In this study, we examine an alternative approach to achieving collision avoidance in dynamic indoor environments by integrating deep-learning-based object detection (YOLOv5) with 2D LiDAR distance estimation through image-to-angle mapping techniques. The system utilizes a rule-based decision-making framework that dynamically adjusts the robot’s trajectory based on obstacle proximity and predefined safe distance thresholds. Instead of relying on computationally expensive 3D sensors, the proposed method leverages mid-level sensor fusion to align bounding box outputs with LiDAR angular scans, enabling reliable distance estimation for real-time obstacle avoidance necessary for safe navigation. The methodology was evaluated using both RViz-based visualization and real-world testing in a controlled indoor environment. Empirical findings confirm the system’s capability to perform reliable collision avoidance and adaptive navigation. Results showed reliable collision avoidance and adaptive navigation performance. With Kalman filtering, 2D LiDAR distance estimation achieved an average error of 0.06 m. The system recorded a 95% success rate for static obstacles and a less desirable success rate of 60% for dynamic obstacles. This work advances autonomous indoor navigation by providing a computationally simple, cost-saving framework that harnesses deep learning-based computer vision and 2D LiDAR sensing and is scalable and well-suited for multi-robot deployment in dynamic environments.