Servo - December 2016

During a recent discussion with fellow robot enthusiast, Steven Canning, we explored the idea that a proper robot should be able to perform an internal self-diagnostic routine to verify its operational status. My first thoughts were somewhat negative. Let’s look at an example to help clarify my initial concerns.

Assume your robot has a compass to determine its orientation and wheel encoders to measure the movement of its wheels. You could attempt to check the compass by using the wheel encoders to turn the robot 90° and then verify that the compass readings reflect that movement. If the readings are wrong though, the error might not be the compass’ fault because it is possible that the robot is not making the 90° turn due to problems with its wheel encoding sensors (or even with the motors themselves). Of course, you could personally move the robot 90°, but that defeats the purpose of having the robot perform a self-test.

In general, in order to test any of its sensors, a robot has to perform some action that alters the sensor’s state in a predicted way. When the test fails, the question becomes, how do we know if the sensor is failing, or if the robot is not performing the action (or at least not performing it accurately)?

This could be an unsurmountable problem — unless you know, or can verify, that one of your robot’s sensors is, in fact, working correctly. If we knew in the previous example, for instance, that the wheel encoders produced accurate movements, then it would be easy to test the compass. If a robot has numerous sensors, this technique can be even more effective. You could use one working sensor to verify that a second works — then use either or both of those sensors to check the operation of a third sensor. At that point, you would have three sensors to check the operation of a fourth. The more sensors your robot has, the more options you have when creating a diagnostic routine.

Using a Robot Simulator

The RobotBASIC simulated robot is a great way to experiment with this idea for several reasons. First, the simulation has numerous sensors including wheel encoders, Since RobotBASIC is free for everyone (download it from www.RobotBASIC.org), it provides an easy-to-use common platform for testing a variety of ideas before implementing them on your own robot using your own programming tools and sensors. Finally, if your robot is built using the RobotBASIC Robot Operating System (RROS), then programs developed on the simulator can be used to control your real robot with minor modifications. Even if your robot is based on an entirely different platform, RobotBASIC’s simple BASIC syntax (which is very much like its own pseudo code) makes it easy to translate algorithms to your language of choice.

The goal of this article is to present a program that can test many of the simulator sensors listed above. Ideally, the program will demonstrate principles that will inspire others to create similar functionality for their own robots. To that end, let’s examine what is needed to write such a program.

As mentioned earlier, it can be difficult — if not impossible — to determine if various sensors are working properly unless we have at least one sensor (or sensor subsystem) that we know is working correctly. After considering ways of testing each of the simulator’s sensors, it occurred to me that the line sensors might be testable without using any of the other sensors, as long as it can be assumed the robot can perform basic movements (even if such movements are not perfectly accurate). If this could be done, the line sensors can be used to test other sensors as described earlier. Let’s see how the line sensors could be autonomously tested.

The Testing Environment

We will assume that the robot to be tested is placed in a known test environment. One of the properties of this environment is that it will have a line on the floor that is just wide enough to trigger all three of the line sensors. It will also be assumed that at the start of the test, the robot will be placed on this line with all three line sensors

Have you ever packed up your robot and taken it to a club meeting with the intent of demonstrating all your work to a room full of hobbyists only to find that a loose connection made both you and your robot look, not just unprepared, but totally inept? If your robot had an internal self-testing program, it could have saved the day for you. Read on to discover some concepts to help you develop such a program for your robot.