharing request to V 3 .
C ( V 1 , V 5 ) J ( V 5 , V 1 ) C ( V 1 , V 4 ) J ( V 4 , V 1 ) C ( V 1 , V 3 ) J ( V 3 , V 1 ) + X ( V 3 , V 1 ) a 3 0 ,   a 4 0 ,   a 5 0
where C ( m , n ) represents the longitudinal distance between m and n , while a 3 , a 4 , and a 5 represent the longitudinal acceleration of the surrounding vehicles.
The working process of Framework 1 is as follows.
  • A comfort and safety threshold is previously based on the comfort level;
  • AVs use the vehicle dynamics model to input the current vehicle speed and distance from the target position to obtain the roll acceleration and lateral acceleration generated during lane changing. It is worth noting that, at this moment, the AVs have not changed lanes;
  • AVs use Decision Model 1 to obtain the variance of lane changing by taking roll acceleration and lateral acceleration as inputs. If the variance does not exceed the pre-set comfort and safety threshold, the AVs can change lanes now.
w 1 a x + w 2 a p i t c h = a 1 w 3 a r o l l + w 4 a y = a 2 V a r ( a ) = min V a r ( a 1 ) , V a r ( a 2 )
where w represents the weight, w 1 + w 2 = 1 , w 3 + w 4 = 1 . In this paper, the following applies: w 1 = w 2 = w 3 = w 4 = 0.5 ; a p i t c h represents the pitch acceleration; a r o l l represents the roll acceleration; a x and a y represent the longitudinal acceleration and lateral acceleration, respectively; a 1 and a 2 represent the accelerations of two sets after data fusion, respectively; and V a r ( a ) represents the minimum variance, representing comfort. V a r ( a 1 ) represents the variance calculated from the set of data a 1 , and the same principle applies to V a r ( a 2 ) . V a r ( a ) = min V a r ( a 1 ) , V a r ( a 2 ) , which means the following: if V a r ( a 1 ) is less than V a r ( a 2 ) , then V a r ( a ) equals V a r ( a 1 ) . In this case, the AVs decide on braking, as braking is more comfortable than lane changing. Similarly, if V a r ( a 2 ) is less than V a r ( a 1 ) , then V a r ( a ) equals V a r ( a 2 ) . At this point, the AVs decide on lane changing because lane changing is more comfortable than braking. It is worth noting that a x , a p i t c h , a r o l l , and a y have different units and significant numerical differences, so it is necessary to map them into the same range before performing numerical weighting and fusion.
The working process of Framework 2 is as follows.
  • After identifying obstacles, AVs determine deceleration or steering through Decision Model 2. If AVs want to change lanes to the left and avoid obstacles at this moment, but cannot do so yet, the AVs need to be further evaluated through Decision Model 1;
  • AVs use the vehicle dynamics model to input the current vehicle speed and distance from the obstacles to obtain the roll acceleration and lateral acceleration generated during lane changing. AVs obtain the pitch acceleration and longitudinal acceleration generated during braking through Equation (5). It is worth noting that, at this moment, the AVs have not changed lanes or braked;
  • AVs obtain the variance of braking and the variance of left lane changing through Decision Model 1, respectively. If the variance of left lane changing to the left is smaller than the variance of braking at this moment, the AVs will perform lane changing to the left to avoid obstacles.
f d d e t , v e g o = v e g o 2 2 d d e t v m i n < v e g o   v m a x v m i n 2 v e g o 2 2 d d e t 2 + g 2 v m i n < v e g o   v m a x v m a x 2 v e g o 2 2 d d e t 2 + g 2 v m i n < v e g o   v m a x
where the following applies: v e g o represents the speed of AVs; d d e t represents the distance between the vehicle and the obstacle; g represents the road width; v m i n represents the minimum lane speed limit; and v m a x represents the maximum speed limit on the lane. f d d e t , v e g o represents acceleration, which represents comfort and couples traffic efficiency and safety. For safety reasons, the priority rule is that, when the decision is made to have the highest priority for left turns, the braking priority should be second. Similarly, when the decision is made to prioritize a right turn, the braking priority should be second.
S = ( d d d e t ) ( d f b d d e t ) N = N max × S S min S max S min × b a + a
where the following applies: d represents the current distance from the obstacle; d d e t represents the maximum detection range of the LiDAR sensor; and d f b represents a fixed value, which is a positive integer. When the distance between AVs and the obstacle equals this fixed value, the braking force of the AVs reaches its maximum; the range of S is as follows: 0 S 1 , and N max represents the maximum braking force of the AVs. It is an inherent attribute of the AVs; N represents the current braking force, a represents the minimum interval of the mapping, and b represents the maximum interval of the mapping.

