Making TDoA 3 more accessible

2018-09-03 | Kristoffer Richardsson | Leave a comment

We started the work on TDoA 3 in May and it has been functional for a few months, but it is a bit cumbersome to make it work since it requires compiling firmware with special flags and running scripts to configure anchors. To rectify this and make it more accessible we are now working on integrating it just like the other positioning modes; TWR and TDoA2.

Changes

The anchors already contained most of the required functionality. We have added support to change to the TDoA3 mode via LPP, that is using the Crazyflie as a bridge between the client and the anchors, transmitting data to the anchors via UWB.

In the Crazyflie TDoA 3 has been added as a third mode. This means that it is now auto detected when the Crayzflie is switched on and it can be selected from a client – no need for compile flags any more! We have also added a new mapping to the memory sub system to transfer anchor information for a dynamic number of anchors to a client. This means that instead of being available to the client as a long list of log variables and parameters, most of the TDoA3 information and configurations are available in a memory map using the same protocol we use to access real memory like the configuration EEPROM or the deck memories. This way we have much more freedom to define and transfer the data-structure to and from the Crazyflie.

The python client/lib is the piece of software that requires most changes. The UI (and implementation) was designed to handle 8 anchors, but with TDoA3 it must support a dynamic and larger number. The new memory mapping has of course to be implemented in the lib as well. The anchor position configuration part of the LPS tab will be separated into a dialog box to get more space for the controls. We also have some ideas for improvements in anchor position configuration (saving to file and sanity checking of configurations for instance) that will be easier to implement in the future as well.

Feedback

The driver for this work is of course to make the TDoA 3 technology available to anyone that wants to try it out. It is important to remember that it still is experimental and that we have mainly tested it in single room setups with a few anchors. Our hope is that more users will use it in various settings and that we will get feedback and contributions to iron out any remaining problems. We currently lack easy access to larger spaces which makes it hard for us to verify the functionality in a system with many anchors.

The code in the firmware for the anchors and the Crazyflie is mostly ready while there still remains some work in the lib/client, hopefully it can be committed and pushed during the week (see issue bitcraze/crazyflie-clients-python#349). If you want to try it out when the client is fixed, remember to upgrade the anchor firmware (including git sub modules), the Crazyflie firmware (including git sub modules), the python lib and the python client. Since this is still work in progress APIs and protocols may change until the first official release.

Log and param protocol V2

2018-08-27 | Arnaud | Leave a comment

Log and param are the two Crazyflie subsystems that have become the core means of communication with the Crazyflie.

The Log is a subsystem that contains functionality to transfer values of variables in the Crazyflie to a client. The client can setup log blocks, which are a list of variables, and start logging this log block at a certain rate. The Crazyflie will then send radio packets at the requested rate with the current values of the variables, thus enabling the client to read changing variables in the Crazyflie in near realtime. It is very useful for monitoring the state of the Crazyflie and further more, any log variable can be graphed in the python client.

Param is a subsystem that contains functionality to get and set the values of variables in the Crazyflie. This is essentially the opposite of Log, it allows the client to read or write variables that are read-only in the firmware.

Both subsystems are based on a Table Of Content (the TOC): at connection time the client pulls the list of log/param variables. This means that there is no hard-dependency between client and firmware and that we can develop new functionalities in the Crazyflie, adding log and param variables to access it without modifying the client.

The Log and Param subsystems have served the Crazyflie community very well, allowing for quick development of experimental and new functionalities. There has been a limitation that has become more and more painful lately though; we were limited to 255 variables due to the protocol using only one byte to encode the variable ID. This issue has now been fixed in the Crazyflie firmware and in the Crazyflie ROS driver by a pull request from Wolfgang at USC. We have recently also implemented the required changes in the Python lib to make it available in the python client (and any other python script using the lib). In the process, some bugs unfortunately found their way into the code, but they have quickly been fixed by a pull request from simonjwright. Thanks to every one involved!

So now Crazyflie supports up to 65535 log and 65535 param variables. This time we should be good for the foreseeable future! ;-).

