Development of a robotic system with hybrid locomotion for both indoor and outdoor fire detection operations

In this thesis; a robotic system with ladder & obstacle climbing and fire detection capabilities was developed. The robotic system was designed and produced as adaptable for both indoor and outdoor applications. An adaptive three wheel legged locomotion system to provide the obstacle climbing and straight motions was developed. The mathematical models were derived for the hybrid locomotion system. "Direction Based Angle Calculation" approach was proposed and the required experimental tests were applied to check the performance of the robotic system with the developed algorithm. A motion decision algorithm to classify the obstacles as ladder, negligible or negotiable according to the height and shape of the obstacle was structured. "Fire Search and Find" and "Fire Detection" algorithms were constructed to find the fire candidates and to determine the probabilities. Several experimental tests were applied to the mechanicals systems and algorithms of the ladder climbing fire detection robot (Performance tests of motion and transmission systems, local path planning and obstacle avoidance, motion mode decision, fire detection). The obtained test results showed that the transmission and motion systems are capable for both ladder climbing and linear motions. The direction based angle calculation approach is suitable and satisfactory for local path planning and obstacle avoidance applications. The developed fire detection algorithm using the Faster R-CNN deep learning model also determines the probability of a fire source with the accuracy of 93%. Key Words: Obstacle Avoidance, Faster R-CNN Deep Learning Model, Hybrid Locomotion, Local Path Planning, Fire Detection Robot, Fire Search and Find