PiWars on usedbytes:Blog

M0o+ Shepherd's Pi

Mon, 13 Jun 2022 00:00:00 +0000

The last of the three “prescribed” challenges is Shepherd’s Pi.

In this challenge there are 6 sheep which need herding into a pen, and 3 wolves which must not be “killed” (knocked over).

M0o+ Nature's Bounty

Sun, 12 Jun 2022 00:00:00 +0000

After tackling Hungry Cattle, Nature’s Bounty is next on the hit-list.

Lots of videos/pictures in this post, because I think it helps a lot with explaining how it all works.

M0o+ Hungry Cattle

Sat, 11 Jun 2022 00:00:00 +0000

Hungry Cattle is probably the easiest of the three main Pi Wars challenges to do autonomously and so is the one I chose to tackle first.

The basic premise is delivering “cattle feed” (rice) to three troughs, placed at known positions within the arena.

M0o+ Computer Vision

Sun, 29 May 2022 00:00:00 +0000

To be able to locate the different game pieces in the Pi Wars challenges M0o+ has a very basic camera.

M0o+ is using a Pico for all the processing, so there’s really no chance of porting OpenCV to it. There are other frameworks - even some machine-learning ones like TFLite - which could work on the Pico, however having had success with hand-coding my own routines for Mini Mouse, I decided to do the same this year.

M0o+ Software Overview

Sun, 15 May 2022 00:00:00 +0000

I’ve written a number of posts now about different “building blocks” which make up M0o+:

There’s one crucial element missing from that list, which is needed to turn a collection of parts into an actual functioning robot than can compete in Pi Wars: The Software.

M0o+ Boom - Part 3

Sun, 08 May 2022 00:00:00 +0000

Last time, I derived the Forward Kinematics for the boom on M0o+. If you haven’t read the first two posts, then this one won’t make much sense (and I don’t want to repeat myself 😀):

M0o+ Boom - Part 2

Sat, 02 Apr 2022 00:00:00 +0000

This post is the second in a trilogy detailing the boom on my Pi Wars 2022 robot M0o+.

In the last post, I wrote about the mechanical and electronic design. In this middle post, I’ll be describing the equations which let me find out where the end of the boom is, and next time will go into the equations which let me control where the end of the boom is.

M0o+ Boom - Part 1

Mon, 28 Mar 2022 00:00:00 +0000

My Pi Wars at Home 2022 robot M0o+ is intended to resemble a telehandler (long-reach forklift).

Image credit: Jacquesbotha, CC BY-SA 4.0, via Wikimedia Commons.

A fairly crucial part of a telehandler - the thing that makes it a telehandler, really - is it’s boom, which the forks (or other tools) are attached to and can lift and extend.

Pico PIO Camera

Sat, 19 Feb 2022 00:00:00 +0000

This is a lengthy post which goes into lots of detail. The tl;dr is that the code is here: https://github.com/usedbytes/camera-pico-ov7670

In this post I’ll describe how I went about connecting an OV7670 camera module to the Raspberry Pi Pico that drive’s my Pi Wars at Home 2022 ‘bot M0o+. It’s all wrapped up in a separate library which should be easy-ish to incorporate in other projects.

Trying to get this camera working was the very first thing I did, as a sort of Pico “feasibility study” to decide if entering a Pico-based robot was practical. Initially I just bit-banged the interface and did get some results without too much effort, but things have escalated a fair amount since then:

M0o+'s Chassis

Sun, 23 Jan 2022 00:00:00 +0000

M0o+ is my Raspberry Pi Pico-based robot entry for Pi Wars at Home 2022. This post details what I’ve settled on for the basic chassis of the robot.

Rebound (top)

As I wrote in my initial post for the project, I had originally hoped to re-purpose an ancient RC car chassis to be the basis of my robot.

Sadly, it became apparent that all of the things which made it so much fun as an RC car when I was a child, made it almost entirely unsuitable for Pi Wars.

ESP32 WiFi/Bluetooth Bridge

Sun, 26 Dec 2021 00:00:00 +0000

M0o+ is my Raspberry Pi Pico-based robot entry for Pi Wars at Home 2022.

A Pi Wars robot needs to be remote-controllable, and network connectivity is extremely useful for telemetry, debugging etc. As the Pico doesn’t natively have any kind of wireless connectivity, I’m using an ESP32 to provide that functionality. It connects to a bluetooth gamepad, relaying controller input to the Pico, and acts as a WiFi access point to provide network connectivity.

Pico serial bootloader

Wed, 01 Dec 2021 00:00:00 +0000

My Pi Wars 2022 entry, M0o+, is powered by a Raspberry Pi Pico, which itself is powered by the RP2040 chip.

The RP2040 has a built-in boot ROM with a USB bootloader, which allows the chip to show up as a USB flash drive for uploading code. This is usually a very convenient interface, making it extremely quick and simple to upload code to the device.

However, in a Pi Wars robot, I don’t want to have to bring the Pico back to my computer and plug it in to iterate on the code - I see it as essential to be able to download new code to the robot wirelessly without needing to touch it.