3.2.2. AVs Need to Request Sharing of Merging

The scenario where AVs meet the lane-changing conditions but need to send a sharing request is shown in Figure 6. V 1 is an AVs, and V 3 is a shared car. V 1 is not within the exclusion zone of the two vehicles in front of it. The distance between V 1 and V 3 is less than or equal to the total length of V 3 ’s exclusion zone and V 3 ’s shared zone, and greater than or equal to the length of V 3 ’s exclusion zone, judged according to Equation (6). V 3 does not accelerate or decelerate, and the front car does not decelerate or accelerate. When making the final decision to change lanes, V 1 needs to send a sharing request to V 3 .
The working process of Framework 1 is as follows.
  • A comfort and safety threshold is preset based on the comfort level;
  • AVs use the vehicle dynamics model to input the current vehicle speed and distance from the target position to obtain the roll acceleration and lateral acceleration generated during lane changing. It is worth noting that, at this moment, the AVs have not changed lanes;
  • AVs use Decision Model 1 to obtain the variance of lane changing by taking roll acceleration and lateral acceleration as inputs. If the variance does not exceed the pre-set comfort and safety threshold, the AVs can send a sharing request to V 3 ;
  • When V 3 agrees to the request, the AVs begin to perform lane changing. Otherwise, they stop changing lanes.
C ( V 1 , V 5 ) J ( V 5 , V 1 ) C ( V 1 , V 4 ) J ( V 4 , V 1 ) J ( V 3 , V 1 ) C ( V 1 , V 3 ) J ( V 3 , V 1 ) + X ( V 3 , V 1 ) a 3 0 ,   a 4 0 ,   a 5 0
J ( V 3 , V 1 ) + X ( V 3 , V 1 ) = v 3 θ h u m a n + a max , a c c e l θ h u m a n 2 2 + .. v 3 + a max , a c c e l θ h u m a n 2 2 a min , b r a k e v 1 2 2 a max , b r a k e
where the following applies: C ( m , n ) represents the longitudinal distance between m and n ; θ represents the delay of AVs reaction; θ h u m a n represents the delay of human driver; a max , b r a k e represents the maximum deceleration; and a max , a c c e l represents the maximum acceleration.
The working process of Framework 2 is as follows.
  • After identifying obstacles, AVs determine deceleration or steering through Decision Model 2. If the AVs want to change lanes to the left and avoid obstacles at this moment, but cannot do so yet, the AVs need to be further evaluated through Decision Model 1;
  • AVs use the vehicle dynamics model to input the current vehicle speed and distance from the obstacles to obtain the roll acceleration and lateral acceleration generated during lane changing. The AVs obtain the pitch acceleration and longitudinal acceleration generated during braking through Equation (5). It is worth noting that, at this moment, the AVs have not changed lanes or braked;
  • AVs obtain the variance of braking and the variance of left lane changing through Decision Model 1, respectively. If the variance of left lane changing is smaller than the variance of braking at this moment, the AVs can send a sharing request to V 3 ;
  • When V 3 agrees to the request, the AVs begin to perform lane changing. Otherwise, they stop changing lanes.

4. Results

4.1. Simulation Verification

The strategy in this paper was validated through simulation using SCANeR studio 2021.2, Carsim 2021.0, and Simulink® R2022b, as shown in Figure 7. During the process of simulation verification, the roll acceleration, lateral acceleration, pitch acceleration, and longitudinal acceleration could be obtained with this simulation framework.
The more detailed explanation of Simulink (Part 1 and Part 2) in Figure 7 is shown in Figure 8. Carsim provides vehicle status data (yaw angle, pitch angle, roll angle, etc.). The ADAS mainly includes models such as road right, Decision Model 2, and Decision Model 1. The integrated vehicle dynamics function is ComVhcDyn in Figure 7, generating some driver’s commands. Finally, since the sensor detection of obstacles is required in simulation verification, RTGATEWAY is used as the transmission bridge for sensor data.
The simulation validation of driving strategy in this paper utilizes the road shown in Figure 9 as the testing environment. This road has three lanes in the same direction. The road modeling process is as follows: obtain road section data through OpenStreetMap, import the data into SCANeR studio after processing, modify the road surface properties, and complete road-surface modeling.
The simulation validation of the driving strategy in this paper utilized the range of the lidar shown in Figure 10. The number of lidars is 3. The maximum detection range is 230 m, the number of wire harnesses is 180 pieces, the maximum horizontal and vertical angles are 120°, and the heading settings are −60°, 60°, and 180°, respectively.

