The traditional visual Simultaneous Localization and Mapping (SLAM) systems rely on the static-world assumption and cannot handle dynamic objects. This paper presents a novel SLAM system, Semantic Point and Line Features SLAM (DPL-SLAM), that can handle dynamic environments and can be used for real-time operation. To handle dynamic objects, we apply object detection to identify 80 categories within the scene and implement unique handling of features both within and outside the detected bounding boxes using Lucas-Kanade (LK) optical flow and epipolar constraint. Within bounding boxes, we propose an efficient local elimination algorithm to address features that violate the epipolar constraint. We designate nearby and intra-box regions that deviate from the constraint as potential dynamic areas, and conditionally eliminate features within these areas to varying extents, thus minimizing incorrect elimination of stable data associations. Outside the bounding boxes, non-compliant features are regarded as outliers and directly eliminated, making the system robust to unknown objects. We have evaluated DPL-SLAM on the TUM RGB-D and KITTI STEREO datasets and compared it with state-of-the-art SLAM systems. The results show that DPL-SLAM outperforms most SLAM systems in various dynamic scenarios and exhibits excellent robustness and realtime performance, thus effectively handling dynamic noise interference under indoor RGB-D and outdoor stereo modes. Finally, we conduct experiments in a real-world environment to verify the algorithm’s effectiveness.

The paper presents a vision-based obstacle avoidance strategy for lightweight self-driving cars that can be run on a CPU-only device using a single RGB-D camera. The method consists of two steps: visual perception and path planning. The visual perception part uses ORBSLAM3 enhanced with optical flow to estimate the car’s poses and extract rich texture information from the scene. In the path planning phase, the proposed method employs a method combining a control Lyapunov function and control barrier function in the form of a quadratic program (CLF-CBF-QP) together with an obstacle shape reconstruction process (SRP) to plan safe and stable trajectories. To validate the performance and robustness of the proposed method, simulation experiments were conducted with a car in various complex indoor environments using the Gazebo simulation environment. The proposed method can effectively avoid obstacles in the scenes. The proposed algorithm outperforms benchmark algorithms in achieving more stable and shorter trajectories across multiple simulated scenes.

Autonomous vehicles (AVs) promise substantial gains in safety, reliability, and decarbonization, yet safe and efficient interaction in dynamic, heterogeneous traffic remains a key barrier to large-scale deployment. Deep reinforcement learning (DRL) has emerged as a data-driven approach for learning adaptive decision policies that handle complex, unpredictable environments better than rule-based methods. However, different scenarios impose distinct requirements, necessitating scenario-specific algorithms. This survey systematically reviews DRL for four typical scenarios (highways, on-ramp merging, roundabouts, and unsignalized intersections), summarizes road features and recent advances, and evaluates methods using five criteria: driving safety, driving efficiency, training efficiency, unselfishness, and interpretability (DDTUI). Each DDTUI criterion is analyzed with respect to the reviewed algorithms. In addition, a dedicated scenario-centric learning transferability analysis is introduced that systematically evaluates whether each reviewed method demonstrates scene-specific learning improvements and assesses how effectively their designs transfer across the four scenarios. Finally, the challenges for future DRL-based decision-making algorithms are summarized.

Roundabout navigation presents significant challenges for autonomous vehicles (AVs) due to complex multi-vehicle interactions, highly dynamic and uncertain traffic patterns, and the requirement to coordinate with multiple vehicles simultaneously. Existing approaches often struggle to maintain safety and efficiency in uncertain vehicle behaviors and space-constrained situations, especially for mandatory lane changes and exit maneuvers. This paper presents a switched decision system that combines game theory and potential fields to enable robust navigation in roundabouts. The proposed approach first develops a Stackelberg game framework with uncertainty-aware prediction, where AVs dynamically adjust their decisions based on probabilistic estimates of surrounding vehicles' behaviors. Additionally, a hybrid potential field method is introduced that seamlessly transitions to emergency maneuvering when standard game-theoretic solutions become infeasible. Through extensive simulations under varied scenarios, the proposed approach achieves 97% successful navigation rates while maintaining collision rates below 1%, significantly outperforming baseline Nash equilibrium methods. The results demonstrate that integrating uncertainty characterization with strategic cooperation significantly improves the overall safety and efficiency of autonomous driving in roundabouts.

Decision making at unsignalized intersections presents significant challenges for autonomous vehicles (AVs), particularly in mixed traffic scenarios where both AVs and human-driven vehicles (HDVs) must safely coordinate their movements. This paper proposes a safety-critical multi-agent Monte Carlo tree search (MCTS) framework that integrates deterministic and probabilistic predictions to enable cooperative decision making in complex intersection scenarios. The framework incorporates three main innovations: 1) a safety assessment mechanism that systematically handles AV-to-AV (V2V), AV-to-HDV (V2H), and Vehicle-to-Road (V2R) interactions using dynamic safety thresholds and spatiotemporal risk metrics, 2)an adaptive HDV behavior awareness by combining the Intelligent Driver Model (IDM) with probabilistic distributions, and 3)a multi-objective reward function optimization approach that balances safety, efficiency, and cooperation. Extensive simulations demonstrate our framework’s efficacy and superior capability in ensuring safe and efficient intersection navigation across the fully-autonomous scenario (100% AVs) and challenging mixed traffic scenario (50% AVs +50% HDVs). Compared to benchmarks, our method reduces trajectory deviations by up to 37.56% in the fully-autonomous scenario and 62.43% in the mixed traffic scenario, while maintaining significantly lower Post-Encroachment Time (PET) violations (0% and 2.8%, respectively).