Local Sensing using Sonar Object Identification and Infrared Drop-off
Detection on an Autonomously Mobile Chassis.

 

Ethan Bregman

B.S. Electrical Engineering (ES 02)

Presented to the Department, April 30, 2002

 

 

This report focuses on a senior thesis research project into sonar array fabrication and infrared terrain detection for an autonomous chassis. Specifically it identifies the strategies used to create a dense sonar array from only a single 6500 series ranging modules. The chassis itself part of a group distributed systems project involving EE, CS and ME seniors at Yale University.

0.0 Report Overview

 

Section 1 Introductory Information

1.1) Introduction

1.2) Whitney Overview

 

Section 2 Project Goals

2.1) Local Sensing

2.2) Sonar Array System

2.3) Infrared System

 

Section 3 Sonar Array Development

3.1) Sonar Development process

3.2) Mounting Bracket

3.3) Sonar Control Circuitry

 

Section 4 Infrared Terrain Detection Development

4.1) Infrared Development Process

4.2) Infrared Mounting

 

Section 5 Results

5.1) Sonar Array Results

5.2) Infrared Terrain Results

 

Section 6 Conclusion

 

Section 7 Appendix

 

1.1 Introduction

 

Whitney is an autonomous chassis that began development under the guidance of Prof. Roman Kuc in the fall semester of 2001 at Yale University. Being a distributed processing and sensing chassis for multiple experiments, Whitney is quite complex. This report focuses on the local area sensing that Whitney uses to determine its immediate surroundings.

 

1.2 Whitney overview

 

A Geode EBX class embedded processor conducts the main processing of Whitney. This system is compatible with the PC104 form-factor embedded systems, but is slightly larger. The Geode was selected over a standard PC104 because of the significant reduction in cost for a small sacrifice in footprint.

Sensing, processing and control functions are distributed from the Geode to several (3) PIC 16F877 embedded microprocessors, over independent RS-232 communication channels. Whitney s global sensing includes a GPS satellite receiver linked directly to the Geode, a digital compass, and an optical odometry system. Additionally, Whitney carries a modulated infrared beacon to guide an autonomous scooter that follows it, and a radio frequency modem to broadcast its status and readings to an off-board computer.

Since Whitney is a group development project, and a project that will continue into the future past the current group s graduation, it is vitally important that all elements be designed with modularity and reconfigurability as primary elements.

 

2.1 Local Sensing

 

Local sensing for Whitney revolves around two independent systems. It is necessary for the safety of Whitney and those around it that it is able to identify potential obstacles and imminent collisions and react accordingly. For the purpose of obstacle and collision avoidance Whitney uses a dense forward-looking sonar array. It is also necessary that Whitney protect itself from dangerous variations in terrain, for example, a pothole that is to deep, or a curb that is to tall. For surface detection Whitney uses a pair downward looking infrared sensors.

 

2.2 Sonar Array Concept

 

The initial concept for the Whitney sonar array is to have a dense forward-looking sonar array of 8-16 transducers. These are mounted near the front of the chassis. The immediate processing of the sonar signals is handled by a dedicated microprocessor that is responsible detecting objects and potential collisions, and will recommend course adjustments to the Geode brain. In the event of an unavoidable collision, it will signal the motors to be shut down directly.

 

2.3 Infrared Concept

 

The initial terrain detection concept was to have a pair of downward looking laser emitter/detector pairs. Due to integration problems, (and prohibitive cost) this solution was scrapped and the concept change to infrared. The infrared system concept is a pair of downward looking infrared emitters and detectors that will determine if Whitney is about to reach a curb or insurmountable difficulty in road surface. The emitter/detector pairs are be mounted as far forward on the chassis as possible to give the system enough time to correct or come to a halt as necessary.