4.1.1. AVs Actively Change Lanes

As shown in Figure 11, based on the SCANeR studio scenario, the following road conditions were deployed: The speed of AVs is 105 km/h. Due to driving tasks, it is necessary to make a left turn on the fast lane and increase the speed to 110 km/h. According to experience, the comfort and safety threshold for AVs was set at 0.15. To verify the effectiveness of the strategy, the shared car is controlled via the SCANeR studio scene script, so that AVs are within the shared zone of the shared car.
According to Decision Model 1, the variance value of active lane changing is 0.1145, which does not exceed 0.15. AVs want to turn left. But at this moment, the left turn is not executed. When the shared car agrees to the request, AVs start turning left, and the vehicle dynamics parameters of the AVs are shown in Figure 12.
The two actions of turning left and returning to the right direction after entering the left lane via the AVs result in the peak values of roll acceleration and lateral acceleration reaching their respective peaks within 1 s. Due to the short suspension-adjustment time, the roll acceleration and lateral acceleration exhibit nonlinear fluctuations between peak values. The larger the peak value, the greater the variance of lane changing, the greater the acceleration of the vehicle, and the less comfortable the passengers feel.

4.1.2. AVs Passively Change Lanes

As shown in Figure 13, based on the SCANeR studio scenario, the following road conditions are deployed. The speed of the AVs is 105 km/h, and they are driving normally in the middle lane. At this time, when encountering a faulty vehicle 125 m ahead, the AVs can choose to brake to avoid obstacles or change lanes to avoid obstacles. The shared car is controlled using SCANeR studio scene scripts to keep the AVs within the shared zone of the shared car.
The speed of the AVs is 105 km/h, and the distance from the faulty vehicle is 125 m. As shown in Figure 14a, based on the Decision Model 2 and priority rules, the AVs prioritize turning left, then braking, and finally turning right. The variance value is obtained via Decision Model 1. The braking variance value is 0.1354, and the left-turn variance value is 0.1139. The AVs want to turn left but have not executed a left turn at this moment. When the shared car agrees to the request, the AVs start turning left, and the roll acceleration and lateral acceleration of the AVs are shown in Figure 14b,c. When the shared car refuses the request, the AVs begin to decelerate and avoid obstacles. The braking force, longitudinal acceleration, and pitch acceleration of the AVs are shown in Figure 14d–f.
When the AVs encounter a faulty vehicle 125 m ahead at a speed of 105 km/h, the Decision Model 2 calculation result is as follows: within the range of 90 km/h to 100.6 km/h, the AVs turn right; within the range of 100.6 km/h to 110 km/h, the AVs turn left; when the speed is not within the high-speed range and there is a slow traffic jam, the AVs brakes. The shared car consent request situation is as follows: the AVs turn left and return to the right direction after entering the left lane, both of which cause the roll acceleration and lateral acceleration to reach their respective peaks within 1 s. Due to the long adjustment time of the suspension, the roll acceleration and lateral acceleration gradually move from nonlinear fluctuations to zero between peak values. The shared car rejection request situation is as follows: the AVs use nonlinear braking force to decelerate and brake to avoid obstacles. At the beginning of braking, the negative longitudinal acceleration increases sharply to slow down the vehicle, reaching its peak at 20 s during the simulation process. The AVs come to a complete stop between 20 s and 21 s, during which there are two peaks in pitch acceleration. The second peak is caused by inertia and suspension adjustment, and the short-term peak becomes the main factor affecting comfort.

4.2. Real Vehicle Verification

4.2.1. Experimental Platform

The testing site was a dual-lane road in the same direction, which was approximately 100 m long. The test vehicle model was Geometry C, as shown in Figure 15.
As shown in Figure 16, the vehicle uses the INS570D (Asensing, Guangzhou, China) high-precision vehicle integrated navigation and positioning system, which can achieve precise three-dimensional positioning and attitude measurement.

