Close

Requirements and specification

A project log for Seek_r

Seek_r is an autonomous rover designed for exploration. Fitted with a number of sensors, can navigate spaces & react to the things it finds.

Dev JoshiDev Joshi 04/23/2014 at 19:080 Comments

Discussion of ‘hardware’ in this context refers to the electro mechanical components required to build Seek_r. It does not (directly) refer to the hardware requirements in terms of clock speeds, RAM space and other electronics. That comes later.

The mechanical configuration of this project must either support or facilitate a number of functions – these include:

* Some kind of frame, to which things can be mounted

* A reliable method of steering and locomotion

* Allow for easy access to electronics and connectors and so on for simple maintenance

* A mechanism to extinguish small fires


Steering & Locomotion

Steering and locomotion can be accomplished through a single system by implementing some kind of differential drive. Contenders for the position are:

PRO: Look badass

PRO: Zero turning circle

PRO: Capable of crossing uneven terrain

CON: Requires a precise implementation for reliability

CON: Pre-fab implementations are inflexible and expensive

CON: Prone to damage due to the number of parts involved

PRO: Very stable

PRO: Relatively easy to implement

CON: Steering not particularly accurate

CON: Traction control issues

PRO: Very stable

PRO: Relatively easy to implement

CON: Steering not particularly accurate

CON: Traction control issues

A direct differential drive – in the format of two directly driven wheels (motor --> gearbox --> wheel) seems like the most straight forward way to build something like this. The specific implementation of the differential drive could be done in one of the following ways (though the considerations listed here mostly apply to all of the methods motioned above):

PRO: Afford precise positioning and indexing

PRO: Inherently position aware (assumed only)

PRO: Wide range of options and good availability

PRO: Easy to lock output shaft position (braking)

PRO: Good torque response

CON: Expensive

CON: Requires a relatively complex hardware and/or software interface

CON: Positioning is implicit, not actually known

CON: Non trivial drivers required


PRO: Cheap and readily available

PRO: Easy to configure torque/speed/size with an appropriate gearbox

PRO: Simple hardware interface

CON: No position/velocity feedback without additional parts

CON: Gearbox almost certainly required (at additional cost)

CON: Potentially high peak current draw to deal with


PRO: Readily available in a number of sizes/configurations

PRO: Simple, well documented interface

PRO: Position+PID based control makes positioning parametric vs. Implied positioning from a velocity controlled system

PRO: Low voltage, built in gear box

CON: Relatively expensive

CON: Continuous rotation requires hardware modification

CON: PID implementation isn’t great in some instances

CON: Cheaper units are not super reliable, can be prone to backlash

For mechanical simplicity, drive wheels should be connected as directly as possible to motor/gear box output shafts. Belts and chains introduce all manner of issues into systems like this. With an inline motor-gearbox type assembly, having both driving units on the same axis ill likely result in a fairly wide chassis which should be relatively stable.

Given the cost complexity trade off and the potentially sizeable payload the rover will need to move (at some speed) a DC motor solutions looks to be best fitting in this scenario.

With this in mind, whilst looking for a suitable combination of drive wheels, motors, gearboxes and controllers I came across the RD02 from DevanTech.

The RD02 kit comprises:

2x EGM30 DC motors with pre fitted encoders: http://www.robot-electronics.co.uk/htm/emg30.htm

2x 100mm diameter wheels on 5mm hubs

2x Wheel-motor mounting brackets

1x MD25 2 channel motor control board with serial interface: http://www.robot-electronics.co.uk/htm/md25tech.htm

Making use of the RD02 kit greatly simplifies the construction and management of the locomotive subsystem in the rover. Encoder feedback, DC motor control and software interface development of this sort constitute an entire project in their own right.

The MD25 has a simple serial communications interface (UART or I2C) and hosts all of the required hardware/firmware to drive two encoded motors concurrently. The Interface it presents hides the majority of motion control complexity from the main MCU requiring simple command to manipulate the drives.

