Almost exactly one year after the first PX4 announcement, we would like to introduce our newest member of the family, Pixhawk! For those familiar with the existing PX4 electronics, it is the all-in-one board combining PX4FMU + PX4IO, combined with a processor and sensor update and a number of new features. The current board revisions will however remain in full service and active development and are fully compatible. Pixhawk is designed for improved ease of use and reliability while offering unprecedented safety features compared to existing solutions.
Pixhawk is designed by the PX4 open hardware project and manufactured by 3D Robotics. It features the latest processor and sensor technology from ST Microelectronics which delivers incredible performance and reliability at low price points.
The flexible PX4 middleware running on the NuttX Real-Time Operating System brings multithreading and the convenience of a Unix / Linux like programming environment to the open source autopilot domain, while the custom PX4 driver layer ensures tight timing. These facilities and additional headroom on RAM and flash will allow Pixhawk the addition of completely new functionalities like programmatic scripting of autopilot operations.
The PX4 project offers its own complete flight control stack, and projects such as APM:Copter and APM:Plane have ported their software to run as flight control applications. This allows existing APM users to seamlessly transition to the new Pixhawk hardware and lowers the barriers to entry for new users to participate in the exciting world of autonomous vehicles.
The flagship Pixhawk module will be accompanied by new peripheral options, including a digital airspeed sensor, support for an external multi-color LED indicator and an external magnetometer. All peripherals are automatically detected and configured.
Features
32 bit ARM Cortex M4 Processor running NuttX RTOS
14 PWM / Servo outputs (8 with failsafe and manual override, 6 auxiliary,
high-power compatible)
Abundant connectivity options for additional peripherals (UART, I2C, CAN)
Integrated backup system for in-flight recovery and manual override with
dedicated processor and stand-alone power supply
Backup system integrates mixing, providing consistent autopilot and manual
override mixing modes
Redundant power supply inputs and automatic failover
External safety switch
Multicolor LED main visual indicator
High-power, multi-tone piezo audio indicator
microSD card for long-time high-rate logging
32bit STM32F427 Cortex M4 core with FPU
168 MHz
256 KB RAM
2 MB Flash
32 bit STM32F103 failsafe co-processor
ST Micro L3GD20H 16 bit gyroscope
ST Micro LSM303D 14 bit accelerometer / magnetometer
MEAS MS5611 barometer
5x UART (serial ports), one high-power capable, 2x with HW flow control
2xCAN
Spektrum DSM / DSM2 / DSM-X® Satellite compatible input
Futaba S.BUS® compatible input and output
PPM sum signal
RSSI (PWM or voltage) input
I2C®
SPI
3.3 and 6.6V ADC inputs
External microUSB port
Power System and Protection
Ideal diode controller with automatic failover
Servo rail high-power (up to 10V) and high-current ready (10A +)
All peripheral outputs over-current protected, all inputs ESD protected
- Monitoring of system and servo rails, over current status monitoring of peripherals
Dimensions
Weight: 38g (1.31oz)
Width: 50mm (1.96")
Thickness: 15.5mm (.613")
Length: 81.5mm (3.21")
Availability
This announcement is a service to our users and developers to allow them to plan their hardware roadmaps in time, and to show what we're currently working on. The board will not be immediately available, but 3D Robotics is taking pre-orders for Pixhawk now, and will begin shipping in late October [Update 11/11: the current expected ship date is late Nov]. The price is $199.99.
Comments
as for me, I would buy complete solution like NAZA, where everything is included, no need to think about how to get vibro isolation, how to get power, and other things, just simple plug and play.
Will this board be less susceptible to outside RF, electrical, and magnetic noise? IMHO, APM hardware is horrible at noise blocking and protection of all kinds. The new GPS/compass to mount far away from the FC and the whole compassmot procedure seem to be proof of underlying shielding and filtering problems.
Thanks Bill, did that already, this is a recurring issue reported over and over and over, akin to suicide for any business.