M0o+ - Pi Wars at Home 2022

Sun, 14 Nov 2021 00:00:00 +0000

Pi Wars (at home) 2022 is upon us, and after taking 2020 “off”, I’ve had an application accepted for the competition. I’ll be blogging about my progress here the same way I did for 2019 and Mini Mouse.

The “in person” 2020 competition was postponed on account of the Covid-19 pandemic, and hasn’t yet taken place. In 2020, an “at home” competition took place, with competitors submitting videos for each challenge. The 2022 “at home” competition will follow the same model, with competitors constructing the courses at home and submitting videos of challenge attempts.

Run at Startup (without rc.local)

Sun, 24 Nov 2019 16:13:28 +0000

A question that comes up quite frequently in Pi Wars circles is “how do I make my robot program run at start-up?” Someone asked me that directly at the Pi Wars Miniconference - and I gave the same answer I always do: “Use systemd”.

In itself, that not a very useful answer, because systemd is a large and varied project with lots of tools and functionality. It’s just a bit too complex to describe over voice without context. However, with the help of examples, it really is quite straightforward to make a simple systemd service. There’s also all kinds of bells and whistles you can try out too if you’re feeling adventurous!

Pi Wars: Post Mortem

Sun, 31 Mar 2019 20:20:41 +0100

Pi Wars happened! It’s over! I can start doing something else with my life!

I had an awesome time at Pi Wars today and yesterday, with Mini Mouse performing pretty solid attempts at all of the challenges. I ran into a few hiccups on some challenges, but far outperformed my expectations on others.

Overall, I placed 3rd in my category, and am really happy to have been awarded first place for this blog, and Artistic Merit (for being “beautiful on both the inside and the outside” - thanks Rachel :-)

Pre-Competition Status Report

Thu, 28 Mar 2019 20:58:21 +0000

With two days to go until competition day (only one day for schools and clubs!), I thought I’d capture the status of Mini Mouse, for posterity and for something to compare against in a Post-Competition Post-Mortem.

I’m quite happy with where I’ve got to. The autonomous challenges are all working mostly reliably in practice runs, and the performance seems to be competitive. Of course there’s no guarantee that it’ll perform “on the day”, but I’ve done all I can at this point!

Battery Discharge Profile

Sat, 16 Mar 2019 22:16:03 +0000

Earlier, on the PiWars Discord, we were having a spirited discussion about the discharge profile of lithium batteries, disagreeing about the relationship between cell voltage and remaining charge level. I’ve never actually checked this for myself before, so I thought I’d run a little experiment to see.

When fully charged, a lithium polymer (LiPo) cell should be at 4.2 Volts. The “nominal” voltage for a LiPo cell is 3.7 V, and figures for the “minimum” recommended discharge level vary - but quoted figures range from 2.75 V to 3.0 - 3.3 V. As we’ll see later on, towards the tail end of the discharge curve the voltage of the cell drops very steeply, meaning that the precise “cut off” voltage doesn’t matter so much from the point of view of using the maximum amount of energy out of your battery.

Autonomous Challenge: Hubble

Wed, 13 Mar 2019 16:15:56 -0700

The Hubble Telescope Nebula Challenge (previously Somewhere Over the Rainbow) is one of the autonomous challenges, requiring the robot to visit the four corners of a square arena, starting from the centre. In each corner, there’s a coloured board - and for maximum points the robot must visit the corners in an order specified by their colours (Red, then Blue, Yellow and finally Green), as quickly as possible.

My overall strategy for this challenge is to spin on the spot at the starting position, to identify where each colour is. Once all four corners have been identified, they can be visited in the appropriate order. Originally, I was also intending to use the camera to make sure the ‘bot stays on-course whilst en-route to the corners, and to detect when a corner has been “visited”. Ultimately, I’ve found dead-reckoning using the IMU and counting wheel rotations is reliable enough, and so I’ve skipped the added complexity of using camera feedback for the actual “visiting” phase.

Setbacks

Fri, 08 Mar 2019 18:20:44 +0000

Three weeks ago, I got a basic implementation of The Hubble Telescope Nebula Challenge working, which made Mini Mouse’s code “feature complete” from the point of view of entering all of the autonomous challenges.

Honestly, I was pretty darn happy. All three challenges were working well enough to be “competitive”, and with more than a month to go until the competition, all that was left was optimisation. So, I spent a week doing rapid-fire blog posts .

Mini Mouse's Armament

Wed, 20 Feb 2019 20:37:14 +0000

Now, irrespective of personal views on arming robots, unfortunately one of the Pi Wars challenges requires¹ that we weaponise them. For Space Invaders we have to be able to knock down as many targets as we can (up to 5) from 1-1.5 metres away.

Plan A (top)

This challenge has caused me a great deal of head scratching as I sought out the right option for projectile launchers for Mini Mouse. For Bot Matrix the plan was dual-wield repeating rubber band guns - so that was my starting point for Mini Mouse too.

