Autonomous Tank

Introduction

We aim at building an autonomous vehicle, “Homer”, which can stroll around an environment (e.g. our lab, or an environment that we build) without getting stuck at obstacles. That implies that we need a robust algorithm that tells the vehicle how to steer when it gets into different obstacle situations. The goal would be to demonstrate a vehicle that will run by itself and not get stuck or bump into any obstructions.

High Level Description

The following is a high-level description of Homer (our Autonomous tank) :

belts

Separate DC motors

Micro-controller AT90S8515, motor-control circuits (one set for each motor), sensor circuits, batteries, …

Infra-red sensors

GP2D02

A “big picture” of Homer

We used separate speed controllers for the two DC motors which turn the belts separately. Homer turns, or even rotates about an axis, when the motors are run at different speeds. By using the H-bridge speed controller circuit and supplying the circuit with Pulse Width Modulation (PWM) signal as described below, the motors can run at a range of speeds or even backwards. Comparing to the usual car designs which use steering front wheels, this design is preferable because

1. It allows high flexibility in the vehicle movement. It is not constrained by the maximum turning angles of the front wheels.

2. It allows rotation that is useful in getting away from obstacles.

3. It is mechanically easy to built (We think the steering front wheels are hard to build).

We use three infra-red sensors to measure distances from obstacles. The sensor sends out infra-red pulses and then receives back the reflected signal, and afterwards the distance information is output as serial bits. We need to read in these serial bits into the Atmel microcontroller and convert them to a distance by ourselves in our code. The SHARP GP2D02 measures distances in the range 10cm – 80cm and therefore was suitable for our design. SHARP GP2D02 sensor interface with our Atmel microcontroller will be discussed in detail in the Program/HW design section.

We do not have access to the machine shop and hence it is hard to build the physical components of the vehicle by ourselves. Therefore, we purchased a completed toy track vehicle (e.g. tank), and then remove the internal parts and install our own circuitry.

Program/Hardware Design

Sharp GP2D02 Infra-red Range Sensors

Interfacing the Sharp GP2D02 infra-red range sensors were fairly straight forward. Following the timing diagram in the specification sheet from Sharp, I programmed the microcontroller to output a control signal Vin (input signal into the GP2D02) for the infra-red sensor. Since the Vin pin on the sensor is initially in the OFF-state and floats high, the microcontroller outputs the Vin signal to be low to turn on the sensor. The measurement is initiated by forcing the Vin signal to logic low for at least 70ms or until the Vout(output signal from the GP2D02) becomes logic high. Once that occurs, we need to start clocking in the serial bits from the sensor. The microcontroller must toggle the Vin signal at the rate of 0.1ms. We toggle the Vin signal a total of 9 times and we clock in the serial bits on the positive clock edge from Vout. Once the entire byte is read we need to reset the sensor for another reading. Resetting occurs when we float Vin high for at least 1.5ms; after resetting we repeat the process over again for another distance reading.

The serial bits that were read into the microcontroller were converted into a distance of centimeters. We calibrated the sensors using a linearization formula. The sensors output has an inverse relationship to the distance of an object. That is the further the object, the small the serial bits. Since we don't really need the sensors to be too accurate, we only need to know the ranges of the objects from our sensors, we can linearize the curve. We calibrated the sensors individually for two distances(15cm & 30cm). Therefore any distances between 15cm and 30cm, we are confidant of its accuracy (In fact we it seems to be accurate from 10cm to 30cm). Any distance beyond 30cm, we start to see some deviation from linearized curve. We will consider any distance beyond 30cm to be infinite.

NOTE: Linearization Formula was referenced fromhttp://www.barello.net/Papers/GP2D02/multiplexing_gp2d02_sensors.htm

D = Kg/(X-Ko)

Kg is the gain, or overall shape of the curve described by Kg/X

Ko is an offset, shifting the curve up or down.

Let D and X be the distance and output, respectively of your first measurement.

Let D' and X' be the distance and output, respectively, of the second measurement.

Kg = (X'-X) D'D/(D-D')

Ko = (D'X' - DX)/(D' - D)

Since there are three sensors (Left, Center, Right) we have three Vout signals into the microcontroller. All three sensors share the same Vin signal from the microcontroller. Therefore, we take a distance reading for all three sensors at once. It was necessary for us to use a diode between the microcontroller and the Vin signal into the sensor. The microcontroller has 5V logic I/O, while the sensors require 3V logic. According to the specifications, we cannot directly drive the sensors with 5V. Hence we used a diode, since the sensor's OFF-state floats high(3V).

Pulse Width Modulation

We use Pulse Width Modulated (PWM) signals to control the speed of the motors. Timer 1 is set to be in 10-bit PWM mode. To generate PWM signals with different pulse widths, we simply have to change the values stored in OCR1A and OCR1B. The PWM signals are then output via the OC1A and OC1B pins.

In our scheme, when we want the tank to go straight ahead, the outputs from OC1A and OC1B are both steady 5V (not modulated) so that they go with “full speed”, and when we want it to stop, the outputs would be 0V. Cases in between going “full speed” and stopping would results in different PWM signals.

Central Decision Algorithm

Every 0.5 seconds (counted by timer0), the distance information from the sensors are processed in order to understand the instantaneous environment. The sensor readings, which are float values on a continuous range, are compared to thresholds so that the ranges in which they lies are determined:

>= 7 cm && < 20 cm : range 1 (very close)

>= 20 cm && < 30 cm : range 2 (close)

>= 30 cm || <=7cm (infinite, nothing detected)

* Note that the sensors sometimes give values smaller than 7 cm when the distances are actually “infinity”, and that’s why we have the “<=7cm” condition.

Based on the 3 “ranges” determined as above, different control signals are generated. Different combinations of the three ranges form different “cases”, and in each cases, we assign predetermined direction bits (Port A0 and A1) values so that the motors will either turn in the forward or backward direction, and also the right predetermined PWM signals so that the motors will turn with the right speed. We predetermined the control signals in each case by experimentation and “common sense”.

In that sense, the control signals of the vehicle are updated every 0.5 seconds. We want to make sure that it is up-to-date enough to navigate the vehicle.

Power Supply

Since our project required us to demonstrate our vehicle driving around, we need the microcontroller off the development board. Having a 9V battery, we used a power regulator from National semiconductors to supply the microcontroller with 5V. In addition to the power regulator, 2 capacitors(0.2pF and 0.1pF) were also required to build the power regulator circuit for the microcontroller.

We also have 6V powering our motors. The both H-Bridge circuits (LEFT and RIGHT) are supplied with 6V; while the microcontroller with 5V. There is a common ground.

Also, we used a 4Mhz crystal and two 22pF capacitors to drive our microcontroller clock.