2.2 Sensor Technologies
When considering sensor technologies for the robot, the operation environment must be considered. The robot is intended to operate indoors, on conventional carpeted floors. The obstacles encountered will primarily be either furniture, features of architecture, and people. The first two types can be assumed to be stationary, whereas people are likely to move around, and may provide fluctuating sensor returns. It is expected that the Vision System will be able to differentiate between people and non-living objects.
2.2.1 Infra-Red IR proximity ranging has the disadvantage of only realistically providing detect/non-detect information, since the reflectivity of objects to IR is much more variable in an indoors environment [Schur, 2006] . The components, however, are widely available and compact. IR sensors use reflected IR light to detect surfaces. Low frequency modulation of the emitted beam is usually used to eliminate interference from unvarying sources, such as electric lights or the sun. Distance measurements are only possible if the environment has uniform colour an surface structure, and the sensors must be calibrated
3
Robotic Sensors & Control
to do this. This is rarely practical in most scenarios, however. Black or dark surfaces, for instance, are practically invisible to IR sensors, so they are not completely infallible when it comes to proximity detection. It is because of this that IR sensors are generally only effective for object detection, and not distance measuring. Furthermore, since the intensity of IR light decreases quadratically with distance (proportional to d-2), typical maximum ranges are 50 to 100 cm, which may prove too small for the purposes of the project. [Nehmzow, 2000]
2.2.2 RADAR RADAR
provides an accurate picture of the surroundings, and is a well understood technology. The majority of objects in the indoor environment have high radar reflectivity, however there may be significant potential for interference from other radio sources in the CVSSP, due to the Wireless Networking systems in place. It is also uncertain how much power would be needed to operate a RADAR antenna with sufficient power to work effectively over the distances involved.
2.2.3 Inductive, Magnetic, Capacitive In the field of proximity sensing, Inductive sensors may be used to detect the proximity of ferromagnetic materials. However, this method is unsuitable for use in the specified environment, as ferromagnetic materials are unlikely to be encountered in great quantities, making this technology more suitable for industrial and manufacturing robots. In addition to this, the sensor requires motion of a ferromagnetic object to generate an output voltage; stationary objects relative to the sensor have no effect. The inductive proximity sensor also has an extremely short range, typically fractions of a millimetre. This range limitation is another reason why this technology is mainly confined to assembly-line robots.[Fu, 1987] A technology with a potentially greater detection range is the Hall-effect sensor. This device works on the principle of a Hall-effect sensor located between the poles of a permanent magnet. When a ferromagnetic surface/object is brought close to the magnetic poles, the magnetic field across the sensor would be reduced (see Figure 2.1).
4
Robotic Sensors & Control -
This method has similar disadvantages to the Inductive sensor method described above; only ferromagnetic materials can be detected, and the range of detection is reduced. [Fu, 1987] If a sensor is required to detect proximity to a non-ferromagnetic surface, a capacitive sensor may be used. These sensors are capable (with varying degrees of sensitivity) of reacting to all non-gaseous materials. As the name implies, the sensor works by detecting a change in capacitance between two electrodes, effectively using the sensed object (and the air around it) as part of the capacitors dielectric. [Fu, 1987] Capacitance based sensors are once again subject to a limited range. Also, whilst non-ferrous materials will give rise to a response, the level will be markedly less than that of a ferrous material; for example, Iron can cause a response 2.5 times greater than that caused by PVC at the same distance (see [Fu, 1987] pp.281).
2.2.4 Sonar A great deal of work has been done on (ultrasound) sonar sensing in the field of Robotics. In a typical ultrasound sensor system, a 'chirp' of ultrasound is emitted periodically from a reasonably narrow-beam acoustic transducer. This burst of ultrasound will be reflected from nearby surfaces and can be detected at the sensor after a time T. This time interval is the out-and-back time. Since the speed of sound in air is known, it is a simple matter to calculate the distance to the reflecting surface using the relationship between velocity and time. A major advantage of ultrasound sensing methods is that the dependency of the sensor response upon the material being sensed is reduced, when compared to methods such as Opto-sensing and RADAR. This is clearly of benefit in an indoor environment, where a variety of obstacles will be found having different surface
5
Figure 2.1: Operation of a Hall-effect sensor in conjunction with a permanent magnet ©[Fu, 1987] pp279
Robotic Sensors & Control -
compositions may be encountered. A contrasting disadvantage is that the sensor field is in the shape of a cone; the detected object could be anywhere within the sensor cone at the measured distance. The accuracy of the position measurement is dependent on the width of the sensor beam. Also, a phenomenon called Specular Reflections can cause inaccuracies in the measurements. If an ultrasound beam strikes a smooth surface at a sufficiently shallow angle, the beam will be reflected away from the receiver instead of back towards it. This may cause a larger range than actually exists to be read by the sensor.