Edge Detection

Tue, 19 Feb 2019 20:23:05 +0000

Identifying edges is one of the key “features” of the human vision system. Our brains have evolved to identify edges and it’s one of the things which lets us pick out different objects in the world. If we want to teach our computers to see, Edge Detection is one of the key tools needed to reach that goal (at least, according to “classical” methods. Things change with Deep Learning).

Autonomous Challenge: Blast Off

Sun, 17 Feb 2019 19:52:08 +0000

Blast Off: the Straight-ish Line Speed Test is, to my mind, the easiest of the three autonomous challenges in this year’s competition. As such, it’s the one which I started working on first.

The task is to get from one end of the course to the other as quickly as possible. There’s effectively two feasible options:

Follow the walls
Follow the white line

Given that I don’t have any side-facing sensors to follow the walls, to me the easiest option looked like following the white line.

Mini Mouse's Software

Sun, 17 Feb 2019 11:43:26 +0000

Before I dive in to details on individual challenge implementations, I thought it would be sensible to describe how the software on Mini Mouse sits together. Some of this information is duplicated in We’re in, but there’s more detail here, and things have evolved since then.

The diagrams on this page are created with graphviz (or more correctly, dot), which doesn’t give the prettiest of drawings, but does stop me wasting time obsessing over layout.

Final Hardware

Sat, 16 Feb 2019 22:03:44 +0000

It’s been a few frantic months of robot building between my last post and now, a decent portion of it being focussed on finalising the chassis. Since around September I’ve had my prototype chassis working, which has given me a great test-bed for software implementation testing and debugging the hardware design.

There were three main (known/intentional) limitations with the prototype chassis, borne out of a need to get it built and tested before I could finalize some of the details. Specifically:

(Quite) Robust Thresholding

Sat, 27 Oct 2018 16:55:59 +0100

Today, I’m detailing the method I’m using to threshold images for line detection. Also included are some hard-learned lessons about premature optimisation and the importance of re-evaluating your decisions frequently!

Image thresholding is the process of taking an image, and converting it into a two-color (just black and white, no grey) image. This is useful for lots of image processing algorithms - like edge or blob detection.

My initial focus here has been on implementing a white line detector for the “Blast Off” challenge.

We're in!

Sat, 06 Oct 2018 11:09:04 +0100

So last weekend Mike and Tim sent out the confirmations for Pi Wars entries, and Mini Mouse made the cut! We’re going to Pi Wars!

Doug on the Discord suggested the idea of sharing successful applications for people to look at to see if they can improve their chances next time - so here’s mine in case it’s useful.

It’s been a little over a month since my last post, but I have moved things forward a bit in that time.

PCBs

Sun, 26 Aug 2018 10:11:04 +0100

For Mini Mouse, I’ve designed two PCBs - one for motor drivers, and one for the power supply/control circuitry. These were my first boards using the open-source KiCad suite of EDA tools, having previously been an EAGLE user.

I got the boards made at dirtypcbs.com, panelising the two boards into a single 5x5 square. This is the third set of boards I’ve made with dirtypcbs, and I’ve always been very pleased with them.

Power

Sun, 26 Aug 2018 09:50:46 +0100

Every robot needs power, generally at two or more voltages - one “high” voltage directly from the battery to feed the power hungry motors, and one “logic” voltage to drive the electronics. The motor voltage needs to be able to provide high current, and the logic voltage needs to be relatively “clean” to avoid messing up the electronics with excessive noise or voltage drops.

Mini Mouse uses a 3S LiPo battery, which means it’s quite important not to over-discharge the battery, as that might damage the cells (and could even result in fire). On Bot Matrix I had a mechanical power switch to disconnect the battery, but I don’t really have space in Mini Mouse, and I also wanted to try designing an electronic switch as a useful building block for future projects. So, I need an electronic switch design which has no current draw when “off”.

Motors

Sun, 03 Jun 2018 14:55:44 +0100

Motors were the first thing I worked on selecting at the start of the development process. Given the space constraints I’m working with, motors would be a critical component which would inform the design of the rest of the robot.

Stepper vs Brushed DC (top)

In Bot Matrix I had DC motors with hall-effect encoders on the back. I used the encoders to implement closed-loop control for motor speed, and for odometry for dead reckoning. In my opinion, accurate speed control and odometry are must-have features for a robot - and so this weighs heavily in to the motor selection process.

Mini Mouse

Mon, 30 Apr 2018 19:47:31 +0100

In the past 12 months, I built Bot Matrix, aiming for PiWars 2018 (though unfortunately I wasn’t selected as a competitor). It was the first chassis I built from scratch, and I learnt a lot in the process! Bot Matrix was large (on the limits of the PiWars size allowance), and used pretty large motors and drivers.

This year’s project is a little bit different - I always fancied building a robot in a computer mouse and so that’s what I’m going for. Min(n)i(e) Mouse is different from Bot Matrix in nearly every way. Bot Matrix used a Pi 3, two dedicated microcontrollers for off-board processing, tracks, 5 ampere motors, tracks, and 3072 LEDs. By comparison, Mini Mouse will have a Pi Zero, only. All processing and control will be handled on-board by the Pi.