Self-Tuning Method for Increased Obstacle Detection Reliability Based on Internet of Things LiDAR Sensor Models.

Area of Science:

• Robotics and Autonomous Systems
• Sensor Networks
• Computational Intelligence

Background:

• Vehicle collision avoidance relies heavily on reliable obstacle detection by Light Detection and Ranging (LiDAR) sensors.
• Optimizing LiDAR sensor network reliability is crucial for Internet of Things (IoT) mobility scenarios.

Purpose of the Study:

• To design and implement a self-tuning methodology to maximize the reliability of LiDAR sensor networks for obstacle detection.
• To enhance obstacle localization accuracy in complex driving environments using computational intelligence.

Main Methods:

• Utilized Webots Automobile 3D simulation for sensor interaction emulation.
• Developed a model-based framework with point-cloud clustering and an error-based prediction model library (MLP, KNN, Linear Regression).
• Implemented Q-learning (reinforcement learning) to determine optimal LiDAR sensor count for reliability.

Main Results:

• A five-LiDAR sensor network in an IoT driving assistance scenario was simulated and implemented.
• The self-tuning method demonstrated increased sensor network reliability.
• Minimized detection thresholds were achieved while enhancing obstacle localization.

Conclusions:

• The proposed self-tuning methodology effectively increases LiDAR sensor network reliability for obstacle detection in IoT mobility.
• Computational intelligence, particularly reinforcement learning, provides a viable approach for optimizing sensor network performance.
• This framework contributes to safer autonomous driving systems through enhanced sensor data processing.