There are methods that have been developed to combat Specular Reflections. One method uses so called “Regions of constant depth”. If a 360 sonar scan is performed (for example) a significant section of arc where the ranges measured are constant is termed a Region of constant depth (RCD, see Figure 2.2). These regions can be interpreted by taking two (or more) sensor scans from two differing locations and comparing the arcs of the RCD's. If the arcs intersect, a corner is indicated at the point of intersection. If the arcs are caused by a flat wall, they will be at a tangent to the reflecting plane (see Figure 2.3). [Nehmzow, 2000] A third issue to be overcome relates to arrays of ultrasound sensors. If one sensor detects the reflected pulse from another, so-called crosstalk arises. Solutions to this include coding the sensor signals somehow, or controlling the timing of the sensors to prevent erroneous detections. [Nehmzow, 2000] Ultrasound sensors are effective at much greater distances than the proximity sensing methods mentioned above, even taking into account the increased atmospheric attenuation of sound waves at high frequencies. This means that the robot would have more freedom of movement, and would be able to sense obstacles at a greater range, allowing more time for path-planning computations to be performed. An experiment performed by Mitsubishi Electric Corporation showed that a mechanically scanned ultrasound sensor was able to detect the locations of standing persons within a room ([Pugh, 1986] pp.271). Investigations were also made into the practicality of an electronic scanning system. The advantage of the electronic scanning system over the mechanical system is that the servos used to pan and tilt the sensor beam contribute vibrational noise, and the assembly is by necessity quite large. An electronic scanning system can be used to deflect the beam by unifying the phases of the emitter elements in the desired direction. The study performed by Mitsubishi highlighted the problems with resolution, reliability, and processing time that must be overcome in the implementation of this form of sensor.
Robotic Sensors & Control -
A fixed sensor will not have the flexibility of the scanning sensor, but will be simpler to mount and utilise. Multiple sensors are needed to provide all around coverage.
2.2.5 Laser Range Finders These sensors are also referred to as Laser Radar or 'Lidar'. They are common in robotics, and function in the same manner as the sonar sensors detailed above; instead of emitting a pulse of ultrasound, a pulse of near- infrared light is emitted. The out-and-back time is again used to determine the range to the detected object. However, since the speed of light is much faster than the speed of sound through air at room temperature (order of 106 higher), the means of measuring the out-and-back time must be proportionately more accurate.
Since the wavelength is also much shorter, the probability of total reflection off a smooth surface is reduced, so specular reflections are less of an issue. Accuracy of commercial Laser sensors is typically in the millimetre range. [Nehmzow, 2000]
2.2.6 Shaft Encoders In order to determine the robot's position, some form of odometry is useful. Sensors known as shaft encoders are used to measure the rotations of robot's wheels. If the circumference of the wheels is known, the distance travelled (and possibly the direction) can be determined.
For measuring distance travelled, Incremental encoders are most suitable. The alternative, Absolute encoders, are more suitable for measuring the current position of a rotating shaft. Incremental encoders are suited for summing movement over time. In a typical set up, two photoreceptors are used to read a disc affixed to the shaft. The disc is encoded with two tracks of information, one for each receptor, in such a way that one will always lag in the case of clockwise rotation, and always lead in the case of anti-clockwise rotation (for example). The number of times each receptor is triggered will inform the number of revolutions achieved.
Using shaft encoders to provide odometry, and in turn an estimate of position, is known as dead reckoning. It has been observed in practice that dead reckoning is very unreliable over any significant distance. This is due to motions of the shaft that are not due to locomotive rotation, such as skidding or slipping on a surface. Such issues would be of particular concern in a skid steer system. [Nehmzow, 2000]
When conducting preliminary research for this project, it was noted that Optical (IR and Laser) and Acoustic sensors (ultrasound) are common products available for amateur Roboticists. This may be taken as a reasonable indication of their ease of use and manufacture, and of their suitability for indoor robotic sensing applications.