Power

 The MD25 has a simple serial communications interface (UART or I2C) and hosts all of the required hardware/firmware to drive two encoded motors concurrently. The Interface it presents hides the majority of motion control complexity from the main MCU requiring simple command to manipulate the drives.


PRO: Limitless supply

PRO: No issues delivering required current.

CON: Limited range of movement (there will be a cable attached to the rover)

CON: Bulky

PRO: Cheap and readily available

PRO: Understood technology

PRO: Easy to recharge

PRO: Generally rated to 12V

CON: Bulky and heavy

CON: Relatively low energy density

PRO: Super common

PRO: Reasonable energy density

PRO: Can be high drain

CON: Not rechargeable

CON: Not sustainable

CON: Expensive in the long run

PRO: Excellent energy density

PRO: Did I mention excellent energy density!

CON: Somewhat prone to exploding

PRO: Easy to obtain (for the time being)

PRO: Very stable

PRO: Deep cycle friendly

PRO: Very low self discharge rate

PRO: Good energy density

CON: High cost

CON: Being phased out due to environmental issues

CON: Terrible temperature coefficient

CON: Cells are slightly lower voltage (1.2V vs 1.5V)

PRO: Good availability

PRO: Discharge profile well suited to application (low internal resistance favourable to high current draw)

PRO: Lower environmental impact vs. NiCd

PRO: Easy to charge

PRO: Higher capacity than equivalent NiCd

CON: Relatively expensive

CON: Bulky

CON: Easy to damage during charge/discharge cycle

PRO: Mature technology, well understood

PRO: High voltage, high capacity

PRO: Excellent energy density

PRO: Low self discharge rate, no memory effect

PRO: Low profile

CON: Poor shelf life and life cycle

CON: Expensive

CON: Complex charging procedure

CON: High internal resistance not suited to high drain application

CON: Temperature sensitive

PRO: Very high energy density

PRO: Better life cycle vs. LiOn

PRO: High drain

PRO: Ultra lightweight

CON: Very expensive

CON: Complex charging procedure

CON: Temperature sensitive – explosive hazard

CON: Weight saved vs. Cost insignificant compared to LiOn

PRO: Badass!

PRO: Ultra efficient

PRO: Near limitless fuel supply

PRO: Excellent energy ratings

CON: Dubious energy density

CON: Experimental tech (read: dangerous)

CON: Prohibitively expensive

CON: Recharging and cell management requires actual science

Based on this evaluation, nickel-metal hydride (NiMh) based battery packs appear to be the most suitable for this project.

High capacity NiMh packs are produced en-masse for use in model, radio-control and hobby cars/boats/aircraft etc. Typically cells are configured in to 7.2V battery packs with capacities ranging from 2100-4500mAh.

Ebay landed me a set of 4 7.2V 3800mAh packs for around £30. Connecting two sets in series provides a 14.4V supply (which the motor control board will regulate for use). Two pairs in parallel afford a total capacity of 7600mAh.

Sensors, sensing and fire


Before any finalisations to the hardware and mechanical design can be completed, the other main peripherals in the system need to be specified. The sensors the unit will use to understand its surroundings form a major constituent of this set.

Collision avoidance

Effective and safe navigation requires, amongst other things, the consistent and reliably mitigation of objects and obstacles. This is especially true of some of the objects or obstacles are on fire.

In order to prevent the rover from running in to other objects in a space, it either needs to know how far away it is from other things (non-contact range sensors) or when it is about to run in to something (bump sensors).

Bump sensors can be constructed from any number of combinations of micro switches, levers, whiskers, rollers and armatures. Bump switches trigger (surprisingly) when they come in to contact with another object. The distance can be approximately judged by specifying the offset incurred by the length of the trigger arm. In the context of this project however they are not overly suitable as some of the objects in question may be unsuitable to touch (i.e. they are hot or on fire by which point, close proximity or bumping is too close).

