Reactive/agent based planners which have been previously applied to mobile robots with autonomous tasks are extended to fixed robot arms in an obstacle rich environment. Initial work describes how the environment is represented to provide both information to the planner and mimic potential sensor information. Basic behaviours are then derived which enable a robot to move in planar space, an outline of simple local minima avoidance methods is then given. This 2D planner is then extended to the more general classification of an industrial robot arm with 3D planning with allowance for prismatic and revolute joints. Static and moving obstacles can be handled by the planner and it is especially useful in unstructured environments and real-time situations.