2.3 Control of Motors The robot uses DC electric motors, and in order to control and drive them, a system incorporating a power converter/regulator is needed. The power from the chassis battery must be translated to the 7.2V needed by the DC motors, and regulated in such a way as to provide speed, acceleration, and directional control to the robot. A range of off-the-shelf controllers are available from Devantech Ltd [Devan, 2006] , which perform exactly the task outlined in the above paragraph. These controllers are highly modular, requiring only power and control
7
Robotic Sensors & Control -
inputs, and can be controlled using a variety of methods, including analogue voltage inputs, and Radio Control Model systems. Of particular note is the I2C capability built into many of these products, considering the availability of USB-I2C interface devices from the same manufacturer, although analogue signals can also be used to control the speed/direction. It was possible that a custom circuit could have been designed, incorporating the power regulation and communications capabilities desired. This, however, would have been a significant design undertaking, requiring significant time and effort, to allow for development of a working device. This option was considered infeasible within the time constraints of this project.
2.4 Control Interface The Motor controllers used in the above example robot are acting as slaves on an I2C bus. This communications standard was developed by Philips as a means of communicating between Integrated Circuits, using a minimum number of data lines. A range of sensors are available from the company supplying the motor controller and the chassis that also act as I2C slaves. Elsewhere in the product range, there are sensors based on the same principle, but are triggered by logic levels, with no bus communication functionality. It was felt that the logic-triggered sensors should be combined with a processing interface, as this would allow for more flexibility, and would foster greater understanding of the technology. The laptop computer has several USB ports available (due to an installed USB expansion card), and a USB to I2C translation device is available from Devantech Ltd [Devan, 2006] . This makes the I2C bus a viable choice for a sensor/motor controller interface, as the aforementioned translation device is treated as a serial (COM) port by Windows (through the use of the freely available drivers), and the writing of Windows programs to access serial ports is a simple task. Another option for the connection of sensors to the laptop computer is an RS232 interface to one of the serial communications ports. This would have required more time to implement, however, and in order to achieve a working solution quickly, the USB-I2C interface was deemed to be the best choice. In addition to this, RS232 is an older technology, which may not be supported on future laptops. Therefore, in order to include an element of 'future-proofing' to the system, USB is a better choice. If the bus-enabled sensors were chosen, they would have been connected to the laptop and controlled directly via the Hardware Interface Software. If the logic-triggered sensors were chosen, then some intermediate device needed to be in place to govern communications between the sensors and the laptop PC. There also needed to be some provision for the possibility of using other types of sensors, and sensors from other manufacturers. The I2C enabled sensors appear to be unique to Devantech Ltd [Devan, 2006] , and this should be considered with regard to the long term maintainability of the system. Such an intermediate device needs to either have inbuilt I2C functionality, or be sufficiently customisable that an I2C interface can be implemented. Devantech Ltd offer a range of I2C to IO devices that can do this job very simply. A more sophisticated solution is to use a PIC (Programmable Interface Controller), manufactured by Microchip Inc. These chips come with a wide range of features (including in-built I2C functionality) and are a
8
Robotic Sensors & Control
very popular and widely understood product range. Many of Devantech Ltd's products are based around PIC micro controllers, which suggests that the PIC product family is trusted and well-supported by the robotics community. In addition to this, facilities for programming PICs are available in the Undergraduate Labs. Such facilities include MPLABtm [Micro, 2000] software, which allows programs to be composed in assembly language, and, with the installed C18 compiler, in C. Since the author is familiar with C from a level 1 programming course, this does not require learning a new language. The MPLAB software includes sophisticated debugging tools, allowing code execution to be 'stepped through', whilst displaying the values of any program variables. Debugging can be done in hardware, if an In-Circuit Debugger tool is connected to the correct pins on the PIC. The undergraduate lab has a number of these tools, as well as PICStart Plus programmers, which are simply used to program PICs before they are installed into a circuit. Also useful are development boards which provide a variety of tools for testing programs and concepts (switches, keypads, displays, etcetera).
