You interact with your world through your five senses: sight, sound, smell, taste, and touch. Robots can have all of these senses (okay, I don't know of any bots that can actually taste), and dozens more. A sensor is any device on a robot that receives input from the outside world and passes that input on to the robot's control system. Let's look at a few of the most common types of robot sensors.
A pressure sensor is basically a switch that, when turned on (or off), sends a signal to the control circuitry of a robot to do something (usually back up or otherwise move to a new spot). Probably the most common type is the bump sensor, often a fender or skirt on the robot that, when hit, presses a switch to which the controller responds. Another common type is the feeler or whisker (common on robo-critters). This is simply a wire that, when pressed or bent, engages a switch. Pressure Sensor are probably the most common type of sensors around, and robot builders are constantly experimenting with more effective ways of building them.
There are a number of ways of using light sensors, but the basic idea is to use light waves sent out and received by special photosensitive components to control a robot's actions. In one frequently used navigation scheme, this is accomplished by putting an infrared (IR) transmitter on the bot that sends an invisible light beam out into the environment. That light is reflected off of objects back to a special infrared receiver elsewhere on the bot. The angle of that reflected light changes depending on the proximity of the robot to the object that's reflecting the light. The robot can use this change in angle to measure the distance and trigger an appropriate action (such as an obstacle avoidance sequence).
Grey Walter came up with a unique bump sensor design. The shells of his robo-tortoises hung from a "mast" attached to their bases. When one of the robots bumped into anything, in any direction, the shell would press against the mast, closing a switch, and the bot's control circuitry would trigger an obstacle avoidance routine.
Light sensors are also used in line-following navigation. Commonly, a black or white line is put down to create a "track" for the robot to follow. Photosensitive sensors, on the front or bottom of the bot, straddle the line. If one of the sensors registers a high degree of change in light intensity (by crossing the line), it triggers the motors to adjust course to keep the line between the two sensors (and the robot on track).
Another use of light sensing is to engineer circuits outfitted with light sensors that make the robot "attracted" to light (in other words, it moves in the direction of the most intense light source), or repulsed by the light (it moves away). These "behaviors" are called, respectively, photovoric (photo meaning light, vore being Latin for "swallow up") and photophobic.
There are basically two ways of using reflected light intensity in robot navigation. Inproximity detection, the photosensitive receiver is simply looking for a set trigger value (for example, a high light intensity level), that when reached, will prompt the robot's response. The robot will wait until that trigger value is reached before doing anything. Proximity detection is like an on/off switch. Distance detection, the more sophisticated of the two approaches, uses the changing angle of reflected light to actually measure the changing distance between the robot and obstacles. This allows for the taking of different actions depending on perceived distance.
The most common type of sound sensor makes use of sonar technology. Sonar sensors basically use the speed of sound to measure distance and use that information to aid in robotic navigation. Inaudible sound waves (outside the range of human hearing) are projected out from a transmitter on the robot, bounce off of surfaces, and return to a receiver on the robot. The time it takes for the sound waves to return to the sonar sensor (also called an ultrasonic range sensor) is used to calculate the robot's distance from an approaching object. This data is then passed along to the control circuitry of the robot and appropriate action is taken.
We'll be using light sensors in Project 2 and will discuss them in more detail then.
Another type of sound sensing involves the use of microphones. A robot outfitted with mics can be programmed to move toward a sound source, move a certain body part, or move all of its body parts. Entertainment robots such as Sony's forthcoming SDR-4X humanoid take in sounds via microphones so that the robot can dance (the sound input activates the robot's many servo motors). Robots are also being programmed to respond to certain voice characteristics (such as a raised voice or lowered voice). Of course, microphones are also used on robots that are programmed to respond to sets of human voice commands.
Inexpensive sonar systems for robots became popular when builders began to cannibalize the sonar range finders built into Polaroid's SX-70 instant camera (introduced in the 1970s). The sonar sensors in the camera were designed to measure distance for auto-focusing a shot. The demand for these sensors (and those in subsequent Polaroid cameras) became so great that they are now sold as standalone components.
Cameras have been mounted on robots for decades (going all the way back to the bulky one that gave SRI's Shakey its shake). Cameras are most commonly used on robots that are teleoperated, that is, operated by a human at a distance (either at the end of a tether of control cables or over a radio link). With the advent of the Internet and cheap digital Web cameras, robots can now send video signals through the Net so that humans can see what the robot sees, or even control the robot remotely.
Getting a robot vision system to analyze its video input (moving, real-time, unpredictable input) and make sense of it is still a daunting challenge. However, those in the field of biometrics are making inroads into this challenge. Human facial identification is becoming a more common technology, especially in security systems. Here, face size and proportional information, skin tone, and other factors can be used to match a face with a database of stored faces. Unfortunately, this only works if the person is staring directly into the camera lens.
Robot vision systems are getting better at detecting humans in a "scene," tracking moving objects, making out saturated colors, and roughly understanding three-dimensional objects. They still struggle with telling men from women, interpreting objects if the camera is in motion, determining what materials objects are made of, and determining "gaze direction" (where the person is looking). Honda has made some advances with the latter problem. The current version of its Asimo humanoid can tell where a person is looking if the person both looks at and points to the target object.
Security robots and others designed to detect fire are outfitted with what are calledinfrared pyroelectric sensors. These sensors detect a combination of heat and movement. If they see some nasty, dancing flames in front of them (registering both movement and heat), the sensors are triggered (and the robot will usually call for human backup). These same sensors are used as motion detectors and will detect the heat and movement of a person within a certain range.
Robots can be outfitted with "electric noses" that can sense a variety of toxic gases. Common on security or industrial robots, triggering this type of sensor usually means a call to a human who decides the nature of the threat. Robots might have become fairly sophisticated, but we're still not comfortable letting them handle a ruptured gas main on their own.
Other Types of Sensors
There are dozens more sensor types than what we've already covered. Here are just a few more and a word or two on each:
Encoders— Optical or magnetic sensors that "read" the rotation of a robot's wheels to measure distance traveled.
GPS receiver— A device that can access the Global Positioning Satellite (GPS) system, using the location of several orbiting satellites to pinpoint a robot's location on the ground (outdoor robots only).
Strain gauges— For measuring the physical force exerted on an object. Used in touch sensors, gripper force feedback, and collision detection.
Tilt sensors— Used to indicate the attitude of a robot. Sometimes used for balance, especially in walking robots. If the sensor detects that the robot is on a dangerous angle, and at risk of falling over, it will "ask" the robot's controller to back it away from the threatening terrain.