Robotics technology before the year 2000

 ... was remarkable less developed. The most advanced approach available during this time period was model predictive control with the RRT algorithm. This allows a robot to follow a fixed trajectory. For example if the robot is 10 centimeter away from the floor's trajecoty, the MPC planner ensures that the robot gets back on the track. No additional features are available, but moving on a fixed trajectory for example in a warehouse was everything which can be realized with an RRT based motion planning algorithm.

In addition, it should be mentioned that the combination of rapidly exploring random tree and model predictive control was a highly advanced technique before the year 2000. RRT is more efficient than other solvers like A*, and model predictive control is based on a physical model of the robot including its movement abilities. This allows to control a wheeled robot and a UAV both.

From today's perspective its surprising, that the described mathematical optimization algorithm requies on the one hand advanced knowledge in computer science including artificial intelligence on a university phd level and at the same time, the resulting robot is a simple line following robot which can't be scaled up to more advanced problems. This my explain why AI before 2000 was seen as difficult to realize and most of the projects have failed. That means, even advanced mathematicans with 20 years of practical experiences in optimization problems were only able to program a line following robot which was able to move along a fixed line on the ground. From this dispointing reality it seems rational to assume that AI can't be realized at all.