A study of object detection and avoidance for an autonomous roadside mowing application

California is a state of diverse landscapes and an ever developing population. The citizens of California depend heavily on the state's vast highway system for transportation. As one could imagine, maintenance of this highway system is a very difficult task. Specifically, vegetation control along the roadside poses a very challenging problem. It also creates a safety hazard to those that are assigned with the task of maintaining it. The California Department of Transportation (Caltrans) is the chief organization responsible for roadway maintenance. Caltrans has demonstrated interest in discovering new ways to improve their current methods of vegetation control. These new methods would promote roadside worker and public safety, help decrease the amount of herbicide application, increase process efficiency, and reduce costs.;The Advance Highway Maintenance and Construction Technology (AHMCT) Research Center has been assisting Caltrans with the study needed to expand upon new methods of vegetation control. There are numerous possibilities that have been considered for further research and development including but not limited to hot foam or steam application, the use infrared heat, hydraulic obliteration and vegetation cutting attachments for the Automated Roadside Debris Vacuum (ARDVAC). Most recently, AHMCT has conducted research for Caltrans to further develop the concept and implementation of an automated roadside mowing system. The ultimate goal of the system would be to have a fleet of autonomous mowing robots. The fleet could be deployed on the roadside at certain intervals where they would commence mowing of the roadside. All the while the robots would need minimal interaction from a Caltrans worker, as they would be completely autonomous. After the sections of roadside, to be mowed, were completed the robots could be recovered.;A robotic test platform was designed and constructed based on the concept of a GPS guided, autonomous mower also known as, "Autonomower". A control system to guide the robot on a back and forth, "strip fill" type pattern was also designed and implemented in software. The control system was tested as to the accuracy at which it could maneuver the platform along this path across an area that was input by a user. The purpose of this platform, in terms of research, was to provide the tools necessary to develop additional support systems that would be required for an autonomous roadside machine of any kind.;The purpose of this thesis will be to take a look at one such support system, object detection and avoidance. First, a complete understanding of the typical roadside environment that may be encountered by the robot is necessary. The environment will have some effect on what sensors and algorithm are finally selected. The hardware and software elements required to perform the task of obstacle detection and avoidance will need to be determined. A literature review will be conducted for object detection and avoidance algorithms that have been used in previous research projects. Not all obstacle avoidance algorithms are created equal, so each algorithm will be evaluated to determine its perceived effectiveness for an autonomous roadside mowing application. A thorough search of object detection sensors will also be conducted. The goal of this search will be to determine what sensor or combination of sensors would allow for the most accurate and repeatable detection of obstacles. A single sensor type will be selected for initial testing and data quantification. Next, as recommended from previous research, the radio control feature will be implemented on the robot. This feature could pose some useful functional possibilities such as easy robot deployment and retrieval, mowing boundary teaching, and remote emergency stop. Finally, conclusions and future research considerations will be addressed.