Obstacle avoidance and trajectory optimisation for a power line inspection robot.

Abstract

This dissertation presents the research, development and application of trajectory creation, obstacle avoidance and trajectory optimisation methods for an existing serial manipulator power line inspection robot (PLIR). The obstacle avoidance implementation allows the robot to navigate around an obstacle obstructing its navigation along the line. The algorithm generated end effector trajectory waypoints autonomously based on bounding box obstacle descriptions in Cartesian space, and connected them with a fifth order basis-spline end effector trajectory command. The trajectories were created taking into account the dynamic torque and velocity constraints of the robot while ignoring non-linearities. Performance was inspected and evaluated in a simulated workspace environment. The trajectory optimisation was designed to maximise the robot’s operating range, with constraints on the battery power supply, by minimising charge consumed during obstacle avoidance trajectories. The temporal components of the basis-spline trajectories were optimised by minimising a timeenergy type of cost function subject to the dynamic constraints of the robot. Cost function analyses are presented for a simple frictionless robot model based on the recursive Newton-Euler method, and for a more realistic model including viscous, Coulomb and static friction as well as gearbox backlash. It is shown that the Nelder-Mead simplex method was appropriate for optimisation. For representative trajectories that were studied, the optimiser was capable of finding global minima with satisfactory speed and accuracy in simulation. The validity of trajectory optimisation with regard to the cost function behaviour was confirmed. This was based on experiments carried out on the robot hardware in the laboratory, examining the predicted and actual actuator current profiles. The engineering design and implementation of hardware and software for the base station and on-board system is presented, together with the layout of the PLIR’s control system and PID (proportional-integral-derivative) controller design. Trajectory commands are sent from the base station to the robot via Wi-Fi for execution. Furthermore, live video feed from the robot can be sent to the ground station computer. Furthermore, high voltage testing of the PLIR showed that the engineering design of the robot and communication platform is robust.