A telecontrolled robotic platform

The present work is part of the development of a semi-autonomously controlled robot. For our purposes an instance or example of a semi-autonomous robot is one that would be able to locate, identify and exterminate weeds in a lawn. The present system is controlled remotely with the help of a vision system, aided by an obstacle avoidance subsystem. The semi-autonomous functionality would become active once the robot was coarsely positioned. In this manner an operator would maneuver the robot assisted by sensor feedback and once in position an apparatus would take over the semi-autonomous function or identifying and extracting weeds. The main focus of the thesis research here is the platform itself and the coarse telecontrol positioning of the robot aided by sensor feedback. This includes the design an implementation of an extendable architecture borrowing and integrating concepts from embedded systems, networking, network programming, and power electronics. The robot consists of sensors, motor drive circuits and different machine elements. The platform is an example of a new generation of personal service robot as identified by S. Thrun [1] among others who have noted that autonomy is one of the major challenges facing the robotics community. The long term goal of the research is to contribute to machine learning operating in conditions of uncertainty with the platform developed providing a research vehicle for future research.

The actual semi-autonomous task will make use of a decision making controller or similar process for locating weeds in its field of view once maneuvered into position under remote control of an operator. More specifically in the present instance an operator can remotely control the system using a steering system with feedback provided by video and relatively simple sensors. Ideally, a number of accelerometers would be mounted to the chassis to sense the vibrations caused by the environment. These would then be transmitted to the user along with the video image. In this manner the user can feel the vibrations at the control console reflecting those encountered on the real chassis. Challenges at present include the synchronization of the sensory data along with the video. The variable delay associated with the video also presents a major challenge for any remotely controlled device and even more challenging here as all control is over an Internet Protocol (TCP/IP).

In summary this thesis examined various architectures and developed an extendable platform or testbed suitable for addressing robotic autonomy. The core technologies include capitalizing on Internet protocols, Ethernet onboard the mobile, embedded linux processors or reconfigurable hardware that when combined provide for a modular testbed. The wireless technology is 802.11 although could be extended to any that support TCP/IP. The work is oriented towards open source and standards making it of greater utility to those who want to extend or use the base platform. Perhaps the most significant aspect of the work is that the basic platform developed here, can be used for a variety of applications requiring the inclusion of telecontrol and semi-autonomous functionality without having to reinvent the core building blocks.