2.5 Technologies Selected Based on the available products, and the literature researched (see References), Ultrasound Sensors were selected for collision avoidance and obstacle detection. The model chosen was the Devantech Ltd SRF05. This Sensor is simple to use, and has a range between 4m and 1cm, suitable for the distances encountered in the CVSSP. The Motors is controlled by an MD22 Motor Controller. As mentioned above, this was shown to work well with the chassis and motors in a demonstration video. In addition, buying as many components as possible from the same manufacturer was intended keep postage costs down. A PIC18F2220 micro controller is used as a sensor controller, and communicates with the Laptop using the built-in I2C module, through a USB to I2C interface (USB-I2C). It should be noted that the I2C module is part of a configurable serial port system on the PIC, allowing the use of other serial protocols in the future. The configurable logic outputs are used to trigger the SRF05 sensors (see below), and the internal timers are used to measure the length of the return pulse. The sensors used in the design are triggered by logic signals applied to their control pins. A range of sensors are also available with I2C bus functionality, allowing them to act as slave devices and respond to commands in the same fashion as the MD22 motor controller. The simpler sensors were selected to keep the Hardware Interface Software as simple as possible, and in order to provide a wider range of learning opportunities, such as PIC programming. It was reasoned that should the PIC based solution prove unworkable, the I2C-ready sensors could be used with a minimal number of changes to the Hardware Interface Software. In addition, a possible future project could be based on constructing Ultrasound rangers from scratch, based on the commercially available model. If this was the case, the rangers would have an interface already in place, although this was not necessary.
9
Robotic Sensors & Control -
The sensors are mounted at positions on the chassis of the robot, at equal angular spacing (see Figure 2.4. This will allow a model of the robots surroundings to be produced, by frequently polling each sensor. In this way, a constantly updating navigational map can be constructed. The eight sensors, mounted as shown in Figure 2.4, should allow for maximum information about the environment to be collected, as well as the possibility of using Regions of Constant Depth (see 2.2.4). It was hoped to add a magnetic sensor to the robot to allow orientation to be determined. However due to time constraints this was not implemented, and there was no significant research done on this type of sensor.
When considering sensor technologies for the robot, the operation environment must be considered. The robot is intended to operate indoors, on conventional carpeted floors. The obstacles encountered will primarily be either furniture, features of architecture, and people. The first two types can be assumed to be stationary, whereas people are likely to move around, and may provide fluctuating sensor returns. It is expected that the Vision System will be able to differentiate between people and non-living objects.
2.2.1 Infra-Red IR proximity ranging has the disadvantage of only realistically providing detect/non-detect information, since the reflectivity of objects to IR is much more variable in an indoors environment [Schur, 2006] . The components, however, are widely available and compact. IR sensors use reflected IR light to detect surfaces. Low frequency modulation of the emitted beam is usually used to eliminate interference from unvarying sources, such as electric lights or the sun. Distance measurements are only possible if the environment has uniform colour an surface structure, and the sensors must be calibrated
3
Robotic Sensors & Control
to do this. This is rarely practical in most scenarios, however. Black or dark surfaces, for instance, are practically invisible to IR sensors, so they are not completely infallible when it comes to proximity detection. It is because of this that IR sensors are generally only effective for object detection, and not distance measuring. Furthermore, since the intensity of IR light decreases quadratically with distance (proportional to d-2), typical maximum ranges are 50 to 100 cm, which may prove too small for the purposes of the project. [Nehmzow, 2000]
2.2.2 RADAR RADAR
provides an accurate picture of the surroundings, and is a well understood technology. The majority of objects in the indoor environment have high radar reflectivity, however there may be significant potential for interference from other radio sources in the CVSSP, due to the Wireless Networking systems in place. It is also uncertain how much power would be needed to operate a RADAR antenna with sufficient power to work effectively over the distances involved.