Short range non contact range finders often make use of infra red light to measure distances which may not be suitable in a space filled with spurious IR heat sources. Effective in the 150-1000mm range, they can quickly provide accurate results and do not require tuning or calibration. They are however, sensitive to surface finish and reflection making them unsuitable for use in all but the most controlled spaces.

Longer range finders are almost exclusively sonar based within the confines of hobby robotics. This type of sensor is easy to implement using a small MCU and an ultra sonic transducer but pre-fab modules are available, presenting a number of interfaces.

The SRF02 I2C /UART range finder module "[Data Sheet]":http://robot-electronics.co.uk/htm/srf02tech.htm uses a single transducer to both pulse and listen when range finding. While this does limit the minimum sensing range to 150mm, the cost, size and minimal configuration benefits far outweigh the pros of a large, more accurate, more expensive unit. Given the application of object avoidance, a 150mm minimum range provides a good bubble to work in as a means of keeping the rover a safe distance away from other things. 

Thermal sensors

In order to detect a heat source at range – any temperature sensor specified for this project will need:

Thermocouples/thermistors/RTDs have limited range and repeatability ruling them out straight away. Discreet devices such as the "LM34":http://www.ti.com/lit/ds/symlink/lm34.pdf, "TMP102":http://www.ti.com/lit/ds/symlink/tmp102.pdf and "DS1820":http://datasheets.maxim-ic.com/en/ds/DS18S20.pdf are out of the running too. While it would be possible to use a device like this to measure changes in the air temperature and then use transfer function predictive filtering in firmware to guess where heat sources might be, it does not seem like the most reliable method. Further, the firmware would encompass a sizeable chunk of mathematics not well suited to the 8-bit integer arithmetic cores in the MCUs in the fray.

In the case of a lit candle, a sensor which can pick up a remote heat source at a distance need only have to report the difference between the candle temperature and the ambient room temperature. We are uninterested in any absolute values, only really the presence of something which seems much hotter than its surroundings. A sensor capable of detecting electromagnetic radiation in the IR band gap (0.9-14µm for near infra red, though technically the IR spectrum covers 700nm – 300µm).

Devices which can “see” heat sources include pyrometers, thermopiles, “infra-red cameras” and other single wave band radiation thermometers – dedicated thermal imagers, optical flame detectors, fibre optic temperature sensors and ultra sonic thermometers (yes, they are a thing). Spoilt for choice!

At the most basic end of the scale for this kind of device is the PIR (passive infra red) sensor. Present in most home security monitors, they are usually configured to provide a logic signal when a threshold has been breached within the device’s field of view (FOV). IR cameras are obviously more complex, providing location and magnitude information for heat sources in a 2D plane. "[Jonny Lee’s work]":http://johnnylee.net/projects/wii/ with Wii-motes successfully demonstrates the interfacing of IR cameras and MCUs to this effect. Thermal imaging of this complexity and detail is somewhat overkill for what is trying to be achieved here.

The TPA81 "[Data Sheet]":http://www.robot-electronics.co.uk/htm/tpa81tech.htm is an 8x1 pixel thermopile array configured as an I2C slave which is sensitive to radiation in the 2-22µm wavelength range. It can (and I quote) "... detect a candle flame at a range 2 metres (6ft) and is unaffected by ambient light".

Each pixel has a 5.12°x6° viewing scope, resulting in a total FOV of around 41° for the entire array. 

Extinguishing equipment

The number of options available for extinguishing fires in this context are as varied as they are numerous. Aerosols, "[acoustic waves]":http://www.youtube.com/watch?feature=player_detailpage&v=XVLXg9lcECQ#t=20s, robotic snuffers, "[science fiction]:http://www.youtube.com/watch?v=PQMHRqckNl4 and more came in to the fray at one point or another.

After (unsuccessfully) experimenting with a "[small pump]":http://www.cornwallmodelboats.co.uk/acatalog/water_pumps.html and the mister (atomiser) from a spray bottle the conclusion was reached that the simplest and safest way to douse the candle flames was to simply blow them out. 

Discussions