Raspberry Pi Zero power deck

2018-08-20 | Tobias | 7 Comments

Ever since the Raspberry-pi zero was released we wanted to find-out what it would take to fly one with the Crazyflie 2.0. One immediate issue is the size and weight of the R-Pi-Zero. It is just a bit to big and heavy to make it work without modifying the Crazyflie 2.0. Also it requires 5V power which is something the Crazyflie 2.0 doesn’t provide if USB isn’t connected. Actually the R-Pi-Zero works well down to ~3.6V but this is still too high to reliably run directly from a single LiPo cell. So to begin with we created a Raspberry Pi Zero power deck. It is reusing the same step-up/step-down (STBB1) as used on the LED-ring to make things simple and the output is set to 3.8V. Other than that the UART and the I2C interfaced has been connected so that the raspberry pi zero could control the Crazyflie.

The raspberry pi zero would then be soldered to the deck with 0.1″ header pins. The result can be seen below and the power part works well. We chose to solder the deck header pins to the deck, instead of using the female deck connectors, to make it more sturdy. Another thing we did was fitting a Pi-camera using a 3D printed mounting bracket we designed. We think this is one of the interesting use cases, to run computer vision or maybe neural networks :-).

Well unfortunately this only solves the first part, powering the R-Pi-Zero from the Crazyflie 2.0. Next step will be to modify the Crazyflie 2.0 with bigger motors/props so that is can carry it for a decent time. So story to be continued…

Crazyflie RZR control board progress

2018-08-06 | Arnaud | Leave a comment

We have been thinking for a while about making a Crazyflie control board that could be used to make a bigger quadcopter using the Crazyflie firmware and deck. This idea has materialized in the Crazyflie RZR project.

The Crazyflie RZR is a quadcopter controller board based on the Crazyflie design, as pointed if our previous blog post, it is intending to bring the strength of the CF2 but in a little bit bigger package :-). It runs the Crazyflie firmware and feature the Crazyflie 2.0 deck port. It is capable of driving brush-less motor controller and has an uFL port for an external 2.4GHz antenna. It also contains the new quadcopter-optimized Bosch BMI088 IMU. We have made some progress lately on the Crazyflie RZR, we have just got the first initial sample from the manufacturer shown in the picture above.

We are not sure yet when the RZR will be in the shop, but the project is definitely going forward. We will keep posting information about the project as it develop.

Lighthouse prototype deck development

2018-07-30 | Arnaud | 4 Comments

We already wrote in a previous blog post that we where working on a Lighthouse positioning receiver deck for the Crazyflie 2.0. In this post we will describe a bit what has been the development process so far for this deck as it is an example of how to develop with the Crazyflie. Basically, our way of working often is to try to get one things working after another, this is what we have done here: we start from a hack and then we replace hardware and software pieces one after the other to make sure we always have one half (hardware of software) we can relie on.

The lighthouse deck started as a Fun Friday project, and as such we usually want to hack something together to see if the idea can work. So I looked around the web to get some information as of how to receive the lighthouse positioning signals and decode it. I found the vive-diy-position-sensor GitHub project by ashtuchkin. The project describe the schematic and contains the software for a Teensy board to receive a lighthouse 1.0 signal and calculate the position of the receiver. I went forward and cabled the circuit on a Crazyflie prototyping deck and attached a Teensy board to another prototyping deck. The idea is to install these two board above and bellow a Crazyflie:

Discreet-component Lighthouse receiver

Teensy to decode the lighthouse signals

The signal from the lighthouse receiver goes to the Teensy, then the serial port of the Teensy is connected to the serial port of the Crazyflie. As a first approach the Teensy was configured and we could get the position data using the Teensy USB port. When everything was working correctly I could implement a small deck driver in the Crazyflie to receive the position and push it in the Kalman filter. This way I could get a Crazyflie 2.0 flying in lighthouse with minimal firmware work.

The obvious next step was to get rid of the Teensy, this was done by implementing the lighthouse pulse acquisition and interpretation in the Crazyflie. Once that was done, we could make our own deck. Instead of using op-amp we used the official receiving chip available at this time, the TS3633:

First lighthouse receiving deck prototype

This board implements up to two receiver which would allow to get the orientation as well as the Position of Crazyflie. Due to questionable soldering only one receiver has ever worked but the prototype was useful to test the concept anyway, one of the lesson learned is that the receiving angle of the two flat is not big enough to fly very high, with the two lighthouse base station near the ceiling we could only fly up to ~1.5m before loosing the signal. We would need a microcontroller or other chip capable of acquiring the signals on the deck since the Crazyflie 2.0 deck port only has two input capable of acquiring the pulses.