4.2.2. Verification of Feedback Testing

The actual vehicle verification process is shown in Figure 17.
Due to venue limitations, only active lane changing and shared-road-rights scenarios were verified. To consider mixed-flow scenarios, the shared car was driven by humans. According to experience, the comfort and safety threshold for AVs was set to 0.01. As shown in Figure 18, the speed of the shared car was 14 km/h, and the speed of the AVs was 13 km/h. The lane-change variance value obtained via Decision Model 1 was 0.0035, which did not exceed 0.01. The AVs wanted to turn left, but the left turn was not executed at this moment.
As shown in Figure 19, when the shared car agreed to the request, the AVs started to turn left and accelerate. The roll acceleration and lateral acceleration of the AVs are shown in Figure 20a,b, respectively.
The two actions of turning left and returning to the right direction after entering the left lane of the AVs resulted in the peak values of roll acceleration and lateral acceleration reaching their respective peaks within 1 s, with slower vehicle speeds and overall smaller peak values. Due to the acceleration and lane changing of the vehicle, the peak values of roll acceleration and lateral acceleration between 12 s and 13 s were greater than those between 9 s and 10 s. In comparison, acceleration caused the vehicle to shake more. The suspension had a long adjustment time, and the roll acceleration and lateral acceleration gradually fluctuated from non-linear to zero between peak values.

5. Discussion and Conclusions

This paper introduces an innovative driving strategy scheme, presenting a novel viewpoint and methodology to tackle the coexistence challenge of AVs and conventional vehicles. Table 1 comprehensively contrasts and highlights the key differences and unique features of the proposed driving strategies vis-à-vis the prevalent mainstream strategies. The driving strategy presented in this paper incorporates a dual-framework decision-making mechanism, grounded in comprehensible mathematical formulas that facilitate understanding. By utilizing a transparent mathematical model to allocate the right of way, the strategy enhances the transparency and predictability of autonomous driving decisions. Comfort is evaluated through the introduction of acceleration variance, enabling adjustable ride experiences. A sophisticated framework for passive lane changes is established, equipped with straightforward mathematical formulas to tackle emergency obstacle avoidance, thereby improving system responsiveness and safety while protecting both passengers and road users. The strategy proved effective across various scenarios, including high-speed, low-speed, and different-distance settings, demonstrating strong adaptability.
Despite the achievements of this driving strategy, several challenges remain. Firstly, to ensure smooth conduct in validation, the process of identifying the motion-state information and intentions of traditional vehicles via AVs was simplified. Secondly, there is room for improvement in the accuracy of the strategy, necessitating the recalculation and updating of traffic-state information once autonomous vehicles are approved. Lastly, during the validation process, the thresholds for comfort and safety were set based on experience. By setting different thresholds for comfort and safety, various driving styles can be derived, as shown in Table 2. Future work will investigate the relationship between driving styles and these thresholds for comfort and safety. Consequently, these issues require further discussion and resolution, and they will serve as key directions for our future research.
In conclusion, in response to the issue of right-of-way conflicts in mixed traffic, this paper has introduced an innovative shared-right-of-way driving strategy that comprises two decision-making frameworks, both grounded in straightforward mathematical formulas for ease of comprehension. One framework is dedicated to proactive lane changing; the right of way is evaluated using formulas, and the decision to change lanes is based on the variance of roll and lateral acceleration to ensure a comfortable transition. The other framework addresses reactive lane changing; acceleration is calculated upon encountering obstacles, and the obstacle avoidance method (either lane changing or deceleration) is determined based on vehicle speed and distance. Prior to lane changing, the framework also considers the right of way and requests for sharing if necessary, incorporating multi-axis acceleration variance to optimize the lane-changing experience. The strategy is validated as effective in various scenarios, including high speed (105 km/h), low speed (13 km/h), and with obstacles at a distance of 125 m, demonstrating its broad adaptability.

Author Contributions

Conceptualization, M.L.; Methodology, M.L., G.L. and C.S.; Software, M.L., G.L. and H.L.; Validation, G.L. and H.L.; Formal analysis, G.L.; Resources, M.L. and C.S.; Data curation, G.L. and C.S.; Writing—original draft, C.S., J.Y., H.L. and J.L.; Writing—review & editing, J.Y. and F.L.; Visualization, J.Y.; Supervision, J.Y.; Project administration, M.L.; Funding acquisition, M.L., C.S. and J.L. All authors have read and agreed to the published version of the manuscript.