 

3.1 Sonar Developmental Process

 

Traditional Yale sonar arrays, as realized by the Yale Intelligent Sensors Laboratory, generally use stereoscopic sonar detection to identify obstacles, (or in the case of Rodolph, identify which side of a penny is facing up). However when small arrays (2-4 transducers) are applied to mobile chassis, they tend to leave large blind spots. By creating a dense array (8-16 transducers) Whitney will be able to better protect itself from obstacles of all sizes and textures. The selected transducers are Polaroid Instrument Grade; the data sheet is available in appendix D. They are round transducers, approximately 1.7 inches in outside diameter.

 

3.2 Mounting Bracket

 

For the sonar array to function properly, the transducers need to be mounted to the Whitney chassis securely. There are several factors that affect the bracket design, which are detailed here.

The front of the existing Whiney chassis has hard points where the front freewheels attach. These hard points are ideal for mounting additional hardware, and serve as the attachment point for the sonar array bracket. These points are bolt pairs, described in the picture from a frontal elevation.

The exterior housing for the transducers acts as their ground. Although the housings are rubber coated everywhere, except the blade connector, to prevent shorting if they are mounted in metal, Prof. Kuc indicated that in a mobile chassis, vibration and chaffing might cause the rubber coating to fail. Therefore the transducer mounts must be non-conductive.

The transducers have a beam angle that is most effective within 20 degrees. In order to have an acceptable range of vision (i.e. >100 degrees), the transducers cannot be mounted in a forward-looking plane, which would attenuate the scope of their vision. They must have an appropriate vergance angle between each transducer such that the sonar field has an acceptable range.

 

Taking these factors into account, the preliminary design for the sonar mounting-bracket was a sheet of 1/4inch plexi-glass that would be heat rolled to a curve that would give the proper vergance to the transducers, and then drilled to accept the transducers. After consideration this design was scrapped due to the limitations of heat forming hardware at this facility.

 

As the design process began anew, it was important to revisit the design criteria, specifically, modularity which was not addressed in the first concept. Mounting each transducer in a small, enclosed housing that can be positioned on the array bracket greatly increased modularity. For cost and fabrication reasons, the transducers were mounted in 2 x 3 in plastic project boxes.

In order to mount these boxes appropriately, the bracket needed to be re-thought. Several solutions immediately presented themselves:

a)      A piece of 1/4inch thick by 2inch high aluminum that could be rolled to approximated the shape originally intended for the plexi-glass. While heat forming is challenging, the Yale Gibbs facility has a sturdy enough strap roller to realize the piece in aluminum.

b)      A section of metal tubing, to which the transducer boxes could be attached with hose-clamps

c)      Two smaller wires (i.e. 1/4inch O.D.) running from each of the mounting bolts that the transducer boxes could slide along.

The simplest option appeared to be the rolled strap aluminum, but the final piece would have been heavy enough to offset the balance of the chassis, and was vetoed. The tube did not give enough of a vergance angle to cover the necessary range. The two small wires were not sturdy enough to absorb the unfortunate possibility of a malfunction and a frontal impact.

 

The final implemented design is a straight 1/2inch square strap of 6000 series aluminum. It is drilled every 1.5inches with 10-32 holes. The mounting bracket is attached to the hard points via 1/4inch by 3inch angle aluminum. The angle aluminum s front face is cut to only 1/2inch to reduce weight and keep the hard points accessible for future attachments or additional sensors. The sonar boxes are attached to the bracket via individual L-brackets machined from 1/16 by 3/4inch angle aluminum. Each bracket is double bolted to the back of the sonar boxes, and attached to the main bracket by a single 10-32 thumbscrew, as shown in the pictures. To incorporate the mounting, the track width of the front wheels has been widened 1/4inch on each side.

3.3              Sonar Control Circuitry

 

The pulse and echo detection of the transducers is handled by a Polaroid 6500 series ranging module. From the factory, these units are designed to mate one ranging module per transducer. By using an intelligent cycling technique, and novel circuitry, Whitney is able to run an 8-transducer array off a single ranging module.