At this point informations about Lighthouse 2.0, the next version of Lighthouse tracking that will allow to cover much bigger area, started appearing on the internet and a new receiver chip was release to receive the signal, the TS4231. One big difference was that Lighthouse 2.0 would transmit data in the laser carrier. The data transmitted are in the range of 1 to 10MHz dixit the TS4231 datasheet so it makes them impractical to acquire with a microcontroller. This gives us a perfect opportunity to play with the iCE40 FPGA and the icestorm open-source toolchain that has just been release.

The result is a deck containing enough receiver to cover a much bigger flying space and an iCE40UP5K FPGA to acquire the signals sent by the lighthouse. There is already two prototype of this design: one without SPI flash, so the Crazyflie would have to embed the FPGA configuration bitstream and program it at startup and the latest one has an SPI flash so the deck can start by itself:

First FPGA-Based lighthouse deck prototype

Partially populated second FPGA-Based lighthouse deck prototype, now with SPI flash

As a first approach the FPGA will acquire the Lighthouse 1 pulses and send the raw timing via a serial port to the Crazyflie. The Crazyflie can then decode and interpret the pulse. I am currently playing with the idea of maybe running a picorv32 Risc-V 32 bits CPU core in the deck, this will allow to acquire and interpret the pulses in the deck and send angles to the Crazyflie, this would greatly lighten the processing load on the Crazyflie 2.0. Eventually this FPGA should be able to acquire and decode the Lighthouse 2.0 signals.

This is very much work in progress and we will write more about the Lighthouse deck when we have further results.

Bitcraze at iROS 2018 in Madrid

2018-07-23 | Arnaud | Leave a comment

Last week we received the visit of Wolfgang from USC, he is the creator of the Crazyswarm project. It was great to have him here at the office. One of the subject of discussion was to prepare a demo for iROS 2018 on October 1-5 2018 in Madrid.

We will be in booth 91, if you are attending iROS 2018 feel free to pass-by and say hello. We are planning a couple of demos:

Crazyswarm with at least 6 Crazyflies flying in a Qualisys mocap system.
Running a fully autonomous Crazyflie with the Loco Positioning System.
Hopefully, some demo of autonomous flight using the lighthouse positioning. This is still not fully working but I have at least 2 full months to get something flying :-).

If you would like to see us demo anything more/else tell us in the comments and we will see if we can setup something.

We used Wolfgang’s visit to finalise the Qualisys support for Crazyswarm. It is now pushed and documented, this means that if you have a Qualisys system and a couple of Crazyflies you can now fly them autonomously using the Crazyswarm framework. It also means that we now have Crazyswarm up and running flawlessly at the office, it will help us testing related pull-request and supporting advanced functionality like the high-level-commander in the Crazyflie python lib.

As a side note, Bitcraze is spread very thin these weeks since most of us are in vacation (I am basically alone). We usually miss one Monday post per year, it was last week and the Wolfgang visit is my excuse :-). Sorry in advance if there is any delay to answer mail, forum or other requests. From next week, the rest of the team will slowly start to come back.

Lighthouse positioning for Crazyflie

2018-06-25 | Arnaud | 3 Comments

A while ago we bought an HTC Vive for the Bitcraze office. This was partly for having fun with VR, but is was mostly because we had hope to use the vive tracking system with the Crazyflie. We are making progress with the idea and we just received our latest prototype:

The Lighthouse tracking system is the hardware component of steamvr tracking, it is used by the HTC vive to get the full position and orientation of the Vive VR head mounted display and game controllers. It has sub-millimeter precision and low latency, which is key to achieve immersive VR experience. The system works by having base-stations installed in the room. The base station sweeps two rotating infrared laser planes. A receiver is basically a photodiode, by detecting when the photodiode is hit by the sweeping lasers, the receiver can measure at which angle it is seen by the base station. With enough receivers and/or base-stations, it is possible to calculate the receiver position and orientation. If you want to read more about how lighthouse works, there has been awesome work of reverse engineering and documentation made by the open-source community.