Funding

This work is supported by the National Key R&D Program of China (2023YFB4302600), the Natural Science Foundation of Jiangsu Province (BK20231197, BK20220243), the Science and Technology Program of Suzhou (SYC2022078), the Hubei Science and Technology Talent Service Enterprise Project (2023DJC084, 2023DJC195), the Hubei Science and Technology Project (2021BEC005), the Structural Simulation of High Performance Hydrogen MPV R&D Project of Hainan Haima Automobile Co., Ltd. (HD-KYH-2022271), and the Natural Science Foundation of Hainan Province (521RC497).

Data Availability Statement

Data available on request from the authors.

Conflicts of Interest

The authors declare that this study received funding from the National Key R&D Program of China, the Natural Science Foundation of Jiangsu Province, the Science and Technology Program of Suzhou, the Hubei Science and Technology Talent Service Enterprise Project, the Hubei Science and Technology Project, the Structural Simulation of High Performance Hydrogen MPV R&D Project of Hainan Haima Automobile Co., Ltd., and the Natural Science Foundation of Hainan Province. The funders were not involved in the study design, collection, analysis, interpretation of data, the writing of this article or the decision to submit it for publication.

References

  1. Zhao, C.; Li, L.; Pei, X.; Li, Z.; Wu, X. A Comparative Study of State-of-the-Art Driving Strategies for Autonomous Vehicles. Accid. Anal. Prev. 2021, 150, 105937. [Google Scholar] [CrossRef]
  2. Choi, J.H.; Lee, H.J.; Oh, S.; Nam, K. Development of Vehicle Maneuvering System for Autonomous Driving. Mechatronics 2022, 85, 102798. [Google Scholar] [CrossRef]
  3. Cong, P.; Feng, H.; Li, S.; Li, T.; Xu, Y.; Zhang, X. A Visual Detection Algorithm for Autonomous Driving Road Environment Perception. Eng. Appl. Artif. Intell. 2024, 133, 108034. [Google Scholar] [CrossRef]
  4. Shalev-Shwartz, S.; Shammah, S.; Shashua, A. On a Formal Model of Safe and Scalable Self-Driving Cars. arXiv 2017, arXiv:1708.06374. [Google Scholar] [CrossRef]
  5. Liard, T.; Stern, R.; Delle Monache, M.L. Optimal Driving Strategies for Traffic Control with Autonomous Vehicles. IFAC-PapersOnLine 2020, 53, 5322–5329. [Google Scholar] [CrossRef]
  6. Li, Z.; Zhan, W.; Sun, L.; Chan, C.-Y.; Tomizuka, M. Adaptive Sampling-Based Motion Planning with a Non-Conservatively Defensive Strategy for Autonomous Driving. IFAC-PapersOnLine 2020, 53, 15632–15638. [Google Scholar] [CrossRef]
  7. Wang, Y.; Yuan, X.; Sun, C. Learning Autonomous Race Driving with Action Mapping Reinforcement Learning. ISA Trans. 2024, 150, 1–14. [Google Scholar] [CrossRef] [PubMed]
  8. Azam, S.; Munir, F.; Kyrki, V.; Kucner, T.P.; Jeon, M.; Pedrycz, W. Exploring Contextual Representation and Multi-Modality for End-to-End Autonomous Driving. Eng. Appl. Artif. Intell. 2024, 135, 108767. [Google Scholar] [CrossRef]
  9. Wang, Y.; Zhang, J.; Chen, Y.; Yuan, H.; Wu, C. Automatic Learning-Based Data Optimization Method for Autonomous Driving. Digit. Signal Process. 2024, 148, 104428. [Google Scholar] [CrossRef]
  10. Li, G.; Yang, Y.; Qu, X.; Cao, D.; Li, K. A Deep Learning Based Image Enhancement Approach for Autonomous Driving at Night. Knowl.-Based Syst. 2021, 213, 106617. [Google Scholar] [CrossRef]
  11. Lee, D.; Kwon, M. ADAS-RL: Safety Learning Approach for Stable Autonomous Driving. ICT Express 2022, 8, 479–483. [Google Scholar] [CrossRef]
  12. Hawlader, F.; Robinet, F.; Frank, R. Leveraging the Edge and Cloud for V2X-Based Real-Time Object Detection in Autonomous Driving. Comput. Commun. 2024, 213, 372–381. [Google Scholar] [CrossRef]
  13. Ding, H.; Pan, H.; Bai, H.; Zheng, X.; Chen, J.; Zhang, W. Driving Strategy of Connected and Autonomous Vehicles Based on Multiple Preceding Vehicles State Estimation in Mixed Vehicular Traffic. Phys. Stat. Mech. Its Appl. 2022, 596, 127154. [Google Scholar] [CrossRef]