2.2.3 Inductive, Magnetic, Capacitive In the field of proximity sensing, Inductive sensors may be used to detect the proximity of ferromagnetic materials. However, this method is unsuitable for use in the specified environment, as ferromagnetic materials are unlikely to be encountered in great quantities, making this technology more suitable for industrial and manufacturing robots. In addition to this, the sensor requires motion of a ferromagnetic object to generate an output voltage; stationary objects relative to the sensor have no effect. The inductive proximity sensor also has an extremely short range, typically fractions of a millimetre. This range limitation is another reason why this technology is mainly confined to assembly-line robots.[Fu, 1987] A technology with a potentially greater detection range is the Hall-effect sensor. This device works on the principle of a Hall-effect sensor located between the poles of a permanent magnet. When a ferromagnetic surface/object is brought close to the magnetic poles, the magnetic field across the sensor would be reduced (see Figure 2.1).
4
Robotic Sensors & Control -
This method has similar disadvantages to the Inductive sensor method described above; only ferromagnetic materials can be detected, and the range of detection is reduced. [Fu, 1987] If a sensor is required to detect proximity to a non-ferromagnetic surface, a capacitive sensor may be used. These sensors are capable (with varying degrees of sensitivity) of reacting to all non-gaseous materials. As the name implies, the sensor works by detecting a change in capacitance between two electrodes, effectively using the sensed object (and the air around it) as part of the capacitors dielectric. [Fu, 1987] Capacitance based sensors are once again subject to a limited range. Also, whilst non-ferrous materials will give rise to a response, the level will be markedly less than that of a ferrous material; for example, Iron can cause a response 2.5 times greater than that caused by PVC at the same distance (see [Fu, 1987] pp.281).
2.2.4 Sonar A great deal of work has been done on (ultrasound) sonar sensing in the field of Robotics. In a typical ultrasound sensor system, a 'chirp' of ultrasound is emitted periodically from a reasonably narrow-beam acoustic transducer. This burst of ultrasound will be reflected from nearby surfaces and can be detected at the sensor after a time T. This time interval is the out-and-back time. Since the speed of sound in air is known, it is a simple matter to calculate the distance to the reflecting surface using the relationship between velocity and time. A major advantage of ultrasound sensing methods is that the dependency of the sensor response upon the material being sensed is reduced, when compared to methods such as Opto-sensing and RADAR. This is clearly of benefit in an indoor environment, where a variety of obstacles will be found having different surface
5
Figure 2.1: Operation of a Hall-effect sensor in conjunction with a permanent magnet ©[Fu, 1987] pp279
Robotic Sensors & Control -
compositions may be encountered. A contrasting disadvantage is that the sensor field is in the shape of a cone; the detected object could be anywhere within the sensor cone at the measured distance. The accuracy of the position measurement is dependent on the width of the sensor beam. Also, a phenomenon called Specular Reflections can cause inaccuracies in the measurements. If an ultrasound beam strikes a smooth surface at a sufficiently shallow angle, the beam will be reflected away from the receiver instead of back towards it. This may cause a larger range than actually exists to be read by the sensor.
There are methods that have been developed to combat Specular Reflections. One method uses so called “Regions of constant depth”. If a 360 sonar scan is performed (for example) a significant section of arc where the ranges measured are constant is termed a Region of constant depth (RCD, see Figure 2.2). These regions can be interpreted by taking two (or more) sensor scans from two differing locations and comparing the arcs of the RCD's. If the arcs intersect, a corner is indicated at the point of intersection. If the arcs are caused by a flat wall, they will be at a tangent to the reflecting plane (see Figure 2.3). [Nehmzow, 2000] A third issue to be overcome relates to arrays of ultrasound sensors. If one sensor detects the reflected pulse from another, so-called crosstalk arises. Solutions to this include coding the sensor signals somehow, or controlling the timing of the sensors to prevent erroneous detections. [Nehmzow, 2000] Ultrasound sensors are effective at much greater distances than the proximity sensing methods mentioned above, even taking into account the increased atmospheric attenuation of sound waves at high frequencies. This means that the robot would have more freedom of movement, and would be able to sense obstacles at a greater range, allowing more time for path-planning computations to be performed. An experiment performed by Mitsubishi Electric Corporation showed that a mechanically scanned ultrasound sensor was able to detect the locations of standing persons within a room ([Pugh, 1986] pp.271). Investigations were also made into the practicality of an electronic scanning system. The advantage of the electronic scanning system over the mechanical system is that the servos used to pan and tilt the sensor beam contribute vibrational noise, and the assembly is by necessity quite large. An electronic scanning system can be used to deflect the beam by unifying the phases of the emitter elements in the desired direction. The study performed by Mitsubishi highlighted the problems with resolution, reliability, and processing time that must be overcome in the implementation of this form of sensor.
