No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Established robot control technology, on the whole, relies on the assumption that the links between joints are fully rigid and therefore can be used to accurately and repeatably calculate where the end effector is in relation to the base, by calculation involving the angles of the joints and the length of the links. Assuming the links and drive mechanisms are rigid and accurately machined, this can be used to create a robot that is very accurate and highly repeatable - two key properties of manipulators. Accuracy can be described as a measure of the absolute error in manipulator position from the chosen target, repeatability as a measure of the spread of positions on successive iterations of the same process. A system can exhibit high repeatability but low accuracy; repeatedly reaching exactly the same location, but some distance from where it should be. A system cannot exhibit high accuracy and low repeatability; accuracy implies an ability to move close to a target on each attempt. The advantage of robots with accurately machined, rigid links connected by high precision encoders and drive gears are the relatively simple calculations of kinematics and dynamics of the system. The disadvantages are: • The financial expense of the precision machining of the links. • The weight and size of the links required for lifting reasonable loads without link flex destroying the accuracy of the end effector position data. • The cost of the j'!int motors and encoders - high precision, high torque motors are expensive. When the robot cannot be manufactured as a rigid system, such as for use in tasks where low manipulator mass is crucial, the equations of motion of the robot are very complex and specific to that manipulator and payload. Methods of reducing the modeling required use large numbers of additional sensors, their signal processing systems add complexity and cost to the systems. These increased costs are justifiable in applications such as operations in space, however they counteract the savings that could be used to reduce the cost of general-purpose robots. By the application of end-effector mounted computer vision, so called Eye-In-Hand (Elli), providing feedback control to the basic control system, flexibility in the links can be tolerated and compensated for without modeling of the link flexibility. This allows the robot to be manufactured from cheaper links and drives, significantly reducing the cost of the robot. This may allow the introduction of robots into areas not currently viable for financial reasons such as low-value-adding processes where a robot would be considered useful but too expensive to justify. Existing methods of working around flexibility employ either fault-tolerant gripper technology or complex modeling and feedback from a large number of sensors. This thesis details the creation and testing of a single high speed EIH visual control system to compensate for unknown manipulator link flex. This is done with a view to assessing the feasibility of producing and' controlling a compliant robot with minimal sensors or modeling, which would allow reduction in its manufacturing costs.

No description available.