@Lorenz: Thank you for the thorough reply. You and your group at ETH have done some wonderful work, along with your colleague Rafaello D'Andrea, who put on an amazing show at TED this summer. Thanks for pointing me toward the Odroid boards, which I only looked at in passing before. I have been playing with a little embedded computer vision with the SoCs I mentioned, as kind of a side applied science research project. Both will do the job (one better than the other). I really like the BeagleBone for its power management, in addition to processing power. It has a much smaller community than RasPi though! I think some really cool autopilot capabilities are just around the corner, particularly with SoCs and centimetre accuracy GPS. I can't wait to play with the new Pixhawk for starters.
@chris,
ordering not possible outside US? I tried to order on 3d robotics but it always says unable to ship to my address while this is the exact same I used for all my previous orders... (I am in united arab emirates)
Ok Chris,
Despite what I wrote earlier I tried to order PixHawk
I get this in response to a valid address in New South Wales AUSTRALIA
We are unable to ship the selected items to your shipping address. Please choose another shipping address.
I have on numerous occasions in the past purchased from 3DR using the same valid address, so what has changed, too hard to ship to Australia now ?
What do I do now ?
@Anyone asking about sensing: First, why do you conclude that a two-year old sensor design (MPU6K and HMC5883) is better than one released by ST this summer? You also might be delighted to hear that the sensor technology in the HMC5883 is actually either based on ST technology or at least pretty much the same, as you will find out if you read the HMC5883 press release. So if Honeywell does rely on ST and you're appraising Honeywell, they can't be so bad, can they?
What also might have misguided you if you did testing is that the internal low pass filters of these different sensors have different minimal cutoff frequencies, and the ST sensors have higher cutoff frequencies, which need to be accounted for. In our MAV / sUAS application vibrations are common and important. If you ran your default flight control code without adjusting your estimation algorithms correctly, you therefore compared apples and oranges.
We did proper validation and head-to-head tests, and the PX4 drivers come with an internal Butterworth lowpass designed by Leonard Hall for this sensor, and in contrast to the APM 2.5+ setup which reads them at just 200 Hz, the sensors are read at 800 Hz and prefiltered, which yields great results as you will see in the flight videos that will appear within the next weeks.
The CAN support alone will turn lots of heads.
@Adam: As you can safely assume we're watching the available System on Chips (SoC) quite actively. Your conclusions do not consider however a few important pitfalls: The BeagleBone is so dirt cheap because its an ongoing marketing campaign / PR stunt from Texas Instruments, pretty much like the ST Micro Discovery board series is one from ST. They use is to push their products out, and in contrast to 3DR they do not need to sustain themselves on it (you pay them back with your phone, set top box or car purchase). There is also a difference in volume.
Second, your projections on speed need to be revised again. You're comparing apples and oranges. To give you some real numbers to work with: What currently maxes out the APM 2.5 runs at around 7% CPU load on PX4 (FMU and PIXHAWK), with much better sensor front end filtering at 1 KHz. So we're talking about an order of magnitude here. From a performance / feature perspective there is nothing to gain from a gigahertz in the short term, but we can reassure you that we will keep evolving the platform as the need arises.
Third, while the SoC series used on the BeagleBone supports a number of the embedded buses we need, it does not support all industrial buses like CAN. Reliability and safety (e.g. benefitting from redundant motor setups on hex / octacopters) however require bidirectional interfaces, and CAN has been built for this purpose and serves Boing and Airbus well. The setup costs, part costs and the required companion interfaces for e.g. an OMAP SoC will make an autopilot at the typical volumes more expensive then the Pixhawk price. That is however just a *momentarily* conclusion. We are watching the SoC market very closely, and you can be reassured that there will always be a top-notch PX4 board that has more processing power than the autopilot software running on it actually uses.
We have put so many serial ports on it so that you can have BeagleBone as companion computer - but at this point its not something that will enable you flight-control wise to do new things on the user / functionality level. It would allow you to e.g. add a camera and do image processing - but if you consider that, I really recommend to look at the Odroid board series. With those you can do what the MIT guys do, but for under $200.
-
40
-
41
-
42
-
43
-
44
of 50 Next