The inputs and outputs of the 6500 are quite basic; it takes a digital initialization signal and returns a single digital pulse when it detects an echo. Therefore, some small level of processing (and timing) is necessary to interpret the environment, based on when the echo signal is returned. This basic processing is handled by a separate PIC16F877 microcontroller.

 

Instead of attaching a 6500 directly to each transducer, Whitney has digital selection circuitry that allows the same microcontroller that interprets the data to specify which transducer it would like to listen to.

The first iteration of this circuitry switched the control line from one transducer to another by means of an array of power FETs. The 6500 module consumes relatively little power when in standby mode (~100mA). However, for the small amount of time the sonar pulse is sent out, the power consumption jumps momentarily to 2 Amps, therefore the power FETs proved unacceptable because of the massive inline capacitance over the 2 Amp and ~400V spike run through the system on each sonar chirp.

The immediate solution, and iteration two, was to tie all of the signal lines to the transducers together, and a select which transducer would fire by moving the power FETs to the ground side of the signal wire. While not completely unacceptable, this means there are 8-16 exposed 18awg signal lines vacillating back and forth between ground and +400V, generally not the best solution for a mobile chassis.

The third and final iteration of the selection circuitry is a relay-block based system. Each relay block has two outputs and three inputs. The outputs are the transducer control line, which is the signal from the Polaroid controller board, and the controller board ground. The inputs are those same two signals plus a selector line that activates the relay.

 

The switching mechanism itself is an electromagnectically controlled relay switch (RS275-243) that takes its control through a FET (2N7000). Each relay also has an LED signal light to specify its status. The diode (1N4148) prevents any back impulses from damaging the FET control. Under normal operation the relay sits in the normally closed position shown in the diagram, which connects the signal line to a floating No Connection pin.

 

4.1 Infrared Developmental Process

 

Since the infrared system is relatively simple compared to the sonar array, fabrication began only after many other significant systems were in place, so that the infrared system would not hinder the operation of other systems.

 

The basic principle of the terrain detection system is to assure Whitney does not high-side or turtle itself over changing surfaces. High-siding is an off-roading term for when the undercarriage of a vehicle comes in contact with the ground enough to lift the drive wheels off the ground, or reduce their traction enough to strand the vehicle. Turtleing refers to completely flipping a vehicle such that all it can do is spin its drive wheels helplessly in the air.

The initial implementation of the terrain detection used a pair of laser based emitter/detector units to see potential hazards. This system was quickly discarded, however, due to the excessive cost and power consumption of laser emitters.

The second, and current, iteration of the terrain detectors uses sealed infrared emitter and detector pairs, SunX MR3-00A s, which are used in industry primarily for package counting at shipping firms. The Yale Intelligent Sensors Laboratory often uses these devices with great success, because of their resistance to washout in daylight, and general rugged design.

 

4.2 Infrared Mounting

In order to uniformly protect Whitney, the mounting angle is a prime consideration. Although the IR detectors clearly need to look forward to prevent Whitney from trundling down a staircase, they also need to be aware of the periphery. The detectors need to be placed as far to the outer frontal limits as possible to protect Whitney when it is traveling parallel to a drop-off or turning on axis.

 

The mountings for the IR detectors are 1/8inch aluminum plate that is machined to attach to the ends of the existing sonar bracket. They extend forward 5.5inches, enough to have clear line of sigh over the physical contact bumper system. The plates are secured down using the same 10-32 thumbscrews used to mount the sonar boxes. Since the plate is relatively thin, the sonar boxes can be mounted on the same tap holes used to attached the IR plate.

 

5.1 Sonar Array Results

 

The final array bracket proves that modularity, reconfigurability and simplicity can all go hand in hand. Using only three parts, the bracket can hold up to 18 of the sealed transducer boxes (using both the top and bottom of the bracket), which can be set at any vergance angle. The array could be as dense as 22 transducers with specific vergance angles.

