Development and evaluation of obstacle detection mechanisms for autonomous aerial systems

Autonomous robots rely heavily on avoidance algorithms and do not typically use obstacle contact detection, which limits their usefulness in congested situations. Blind spots or discretized detection locations plague current contact-detection systems, and performance is further limited by stiff platforms that only detect collisions without offering meaningful feedback. We offer a novel architecture for contact sensors that enhances autonomous navigation by detecting physical contact as a solution to these problems. The system uses a flex-sensor-equipped elastic collision platform to measure displacements in the event of a collision. Data collected by flex sensors is transformed into useful contact information by a contact-detection algorithm that were based on a neural network. By utilizing sensor data for real-time collision recovery, collision system was tested with collisions that occurred during both autonomous contact-based missions and manual flights. The testing findings showed that the system could accurately detect collision parameters and contact events estimate, even in scenarios when the parameters were changing. Future research on contact-based navigation systems can build on the suggested methodology, which offers a strong method for improving autonomous navigation in complicated surroundings.