Robotic Sensors & Control -
A fixed sensor will not have the flexibility of the scanning sensor, but will be simpler to mount and utilise. Multiple sensors are needed to provide all around coverage.
2.2.5 Laser Range Finders These sensors are also referred to as Laser Radar or 'Lidar'. They are common in robotics, and function in the same manner as the sonar sensors detailed above; instead of emitting a pulse of ultrasound, a pulse of near- infrared light is emitted. The out-and-back time is again used to determine the range to the detected object. However, since the speed of light is much faster than the speed of sound through air at room temperature (order of 106 higher), the means of measuring the out-and-back time must be proportionately more accurate.
Since the wavelength is also much shorter, the probability of total reflection off a smooth surface is reduced, so specular reflections are less of an issue. Accuracy of commercial Laser sensors is typically in the millimetre range. [Nehmzow, 2000]
2.2.6 Shaft Encoders In order to determine the robot's position, some form of odometry is useful. Sensors known as shaft encoders are used to measure the rotations of robot's wheels. If the circumference of the wheels is known, the distance travelled (and possibly the direction) can be determined.
For measuring distance travelled, Incremental encoders are most suitable. The alternative, Absolute encoders, are more suitable for measuring the current position of a rotating shaft. Incremental encoders are suited for summing movement over time. In a typical set up, two photoreceptors are used to read a disc affixed to the shaft. The disc is encoded with two tracks of information, one for each receptor, in such a way that one will always lag in the case of clockwise rotation, and always lead in the case of anti-clockwise rotation (for example). The number of times each receptor is triggered will inform the number of revolutions achieved.
Using shaft encoders to provide odometry, and in turn an estimate of position, is known as dead reckoning. It has been observed in practice that dead reckoning is very unreliable over any significant distance. This is due to motions of the shaft that are not due to locomotive rotation, such as skidding or slipping on a surface. Such issues would be of particular concern in a skid steer system. [Nehmzow, 2000]
When conducting preliminary research for this project, it was noted that Optical (IR and Laser) and Acoustic sensors (ultrasound) are common products available for amateur Roboticists. This may be taken as a reasonable indication of their ease of use and manufacture, and of their suitability for indoor robotic sensing applications.
2.3 Control of Motors The robot uses DC electric motors, and in order to control and drive them, a system incorporating a power converter/regulator is needed. The power from the chassis battery must be translated to the 7.2V needed by the DC motors, and regulated in such a way as to provide speed, acceleration, and directional control to the robot. A range of off-the-shelf controllers are available from Devantech Ltd [Devan, 2006] , which perform exactly the task outlined in the above paragraph. These controllers are highly modular, requiring only power and control
7
Robotic Sensors & Control -
inputs, and can be controlled using a variety of methods, including analogue voltage inputs, and Radio Control Model systems. Of particular note is the I2C capability built into many of these products, considering the availability of USB-I2C interface devices from the same manufacturer, although analogue signals can also be used to control the speed/direction. It was possible that a custom circuit could have been designed, incorporating the power regulation and communications capabilities desired. This, however, would have been a significant design undertaking, requiring significant time and effort, to allow for development of a working device. This option was considered infeasible within the time constraints of this project.