  14. Shi, R.; Yang, S.; Chen, Y.; Wang, R.; Zhang, M.; Lu, J.; Cao, Y. CNN-Transformer for Visual-tactile Fusion Applied in Road Recognition of Autonomous Vehicles. Pattern Recognit. Lett. 2023, 166, 200–208. [Google Scholar] [CrossRef]
  15. Schitz, D.; Aschemann, H. Path Optimization for Autonomous Driving Using Deep Learning. IFAC-PapersOnLine 2022, 55, 490–496. [Google Scholar] [CrossRef]
  16. Liu, S.; Bai, Y. Multiple UAVs Collaborative Traffic Monitoring with Intention-Based Communication. Comput. Commun. 2023, 210, 116–129. [Google Scholar] [CrossRef]
  17. Eyuboglu, M.; Atali, G. A Novel Collaborative Path Planning Algorithm for 3-Wheel Omnidirectional Autonomous Mobile Robot. Robot. Auton. Syst. 2023, 169, 104527. [Google Scholar] [CrossRef]
  18. Petrillo, A.; Salvi, A.; Santini, S.; Valente, A.S. Adaptive Multi-Agents Synchronization for Collaborative Driving of Autonomous Vehicles with Multiple Communication Delays. Transp. Res. Part C Emerg. Technol. 2018, 86, 372–392. [Google Scholar] [CrossRef]
  19. Ghorai, P.; Eskandarian, A.; Kim, Y.-K.; Mehr, G. State Estimation and Motion Prediction of Vehicles and Vulnerable Road Users for Cooperative Autonomous Driving: A Survey. IEEE Trans. Intell. Transp. Syst. 2022, 23, 16983–17002. [Google Scholar] [CrossRef]
  20. Online Longitudinal Trajectory Planning for Connected and Autonomous Vehicles in Mixed Traffic Flow with Deep Reinforcement Learning Approach. J. Intell. Transp. Syst. 2023, 27, 396–410. [CrossRef]
  21. Peng, B.; Keskin, M.F.; Kulcsár, B.; Wymeersch, H. Connected Autonomous Vehicles for Improving Mixed Traffic Efficiency in Unsignalized Intersections with Deep Reinforcement Learning. Commun. Transp. Res. 2021, 1, 100017. [Google Scholar] [CrossRef]
  22. Li, Y.; Ma, D.; An, Z.; Wang, Z.; Zhong, Y.; Chen, S.; Feng, C. V2X-Sim: Multi-Agent Collaborative Perception Dataset and Benchmark for Autonomous Driving. IEEE Robot. Autom. Lett. 2022, 7, 10914–10921. [Google Scholar] [CrossRef]
  23. Sabouni, E.; Cassandras, C.G. Optimal Merging Control of an Autonomous Vehicle in Mixed Traffic: An Optimal Index Policy. IFAC-PapersOnLine 2023, 56, 2353–2358. [Google Scholar] [CrossRef]
  24. Zhao, C.; Li, Z.; Li, L.; Wu, X.; Wang, F. A Negotiation-based Right-of-way Assignment Strategy to Ensure Traffic Safety and Efficiency in Lane Changes. IET Intell. Transp. Syst. 2021, 15, 1345–1358. [Google Scholar] [CrossRef]
  25. Kaszas, D.; Roberts, A.C. Comfort with Varying Levels of Human Supervision in Self-Driving Cars: Determining Factors in Europe. Int. J. Transp. Sci. Technol. 2023, 12, 809–821. [Google Scholar] [CrossRef]
  26. Hou, X.; Gan, M.; Zhang, J.; Zhao, S.; Ji, Y. Vehicle Ride Comfort Optimization in the Post-Braking Phase Using Residual Reinforcement Learning. Adv. Eng. Inform. 2023, 58, 102198. [Google Scholar] [CrossRef]
  27. Deubel, C.; Ernst, S.; Prokop, G. Objective Evaluation Methods of Vehicle Ride Comfort—A Literature Review. J. Sound Vib. 2023, 548, 117515. [Google Scholar] [CrossRef]
  28. Han, C.; Han, T.; Ma, T.; Tong, Z.; Wang, S.; Hei, T. End-to-End BIM-Based Optimization for Dual-Objective Road Alignment Design with Driving Safety and Construction Cost Efficiency. Autom. Constr. 2023, 151, 104884. [Google Scholar] [CrossRef]
  29. De Winkel, K.N.; Irmak, T.; Happee, R.; Shyrokau, B. Standards for Passenger Comfort in Automated Vehicles: Acceleration and Jerk. Appl. Ergon. 2023, 106, 103881. [Google Scholar] [CrossRef] [PubMed]
  30. McConnell, W.A. Motion Sensitivity as a Guide to Road Design. SAE Trans. 1957, 65, 493–507. [Google Scholar]
Figure 1. The road rights of vehicle-following scenarios.
Figure 1. The road rights of vehicle-following scenarios.
Electronics 13 03214 g001
Figure 2. Road rights in mixed-flow scenarios.
Figure 2. Road rights in mixed-flow scenarios.
Electronics 13 03214 g002
Figure 3. Framework 1: active lane changing for shared road rights.
Figure 3. Framework 1: active lane changing for shared road rights.
Electronics 13 03214 g003
Figure 4. Framework 2: passive lane changing for shared road rights.
