Intelligent Robotics

Walking Robot
We designed control algorithms for a bipedal balancing robot with only pegs for feet

Generic Robotic Challenges

There are a number of challenges involved in controlling a robot. These may be broadly grouped as perceptual challenges, reasoning challenges, and control challenges. For a mobile robot, perceptual challenges may include such tasks as determining the robot’s location at any point in time, determining the locations and sizes of obstacles (including the terrain itself), and understanding spoken commands. For a fixed robot, such as one on a factory floor, perceptual challenges might include determining the location and orientation of a part on a conveyor belt based on images from a camera.

Reasoning challenges are problems of trying to figure out how the robot can accomplish its goal. These include exploration strategies for mapping out the office that a mobile robot must navigate through, finding the best way to get to a room, and determining whether or not it should try to do more before it recharges its battery.

Control challenges involve trying to make the robot’s body perform a desired function. This usually involves a fairly detailed physical model of the robot, although sometimes it is preferable to have the robot learn this automatically. In a complex robot, there are often many ways for the robot to accomplish the same thing, so control becomes a matter of choosing the best strategy, which balances the probability of success, the amount of time the action takes, and the amount of energy the robot expends. All of this can come into play in a task as simple as moving an object from one conveyor belt to another.

Special Factory Robot Challenges

Robots on factory floors or other similar environments have unique challenges. Their sensory and reasoning problems are often simpler than with other types of robots, but their control problems are usually intensified by the requirement of very precise movements. Despite the precision of the robot being of great concern to its user, robotics manufacturers typically do not specify it carefully. As a result, users can purchase the robot, install it into their process, and start it running before finding out that it will require careful tuning or may in fact be completely inadequate for the task they are attempting to use it for.

The accuracy of a robot is the measure of how precisely it can move its flange (the tool attached to it) to a desired location. How close it gets to this target is dependent on what motion the robot performed to get there, so accuracy is the distance from the target using the hardest motion. Since this distance varies from trial to trial, its value is usually taken to be the worst measurement over some reasonable number of trials.

The accuracy is not just a function of the robot, but also a function of the mathematical model of the robot. This model specifies what the joint angles should be in order for the robot’s flange to reach a particular location. If the model of the robot is inaccurate, then the desired joint angles will be off, and the flange will not reach the target. Even when manufacturers specify a model of the robot, it is often just a rough guess based on the design specs of the robot. The best model will vary from robot to robot, even if they are the same type of robot. The only way to get the highest accuracy out of a robot is to test the robot out and develop a model of it.

Whether you are setting up a new robot or just need to improve the performance or increase the intelligence of an existing robot, Shaw Technologies can help you, so please contact us today!