2.4 Control Interface The Motor controllers used in the above example robot are acting as slaves on an I2C bus. This communications standard was developed by Philips as a means of communicating between Integrated Circuits, using a minimum number of data lines. A range of sensors are available from the company supplying the motor controller and the chassis that also act as I2C slaves. Elsewhere in the product range, there are sensors based on the same principle, but are triggered by logic levels, with no bus communication functionality. It was felt that the logic-triggered sensors should be combined with a processing interface, as this would allow for more flexibility, and would foster greater understanding of the technology. The laptop computer has several USB ports available (due to an installed USB expansion card), and a USB to I2C translation device is available from Devantech Ltd [Devan, 2006] . This makes the I2C bus a viable choice for a sensor/motor controller interface, as the aforementioned translation device is treated as a serial (COM) port by Windows (through the use of the freely available drivers), and the writing of Windows programs to access serial ports is a simple task. Another option for the connection of sensors to the laptop computer is an RS232 interface to one of the serial communications ports. This would have required more time to implement, however, and in order to achieve a working solution quickly, the USB-I2C interface was deemed to be the best choice. In addition to this, RS232 is an older technology, which may not be supported on future laptops. Therefore, in order to include an element of 'future-proofing' to the system, USB is a better choice. If the bus-enabled sensors were chosen, they would have been connected to the laptop and controlled directly via the Hardware Interface Software. If the logic-triggered sensors were chosen, then some intermediate device needed to be in place to govern communications between the sensors and the laptop PC. There also needed to be some provision for the possibility of using other types of sensors, and sensors from other manufacturers. The I2C enabled sensors appear to be unique to Devantech Ltd [Devan, 2006] , and this should be considered with regard to the long term maintainability of the system. Such an intermediate device needs to either have inbuilt I2C functionality, or be sufficiently customisable that an I2C interface can be implemented. Devantech Ltd offer a range of I2C to IO devices that can do this job very simply. A more sophisticated solution is to use a PIC (Programmable Interface Controller), manufactured by Microchip Inc. These chips come with a wide range of features (including in-built I2C functionality) and are a
8
Robotic Sensors & Control
very popular and widely understood product range. Many of Devantech Ltd's products are based around PIC micro controllers, which suggests that the PIC product family is trusted and well-supported by the robotics community. In addition to this, facilities for programming PICs are available in the Undergraduate Labs. Such facilities include MPLABtm [Micro, 2000] software, which allows programs to be composed in assembly language, and, with the installed C18 compiler, in C. Since the author is familiar with C from a level 1 programming course, this does not require learning a new language. The MPLAB software includes sophisticated debugging tools, allowing code execution to be 'stepped through', whilst displaying the values of any program variables. Debugging can be done in hardware, if an In-Circuit Debugger tool is connected to the correct pins on the PIC. The undergraduate lab has a number of these tools, as well as PICStart Plus programmers, which are simply used to program PICs before they are installed into a circuit. Also useful are development boards which provide a variety of tools for testing programs and concepts (switches, keypads, displays, etcetera).
2.5 Technologies Selected Based on the available products, and the literature researched (see References), Ultrasound Sensors were selected for collision avoidance and obstacle detection. The model chosen was the Devantech Ltd SRF05. This Sensor is simple to use, and has a range between 4m and 1cm, suitable for the distances encountered in the CVSSP. The Motors is controlled by an MD22 Motor Controller. As mentioned above, this was shown to work well with the chassis and motors in a demonstration video. In addition, buying as many components as possible from the same manufacturer was intended keep postage costs down. A PIC18F2220 micro controller is used as a sensor controller, and communicates with the Laptop using the built-in I2C module, through a USB to I2C interface (USB-I2C). It should be noted that the I2C module is part of a configurable serial port system on the PIC, allowing the use of other serial protocols in the future. The configurable logic outputs are used to trigger the SRF05 sensors (see below), and the internal timers are used to measure the length of the return pulse. The sensors used in the design are triggered by logic signals applied to their control pins. A range of sensors are also available with I2C bus functionality, allowing them to act as slave devices and respond to commands in the same fashion as the MD22 motor controller. The simpler sensors were selected to keep the Hardware Interface Software as simple as possible, and in order to provide a wider range of learning opportunities, such as PIC programming. It was reasoned that should the PIC based solution prove unworkable, the I2C-ready sensors could be used with a minimal number of changes to the Hardware Interface Software. In addition, a possible future project could be based on constructing Ultrasound rangers from scratch, based on the commercially available model. If this was the case, the rangers would have an interface already in place, although this was not necessary.
9
Robotic Sensors & Control -
The sensors are mounted at positions on the chassis of the robot, at equal angular spacing (see Figure 2.4. This will allow a model of the robots surroundings to be produced, by frequently polling each sensor. In this way, a constantly updating navigational map can be constructed. The eight sensors, mounted as shown in Figure 2.4, should allow for maximum information about the environment to be collected, as well as the possibility of using Regions of Constant Depth (see 2.2.4). It was hoped to add a magnetic sensor to the robot to allow orientation to be determined. However due to time constraints this was not implemented, and there was no significant research done on this type of sensor.
No comments:
Post a Comment