Collective Perception: A Safety Perspective

Towards Robust Decision-Making for Autonomous Highway Driving Based on Safe Reinforcement Learning

Reinforcement Learning (RL) methods are regarded as effective for designing autonomous driving policies. However, even when RL policies are trained to convergence, ensuring their robust safety remains a challenge, particularly in long-tail data. Therefore, decision-making based on RL must adequately consider potential variations in data distribution. This paper presents a framework for highway autonomous driving decisions that prioritizes both safety and robustness. Utilizing the proposed Replay Buffer Constrained Policy Optimization (RECPO) method, this framework updates RL strategies to maximize rewards while ensuring that the policies always remain within safety constraints. We incorporate importance sampling techniques to collect and store data in a Replay buffer during agent operation, allowing the reutilization of data from old policies for training new policy models, thus mitigating potential catastrophic forgetting. Additionally, we transform the highway autonomous driving decision problem into a Constrained Markov Decision Process (CMDP) and apply our proposed RECPO for training, optimizing highway driving policies. Finally, we deploy our method in the CARLA simulation environment and compare its performance in typical highway scenarios against traditional CPO, current advanced strategies based on Deep Deterministic Policy Gradient (DDPG), and IDM + MOBIL (Intelligent Driver Model and the model for minimizing overall braking induced by lane changes). The results show that our framework significantly enhances model convergence speed, safety, and decision-making stability, achieving a zero-collision rate in highway autonomous driving.

The Impact of Line-of-Sight and Connected Vehicle Technology on Mitigating and Preventing Crash and Near-Crash Events

Line-of-sight (LOS) sensors developed in newer vehicles have the potential to help avoid crash and near-crash scenarios with advanced driving-assistance systems; furthermore, connected vehicle technologies (CVT) also have a promising role in advancing vehicle safety. This study used crash and near-crash events from the Second Strategic Highway Research Program Naturalistic Driving Study (SHRP2 NDS) to reconstruct crash events so that the applicable benefit of sensors in LOS systems and CVT can be compared. The benefits of CVT over LOS systems include additional reaction time before a predicted crash, as well as a lower deceleration value needed to prevent a crash. This work acts as a baseline effort to determine the potential safety benefits of CVT-enabled systems over LOS sensors alone.

Perception Methods for Adverse Weather Based on Vehicle Infrastructure Cooperation System: A Review

Environment perception plays a crucial role in autonomous driving technology. However, various factors such as adverse weather conditions and limitations in sensing equipment contribute to low perception accuracy and a restricted field of view. As a result, intelligent connected vehicles (ICVs) are currently only capable of achieving autonomous driving in specific scenarios. This paper conducts an analysis of the current studies on image or point cloud processing and cooperative perception, and summarizes three key aspects: data pre-processing methods, multi-sensor data fusion methods, and vehicle-infrastructure cooperative perception methods. Data pre-processing methods summarize the processing of point cloud data and image data in snow, rain and fog. Multi-sensor data fusion methods analyze the studies on image fusion, point cloud fusion and image-point cloud fusion. Because communication channel resources are limited, the vehicle-infrastructure cooperative perception methods discuss the fusion and sharing strategies for cooperative perception information to expand the range of perception for ICVs and achieve an optimal distribution of perception information. Finally, according to the analysis of the existing studies, the paper proposes future research directions for cooperative perception in adverse weather conditions.

Two-Level Sharing and Extraction of Sensing Information with Hybrid V2X Communications

Vehicle-to-Everything (V2X) communications provide opportunities for information sharing among vehicles, edge servers, and cloud services. By the collection and extraction of sensing information from vehicles, such as communication quality or free space size, the edge server in V2X communications can improve its sensing and perception coverage. However, the collection of sensing data from vehicles consumes a large amount of wireless resources and computing resources at the edge server. The objective of this study is to extract object sensing information from vehicles, including the minimum or maximum of the sensing values, with low resource consumption and with high scalability. We propose a method that transforms the extraction of sensing information into a two-level procedure that includes (1) the local sharing and extraction of sensing information among vehicles and (2) the efficient extraction of sensing information at the edge server. Moreover, hybrid communication methods are employed at vehicles, with a short range of communication between vehicles to reduce the consumption of wireless resources for the local sharing of sensing data. The evaluation results show that the proposed method highly reduces the number of reports from the vehicles to the edge server, with a small amount of network resource consumption and scalability.

Compacted Area with Effective Links (CAEL) for Data Dissemination in VANETs

Vehicular ad hoc network (VANET) is a specialized form of wireless network that is solely intended for collaboration between vehicles. Several studies have shown that standard routing protocols cannot be implemented in VANETs because of their unique characteristics such as their significant count of vehicles on the network and the rapid evolution of the network’s design. Because VANET communication links are broken very frequently, it is necessary to address the routing consistency of these highly dynamic networks. The transmission of VANET data may result in a substantial amount of overhead in the routing process; thus, it is vital to address the issue of overhead to enhance the overall network performance. The proposed protocol named compacted area with effective links (CAEL) is designed to focus on decreasing overhead to achieve an enhancement in PDR performance inside the network. The communication between selected nodes that have been judged to be dependable in terms of geographical location and appropriate existing links between vehicles is focused on achieving this goal. With the inclusion of the reliability factor, it is possible to complete the important step of removing extraneous nodes, with the selection of the trustworthy nodes being made based on the link expiration time during the whole routing procedure. When compared to our previously published protocols, i.e., Dynamic Trilateral Enrollment (DyTE) and Reliable Group of Vehicles (RGoV), the results of the simulations demonstrate that CAEL has achieved an overall improvement in the performance of the network.

Context-Aware Content Selection and Message Generation for Collective Perception Services

Collective perception messages (CPMs) inform neighbors about detected objects and their current status. Although CPMs can greatly contribute to improving vehicle perception, their performance will degrade if the wireless channel is saturated and/or vehicles must process excessively high amount of data to build such a perception. Channel saturation and processing overhead can be induced by transmissions of CPMs if vehicles and roadside infrastructure transmit CPMs too often with too much redundant information. Indeed, according to the current draft standard, the CPM generation frequency at individual vehicles and infrastructure can be as high as 10 Hz and depending on the perception capability, individual CPMs can be quite heavy (around 1500 Bytes), containing redundant information. It is hence important to integrate techniques that reduce information redundancy while providing sufficient perception with optimal resource use. In this paper, we propose context-aware communication schemes that control CPM content selection and transmission. Two types of contexts, particularly radio resource use (channel busy ratio: CBR) and infrastructure availability, have been considered in the proposed schemes. Using a network simulator, we evaluated the performances of the proposed schemes in terms of CBR, packet delivery ratio and awareness ratio. The simulation results show that the scheme that takes into account both resource use and infrastructure availability offers the best performance on all the above metrics.

Analysis of Cooperative Perception in Ant Traffic and Its Effects on Transportation System by Using a Congestion-Free Ant-Trail Model

We investigated agent-based model simulations that mimic an ant transportation system to analyze the cooperative perception and communication in the system. On a trail, ants use cooperative perception through chemotaxis to maintain a constant average velocity irrespective of their density, thereby avoiding traffic jams. Using model simulations and approximate mathematical representations, we analyzed various aspects of the communication system and their effects on cooperative perception in ant traffic. Based on the analysis, insights about the cooperative perception of ants which facilitate decentralized self-organization is presented. We also present values of communication-parameters in ant traffic, where the system conveys traffic conditions to individual ants, which ants use to self-organize and avoid traffic-jams. The mathematical analysis also verifies our findings and provides a better understanding of various model parameters leading to model improvements.