Figure 4. Framework 2: passive lane changing for shared road rights.
Electronics 13 03214 g004
Figure 5. AVs can merge directly.
Figure 5. AVs can merge directly.
Electronics 13 03214 g005
Figure 6. AVs need to request the sharing of merging.
Figure 6. AVs need to request the sharing of merging.
Electronics 13 03214 g006
Figure 7. Simulation framework. Part 1 receives the driver’s commands as an input, and it outputs vehicle status data; these data are then used as the input for Part 2, along with complex driver commands, including lane-keeping and lane changes for vehicles. The vehicle status data are also sent to Part 3 for interaction via the ComUDP protocol. Part 2 processes this information and outputs driver commands, including steering, throttle, brake pressure, and vehicle position. These outputs are also the output of Part 3 for further control or display. Part 3 provides scene information, including road information, sensor data, and scripts for controlling a shared car.
Figure 7. Simulation framework. Part 1 receives the driver’s commands as an input, and it outputs vehicle status data; these data are then used as the input for Part 2, along with complex driver commands, including lane-keeping and lane changes for vehicles. The vehicle status data are also sent to Part 3 for interaction via the ComUDP protocol. Part 2 processes this information and outputs driver commands, including steering, throttle, brake pressure, and vehicle position. These outputs are also the output of Part 3 for further control or display. Part 3 provides scene information, including road information, sensor data, and scripts for controlling a shared car.
Electronics 13 03214 g007
Figure 8. The explanation of Simulink (Part 1 and Part 2) in Figure 7.
Figure 8. The explanation of Simulink (Part 1 and Part 2) in Figure 7.
Electronics 13 03214 g008
Figure 9. The road used in simulation validation.
Figure 9. The road used in simulation validation.
Electronics 13 03214 g009
Figure 10. The detection range of the lidars with the road as the background.
Figure 10. The detection range of the lidars with the road as the background.
Electronics 13 03214 g010
Figure 11. AVs are verified their ability to actively change lanes in this scenario.
Figure 11. AVs are verified their ability to actively change lanes in this scenario.
Electronics 13 03214 g011
Figure 12. Vehicle dynamics parameters: (a) roll acceleration; (b) lateral acceleration.
Figure 12. Vehicle dynamics parameters: (a) roll acceleration; (b) lateral acceleration.
Electronics 13 03214 g012
Figure 13. AVs are verified their ability to passively change lanes in this scenario.
Figure 13. AVs are verified their ability to passively change lanes in this scenario.
Electronics 13 03214 g013
Figure 14. Vehicle dynamics parameters: (a) Decision Model 2; (b) roll acceleration; (c) lateral acceleration; (d) braking force; (e) longitudinal acceleration; (f) pitch acceleration.
Figure 14. Vehicle dynamics parameters: (a) Decision Model 2; (b) roll acceleration; (c) lateral acceleration; (d) braking force; (e) longitudinal acceleration; (f) pitch acceleration.
Electronics 13 03214 g014
Figure 15. Experimental vehicle.
Figure 15. Experimental vehicle.
Electronics 13 03214 g015
Figure 16. INS570D integrated navigation and positioning system.
Figure 16. INS570D integrated navigation and positioning system.
Electronics 13 03214 g016
Figure 17. Real vehicle verification process.
Figure 17. Real vehicle verification process.
Electronics 13 03214 g017
Figure 18. Active lane changing.
Figure 18. Active lane changing.
Electronics 13 03214 g018
Figure 19. Shared car consent request.
Figure 19. Shared car consent request.
Electronics 13 03214 g019
Figure 20. Vehicle dynamics parameters: (a) roll acceleration; (b) lateral acceleration.
Figure 20. Vehicle dynamics parameters: (a) roll acceleration; (b) lateral acceleration.
Electronics 13 03214 g020