As far as Crazyflie is concerned the lighthouse system has one major advantage: the position and orientation can be calculate in the tracked object which means that the Crazyflie can be completely autonomous and there is no limit in the number of Crazyflies that can be tracked at the same time.

Lighthouse has been my fun-Friday project for a couple of month and the early results are very encouraging.This is still very much work in progress, so stay tuned for future blog-posts about the subject :-).

A[nother] french in Malmö

2018-06-18 | Benjamin Kuperberg | Leave a comment

Malmö is known to be very beautiful in June. At least that’s what i’ve heard from the Bitcraze team when we talked about having a meet-n-geek in their offices in Sweden.

I’ve been working on the crazyflies and developing artist friendly interface to control them for a year now, and I always was impressed with their philosophy of work and communication. Over the past year, they’ve helped me and my companies a lot to be able to create the shows we want, with our specific needs and constraints, that researchers don’t have, like fast installation time, or confidence in drone take off (you want to make sure that a drone won’t hit the theater’s director’s head at the very beginning of the show !).

For that I created LaMoucheFolle, which is an open-source software with a nice UI to be able to connect, monitor and control multiple drones. LaMoucheFolle is not made to be a flight controller, but acts more as a server that any software will be able to use, like Unity or Chataigne. That way, people and users don’t need to handle all the connection, feedback, warning, calibration process and can concentrate essentially on the flight and interaction.
While the fist version of LaMoucheFolle got me through most of the demos and workshops, I knew that it could be vastly improved if I understood better the drones, so I decided to ask the creators to meet them, and here I am !

As I hoped, being physically there allowed me to understand better what are their practices, and allowed them to better understand mine, so we could find a way to improve both their Crazyflie ecosystem and my contribution to it.

So I refactored, redesigned and improved LaMoucheFolle to a (soon to be released) V2, featuring :
– New drone manager interface with a more intuitive feedback of the drones
– New multi-threaded drone communication mechanism
– New state-safe sequenced initialization and flight control of the drone, with a progressive take-off vastly increasing its stability
– Unity demo app and Chataigne demo session to show how to control from other softwares via OSC

While developping the new version and talking to the team, some key features and improvements for the use Crazyflies in shows were discussed and some of them are now in research/development :
– Health analysis : using the accelerometer’s data and Tobias’ magic brain, it allows to test the motors while the drone is on ground and find out if one or more are problematic (too much or not enough vibration, meaning either the propeller or the motor needs to be repaired/replaced)
– Battery analysis : when the battery is fully charged, this allows to have an automated motor sequence which will find out if the battery has an abnormal discharge behavior
– Steath mode : Shut down all the system leds (the 4 builtin leds on the drone) so it can be invisible, ninja-style
– Normalized battery level and low battery log values : it allows for safe and consistent feedback of the battery level in percent, and an indicator if the drone should land soon. It will also possible to use this value to monitor the charging progression of a drone.
– LedRing fade pattern : This mode allows for easy fading between solid colors on the led ring, so it’s not necessary to stream all the colors from one color to another, but instead having a very beautiful smooth interpolation, using only 2 parameters : the target color and the fade time.

I hope you are as excited as I am about those new features, and if you’re not, please tell us what would make you “vibrate” !

I’ve been spending the last week at their office, and it was a great week : I initially came to improve my software, and discuss about future development of the Crazyflies [which was great], but what i’ll remember the most from this trip is by far the human aspect of the Bitcraze team.
Marcus, Kristoffer, Tobias, Björn and Arnaud are amazing and i’m really happy to have had the chance to see them working, and collaborate with them. I admire their choices of being fully transparent on their work and amongst themselves, and there is a natural kindness mixed to the passion for their project that makes working there feel like everyday’s special. Also, Malmö is very beautiful indeed :)

Thank you very much and I hope to reiterate the experience soon,

Skål

MoCap deck

2018-06-11 | Tobias | 9 Comments

The Crazyflie 2.0 has been flyable in MoCap systems such as Qualisys, Vicon or Optitrack for quite a while thanks to the Crazyswarm project. For the MoCap systems to be able to track the Crazyflie it needs to be fitted with reflective markers. These can be attached on e.g the motor mounts which in some cases might be the best solution, however we also liked the idea of creating a deck where the markers easily can be attached and in many, repeatable configurations, that is why we created the Mocap deck. This deck is now soon to be released but before we start manufacturing it would be great to get some feedback.

