The focus of this dissertation is to improve autonomous navigation in unstructured terrain conditions, with specific application to unmanned casualty extraction in disaster scenarios. Robotic systems are being widely employed for search and rescue applications, especially in disaster scenarios. But a majority of these are focused solely on the search aspect of the problem. This dissertation proposes a conceptual design of a Semi-Autonomous Victim Extraction Robot (SAVER) capable of safe and effective unmanned casualty extraction, thereby reducing the risk to the lives of first responders. In addition, the proposed design addresses the limitations of existing state-of-the-art rescue robots specifically in the aspect of head and neck stabilization as well as fast and safe evacuation. One of the primary capabilities needed for effective casualty extraction is reliable navigation in unstructured terrain conditions. Autonomous navigation in unstructured terrain, particularly for systems with tracked locomotion mode involves unique challenges in path planning and trajectory tracking. The dynamics of robot-terrain interaction, along with additional factors such as slip experienced by the vehicle, slope of the terrain, and actuator limitations of the robotic system, need to be taken into consideration. To realize these capabilities, this dissertation proposes a hybrid navigation architecture that employs a physics engine to perform fast and accurate state expansion inside a graph-based planner. Tracked skid-steer systems experience significant slip, especially while turning. This greatly affects the trajectory tracking accuracy of the robot. In order to enable efficient trajectory tracking in varying terrain conditions, this dissertation proposes the use of an active disturbance rejection controller. The proposed controller is capable of estimating and counter acting the effects of slip in real-time to improve trajectory tracking. As an extension of the above application, this dissertation also proposes the use of support vector machine architecture to perform terrain identification, solely based on the estimated slip parameters. Combining all of the above techniques, an overall architecture is proposed to assist and inform tele-operation of tracked robotic systems in unstructured terrain conditions. All of the above proposed techniques have been validated through simulations and experiments in indoor and simple outdoor terrain conditions. / Doctor of Philosophy / This dissertation explores ways to improve autonomous navigation in unstructured terrain conditions, with specific applications to unmanned casualty extraction in disaster scenarios. Search and rescue applications often put the lives of first responders at risk. Using robotic systems for human rescue in disaster scenarios can keep first responders out of danger. To enable safe robotic casualty extraction, this dissertation proposes a novel rescue robot design concept named SAVER. The proposed design concept consists of several subsystems including a declining stretcher bed, head and neck support system, and robotic arms that conceptually enable safe casualty manipulation and extraction based on high-level commands issued by a remote operator. In order to enable autonomous navigation of the proposed conceptual system in challenging outdoor terrain conditions, this dissertation proposes improvements in planning, trajectory tracking control and terrain estimation. The proposed techniques are able to take into account the dynamic effects of robot-terrain interaction including slip experienced by the vehicle, slope of the terrain and actuator limitations. The proposed techniques have been validated through simulations and experiments in indoor and simple outdoor terrain conditions. The applicability of the above techniques in improving tele-operation of rescue robotic systems in unstructured terrain is also discussed at the end of this dissertation.

The field of autonomous robots have been explored more and more over the last decade. The combination of machine learning advances and increases in computational power have created possibilities to explore the usage of machine learning models on edge devices. The usage of object detection on edge devices is bottlenecked by the edge devices' limited computational power and they therefore have constraints when compared to the usage of machine learning models on other devices. This project explored the possibility to use real time object detection on a Raspberry Pi as input in different control systems. The Raspberry with the help of a coral USB accelerator was able to find a specified object and drive to it, and it did so successfully with all the control systems tested. As the robot was able to navigate to the specified object with all control systems, the possibility of using real time object detection in faster paced situations can be explored. / Ämnet autonoma robotar har blivit mer och mer undersökt under de senaste årtiondet. Kombinationen av maskin inlärnings förbättringar och ökade beräknings möjligheter hos datorer och chip har gjort det möjligt att undersöka användningen av maskin inlärningsmodeller på edge enheter. Användandet av bildigenkänning på edge enheter är begränsad av edge enheten begränsade datorkraft, och har därför mer begränsningar i jämförelse med om man använder bildigenkänning på en annan typ av enhet. Det här projektet har undersökt möjligheten att använda bildigenkänning i realtid som input för kontrollsystem på en Raspberry Pi. Raspberry Pien med hjälp av en Coral USB accelerator lyckades att lokalisera och köra till ett specificerat objekt, Raspberryn gjorde detta med alla kontrollsystem som testades på den. Eftersom roboten lyckades med detta, så öppnas möjligheten att använda bildigenkänning på edge enheter i snabbare situationer.