 

The transducer selection circuitry worked quite well; there was only one significant surprise. The relays would close immediately and appropriately when the +5V high control signal from the PIC was applied, but when the low signal resumed, the relays would occasionally flutter, or stay on. The root of the problem was determined to the be that the low signal floated significantly, causing the FETs controlling the relays to float back and forth over their trigger point. The solution was simple, a 100K Ohm resistor to ground just before the FET to pull it down when no high signal is being applied. The schematics have been updated to reflect this change.

 

5.2 Infrared Terrain Results

 

The infrared mounting bracket revealed itself to have an interesting additional use. The original intention of the IR system was to detect hazardous drop-off, but by mounting the system high, in order to have a clear line of sight over the bumper, it has the added benefit of being able to detect obstacles that were too short to identify themselves on the sonar array. Although the bumper system will clearly detect such obstacles as well, the potential to reduce physical stress on the system and create redundancy in obstacle detection and avoidance is a significant bonus.

 

6 Conclusion

As is to be expected from a project of this scope, especially one with an eye on next year s rising seniors, it could be said more was bitten off than could be chewed. In reality however, the team made many exciting breakthroughs with completely untested and new hardware, and have set all the elements in place for an excellent experimental chassis with room to grow, evolve and expand within Yale Engineering and across the Yale community.

 

On a personal note, it was quite rewarding to work with such a talented and diverse group of engineers. Being one who tends to strike out on his own, it was great to see how the whole team came together.

 

Many thanks for their excellent support to:

Prof. Roman Kuc, Ed Jackson, the YSEA, and of course, my Whitney Team members.
Contents of Appendix

 

Appendix A Part Specifications

 

Appendix B Weekly Status Reports

 

Appendix C Circuit Diagrams

 

Appendix D Hardware Blueprints and Sketches

 

 

Appendix A Part Specifications

 

Polaroid Instrument Grade Transducer

 

This instrument grade electrostatic transducer is specifically intended for operation in air at ultrasonic frequencies.  The assembly comes complete with a perforated protective cover.

 

 

 

 

 

 

 

 

Typical Beam Pattern At 50 kHz

Note: db normalized to on-axis response.
Note: Curves are representitive only.  Individual responses may differ.

http://www.polaroid-oem.com/products/ultrasonic.htm

Polaroid 6500 Ranging Module

• Accurate Sonar Ranging from 6 inches to 35 feet
• Drives 50-kHz Electrostatic Transducer with No Additional Interface
• Operates from Single Supply
• Accurate Clock Output Provided for External Use
• Selective Echo Exclusion
• TTL-Compatible
• Multiple Measurement Capability
• Uses TI TL851 and Polaroid 614906 Sonar Ranging Integrated Circuits
• Socketed Digital Chip
• Convenient Terminal Connector
• Variable Gain Control Potentiometer

6500 Board Schematic

Absolute Maximum Ratings

Recommended Operating Conditions

 

MR3 00A Infrared Emitter/Detector

Triangulation Sensor,
Multi Voltage Types.   

AC/DC Version
"The universal sensor". Utilising diffuse area detection the sensor has distance setting for background suppression, can accept a universal voltage supply from 24 V to 240 V and operates virtually independently of colour or surface condition of the detected object.

 

Characteristics.

		MR3-M100P	MR3-M200P	MR3-M100PT	MR3-M200PT
Sensor type		Diffuse
Rated sensing distance		100 cm	200 cm	100 cm	200 cm
Sensing range		20 - 100 cm	50 - 200 cm	20 - 100 cm	50 - 200 cm
Standard detectable object		White drawing paper 20 x 20 cm
Detectable target		transparent and opaque material
Hysteresis		< 10% of measurement range
Detection frequency/ Response time		25 Hz / 20 ms
Output Relay		max. 3 A / 250 V AC (resistive load)
Wavelength of emitting diode		890 nm, Infrared LED
Max. capacity without load		max. 3 W / 4 VA
Housing material		Plastic		max. 4 W / 5 VA
Protection		IP66
Physical size (HxWxL)		68 x 26 x 68 mm
Connection method		Screw terminals
Operating voltage