The deck has M3 sized holes which is spaced on a 5mm grid. The deck also has footprints for two optional push buttons that can be used to e.g. trigger a take-off or start of a demo. And as the battery holder deck, which has no electronics, so this can be mounted upside down for better fitting of the markers, if the buttons aren’t mounted of course.

We are collaborating with Qualisys, also based in Sweden (in Gothenburg), to make the Crazyflie more moCap friendly and make it easy to use together with the Qualisys system as well as other mocap systems. Qualisys will provide the markers for the moCap deck.

Qualisys and Bitcraze are exhibiting together at IROS in Madrid where we will show some awesome demos with a moCap system as well as other position technologies and we are also hoping to fly a Crazyswarm. We will publish more information when we get closer to the conference.

TDoA 3 progress

2018-05-28 | Kristoffer Richardsson | Leave a comment

We have now worked a few weeks on the new TDoA 3 mode for the Loco Positioning System. We are happy with the results so far and think we managed to do what we aimed for: removing the single point of failure in anchor 0 and supporting many anchors as well as larger spaces.

We finished off last week by setting up a system with 20 anchors covering two rooms down in the lunch area of the office. We managed to fly a scripted autonomous flight between two rooms.

Work so far on the anchors

Messages from the anchors are now transmitted at random times, which removes the dependency on anchor 0 that used to act as a master that all other anchors were synchronized to. The drawback is that we get problems with collisions when two anchors happens to transmit at the same time. Experiments showed that at 400 packets/s (system rate) we ended up at a packet loss of around 15% and 340 TDoA measurements/s sent to the kalman filter for position estimation. We figured that this was acceptable level and added an algorithm in the anchors that reduces the transmission rate based on the number of anchors around them. If more anchors are added to a room they all reduce their transmission rate to target 400 packets/s in total system rate.

The anchors continuously keeps track of the clock drift of all other anchors by listening to the messages that are transmitted. We know that clocks do not change frequency suddenly and can use this fact to filter the clock correction to reduce noise in the data. Outliers are detected and removed and the resulting correction is low pass filtered. We have done some experiments on using this information and compare it to the time stamp of a received message to detect if the time stamp is corrupt or not, but this idea requires more work.

One interesting feature of the anchors is the limited CPU power that is available. The strategy we have chosen to handle this fact has been to create an algorithm that is efficient when handling messages. A timer based maintenance algorithm (@1 Hz) examines the received data and makes demissions on which anchors to include in the messages in the future as well as purges old data.

The Crazyflie

The implementation in the Crazyflie is fairly straight forward. The biggest change to TDoA 2 is that we now can handle a dynamic number of anchors and have to chose what data to store and what to discard. We have also extracted the actual TDoA algorithm into a module to separate it from the TDoA 3 protocol. The clock correction filtering algorithm from the anchors has also been implemented in the Crazyflie.

An experimental module test has been added where the TDoA module is built and run on a PC using data recorded from a sniffer. We get repeatability as well as better tools for debugging and this is something that we should explore further.

Work remaining

The estimated position in the Crazyflie is still more noisy than in TDoA 2 and we would like to improve it to at least the same level. We see that we have outliers in the TDoA measurements that makes the Crazyflie go off in a random direction from time to time, we believe it should be possible to get rid of most of these.

The code is fairly hackish and there are no structured unit or module tests to verify functionality. So far the work has been in an exploratory phase but we are getting closer to a set of algorithms that we are happy with and that are worth testing.

We have not done any work on the client side, that is support for visualizing and configuring the system. This is a substantial amount of work and we will not officially release TDoA 3 until this is finished.

How to try it out

If you are interested in trying TDoA 3 out your self, it is all available on github. There are no hardware changes and if you have a Loco Positioning system it should work just fine. There is a short description on the wiki of how to compile and configure the system. The anchor supports both TDoA 2 and TDoA 3 through configuration while the Crazyflie has to be recompiled to change between the two. The support in the client is limited but will basically handle anchors 0 – 7.

Have fun!