Table 1. Comparison table of driving-strategy characteristics.
Table 1. Comparison table of driving-strategy characteristics.
Driving-Strategy CharacteristicsDriving Strategy in This PaperDefensive Driving StrategyCompetitive Driving StrategyCooperative Driving Strategy
Road-right sharingActively allocate road rights and strictly follow the road rightsLess emphasis on road right sharing, conservative drivingDoes not prioritize road-right sharing, may lead to conflictsInvolves road-right sharing but relies more on collaborative decision-making
Decision optimizationOptimized decision models tailored to different driving scenariosDecisions based on single safety criteriaPursues efficiency maximizationCollaborative optimization decisions based on information sharing
SafetyEnsure safety through multiple decision-making judgmentsHigh safety but may compromise efficiencySafety may be compromised due to competitionCollaborative efforts reduce conflicts, improving safety
Efficiency and comfortBalances safety with efficiency, enhancing passenger comfortLower efficiency but higher comfortHigher efficiency may compromise comfortCollaborative optimization achieves a balance between efficiency and comfort
AdaptabilityStrong adaptability to complex environments, applicable to high-speed and low-speed scenariosStrong adaptability to complex environments but conservative strategyGood adaptability for efficiency gains, but safety needs improvementRelies on collaborative systems, showing strong adaptability to new environments
Research contributionsProvides a new perspective on road right allocation in mixed traffic, advancing autonomous driving technologyEmphasizes safe driving standardsExplores efficient driving strategiesShowcases the potential of cooperative driving, driving intelligent transportation development
Table 2. Comparison of autonomous driving styles.
Table 2. Comparison of autonomous driving styles.
Style TypeCharacteristicsApplicable Scenarios
Ultimate comfortFocuses on passenger comfort, avoids abrupt maneuvers, and optimizes suspension, noise reduction, and seatingLong-distance travel, business transfers, and family outings
Balanced comfortBalances comfort with driving pleasure and flexibly adjusts to road conditionsDaily commuting, urban driving, and short trips
Sport–aggressiveEmphasizes driving dynamics and speed and provides responsive acceleration, high shift RPMs, and precise steering controlMountain roads, racetrack experiences, and performance-car demonstrations
Intelligent–adaptiveLeverages advanced sensors, algorithms, and AI technology to perceive road conditions, driving contexts, and passenger preferences, automatically adjusts driving style, and predicts and adapts to future driving situations for optimal driving experience and safetyAll-terrain, all-weather driving, especially scenarios requiring highly intelligent and personalized driving experiences
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Li, M.; Li, G.; Sun, C.; Yang, J.; Li, H.; Li, J.; Li, F. A Shared-Road-Rights Driving Strategy Based on Resolution Guidance for Right-of-Way Conflicts. Electronics 2024, 13, 3214. https://doi.org/10.3390/electronics13163214

AMA Style

Li M, Li G, Sun C, Yang J, Li H, Li J, Li F. A Shared-Road-Rights Driving Strategy Based on Resolution Guidance for Right-of-Way Conflicts. Electronics. 2024; 13(16):3214. https://doi.org/10.3390/electronics13163214

Chicago/Turabian Style

Li, Mei, Guisheng Li, Chuan Sun, Junru Yang, Haoran Li, Jialin Li, and Fei Li. 2024. "A Shared-Road-Rights Driving Strategy Based on Resolution Guidance for Right-of-Way Conflicts" Electronics 13, no. 16: 3214. https://doi.org/10.3390/electronics13163214

APA Style

Li, M., Li, G., Sun, C., Yang, J., Li, H., Li, J., & Li, F. (2024). A Shared-Road-Rights Driving Strategy Based on Resolution Guidance for Right-of-Way Conflicts. Electronics, 13(16), 3214. https://doi.org/10.3390/electronics13163214

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop