Autonomous driving at unsignalized intersections is still considered a challenging application for machine learning due to the complications associated with handling complex multi-agent scenarios characterized by a high degree of uncertainty. Automating the decision-making process in these safety-critical environments involves comprehending multiple levels of abstraction associated with learning robust driving behaviors to enable the vehicle to navigate efficiently. In this survey, we aim at exploring the state-of-the-art techniques implemented for decision-making applications, with a focus on algorithms that combine Reinforcement Learning (RL) and deep learning for learning traversing policies at unsignalized intersections. The reviewed schemes vary in the proposed driving scenario, in the assumptions made for the used intersection model, in the tackled challenges, and in the learning algorithms that are used. We have presented comparisons for these techniques to highlight their limitations and strengths. Based on our in-depth investigation, it can be discerned that a robust decision-making scheme for navigating real-world unsignalized intersection has yet to be developed. Along with our analysis and discussion, we recommend potential research directions encouraging the interested players to tackle the highlighted challenges. By adhering to our recommendations, decision-making architectures that are both non-overcautious and safe, yet feasible, can be trained and validated in real-world unsignalized intersections environments.

When applied to autonomous vehicle (AV) settings, action recognition can enhance an environment model's situational awareness. This is especially prevalent in scenarios where traditional geometric descriptions and heuristics in AVs are insufficient. However, action recognition has traditionally been studied for humans, and its limited adaptability to noisy, un-clipped, un-pampered, raw RGB data has limited its application in other fields. To push for the advancement and adoption of action recognition into AVs, this work proposes a novel two-stage action recognition system, termed RALACs. RALACs formulates the problem of action recognition for road scenes, and bridges the gap between it and the established field of human action recognition. This work shows how attention layers can be useful for encoding the relations across agents, and stresses how such a scheme can be class-agnostic. Furthermore, to address the dynamic nature of agents on the road, RALACs constructs a novel approach to adapting Region of Interest (ROI) alignment to agent tracks for downstream action classification. Finally, our scheme also considers the problem of active agent detection, and utilizes a novel application of fusing optical flow maps to discern relevant agents in a road scene. We show that our proposed scheme can outperform the baseline on the ICCV2021 Road Challenge dataset (Singh et al., 2023) algorithm and by deploying it on a real vehicle platform, we provide preliminary insight to the usefulness of action recognition in decision making. The code is publicly available at https://github.com/WATonomous/action-classification.

Trajectory planning and control have historically been separated into two modules in automated driving stacks. Trajectory planning focuses on higher-level tasks like avoiding obstacles and staying on the road surface, whereas the controller tries its best to follow an ever changing reference trajectory. We argue that this separation is (1) flawed due to the mismatch between planned trajectories and what the controller can feasibly execute, and (2) unnecessary due to the flexibility of the model predictive control (MPC) paradigm. Instead, in this paper, we present a unified MPC-based trajectory planning and control scheme that guarantees feasibility with respect to road boundaries, the static and dynamic environment, and enforces passenger comfort constraints. The scheme is evaluated rigorously in a variety of scenarios focused on proving the effectiveness of the optimal control problem (OCP) design and real-time solution methods. The prototype code will be released at github.com/WATonomous/control.

Reinforcement learning-based techniques, empowered by deep-structured neural nets, have demonstrated superiority over rule-based methods in terms of making high-level behavioral decisions due to qualities related to handling large state spaces. Nonetheless, their training time, sample efficiency and the feasibility of the learnt behaviors remain key concerns. In this paper, we propose a novel hierarchical reinforcement learning-based decision-making architecture for learning left-turn policies at unsignalized intersections with feasibility guarantees. The proposed technique is comprised of two layers; a high-level learning-based behavioral planning layer which adopts soft actor-critic (SAC) principles to learn high-level, non-conservative yet safe, driving behaviors, and a low-level motion planning layer that uses Model Predictive Control (MPC) framework to ensure feasibility of the two-dimensional left-turn maneuver. The high-level layer generates reference signals of velocity and yaw angles for the ego vehicle taking into account safety and collision avoidance with the intersection vehicles, whereas the low-level motion planning layer solves an optimization problem to track these reference commands taking into account several vehicle dynamic constraints and ride comfort. While training the behavioral SAC-based planning layer, We develop an adaptive entropy regularization technique that results in faster convergence, higher mean rewards, and lower collision rates. We validate the proposed decision-making scheme in simulated environments and compare with other model free Reinforcement Learning (RL) baselines. The results demonstrate that the proposed integrated framework possesses better training and navigation capabilities.

Environment modeling is the backbone of how autonomous agents understand the world, and therefore has significant implications for decision-making and verification. Motivated by the success of relational mapping tools such as Lanelet2, we present the Dynamic Relation Graph (DRG). The DRG is a novel method for extending prior relational maps to include online observations, creating a unified en-vironment model which incorporates both prior and online data sources. Our prototype implementation models a finite set of heterogeneous features including road signage and pedestrian movement. However, the methodology behind the DRG can be expanded to a wider range of features in a fashion that does not increase the complexity of behavioral planning. Simulated stress tests indicate the DRG's effectiveness in decreasing decision-making complexity, and deployment on the University of Waterloo's WATonomous research vehicle demonstrates its practical utility. The prototype code will be released at github.com/WATonomous/DRG.

While supervised detection and classification frameworks in autonomous driving require large labelled datasets to converge, Unsupervised Domain Adaptation (UDA) approaches, facilitated by synthetic data generated from photoreal simulated environments, are considered low-cost and less time-consuming solutions. In this paper, we propose UDA schemes using adversarial discriminative and generative methods for lane detection and classification applications in autonomous driving. We also present Simulanes dataset generator to create a synthetic dataset that is naturalistic utilizing CARLA’s vast traffic scenarios and weather conditions. The proposed UDA frameworks take the synthesized dataset with labels as the source domain, whereas the target domain is the unlabelled real-world data. Using adversarial generative and feature discriminators, the learnt models are tuned to predict the lane location and class in the target domain. The proposed techniques are evaluated using both real-world and our synthetic datasets. The results manifest that the proposed methods have shown superiority over other baseline schemes in terms of detection and classification accuracy and consistency. The ablation study reveals that the size of the simulation dataset plays important roles in the classification performance of the proposed methods. Our UDA frameworks are available at github.com/anita-hu/sim2real-lane-detection and our dataset generator is released at github.com/anita-hu/simulanes.