使用 C MAVLink 库 (mavgen)

The MAVLink C library generated by mavgen is a header-only implementation that is highly optimized for resource-constrained systems with limited RAM and flash memory. It is field-proven and deployed in many products where it serves as interoperability interface between components of different manufacturers.

This topic explains how to get and use the library.

Getting the C MAVLink Libraries

If you are using a standard dialect then download the MAVLink 2 library from Github: c_library_v2.

The MAVLink 2 library supports both MAVLink 2 and MAVLink 1, and is rebuilt for all the standard dialects whenever any of the definitions in the mavlink/mavlink repo change. It supersedes the MAVLink 1 library (c_library_v1), and should be used by preference.

If you need libraries for a custom dialect then you will need to install mavgen and generate them yourself. These can then be used in the same way as the pre-generated libraries.

The libraries can be placed/generated anywhere in your project tree.

Adding Libraries

The example below shows them located in: generated/include/mavlink/v2.0/.

To use MAVLink in your C project, include the mavlink.h header file for your dialect:

#include <your_dialect/mavlink.h>

This will automatically add the header files for all messages in your dialect, and for any dialect files that it includes.

Only include the header file for a single dialect. If you need to support messages from a number of dialects then create a new "parent" dialect XML file that includes them (and use its generated header as shown above).

Do not include the individual message files. If you generate your own headers, you will have to add their output location to your C compiler's search path.

When compiling the project, make sure to add the include directory:

$ gcc ... -I generated/include -I generated/include/common ...

Adding Library to Cmake Project

To include the headers in cmake, install them locally, e.g. into the directory install:

cmake -Bbuild -H. -DCMAKE_INSTALL_PREFIX=install -DMAVLINK_DIALECT=common -DMAVLINK_VERSION=2.0
cmake --build build --target install

Then use find_package to get the dependency in CMakeLists.txt:

find_package(MAVLink REQUIRED)

add_executable(my_program my_program.c)

target_link_libraries(my_program PRIVATE MAVLink::mavlink)

And pass the local install directory to cmake (adapt to your directory structure):

cd ../my_program
cmake -Bbuild -H. -DCMAKE_PREFIX_PATH=../mavlink/install

Build Warnings

`-Waddress-of-packed-member`

Building the headers may result in warnings like:

mavlink/common/../mavlink_helpers.h:86:24: warning: taking address of packed member of ‘__mavlink_message’ may result in an unaligned pointer value [-Waddress-of-packed-member]
   86 |  crc_accumulate_buffer(&msg->checksum, _MAV_PAYLOAD(msg), msg->len);

The warning indicates the potential for hard faults caused by unaligned access to packed data. This does not happen on most of the common architectures on which MAVLink is run, and generally the warning can be supressed.

You can supress the warnings in CMake using target_compile_options(mavlink_c INTERFACE -Wno-address-of-packed-member -Wno-cast-align).

The issue causes hard faults on Cortex-M0 and other platforms listed here. Please raise issues in ArduPilot/pymavlink if you find other hardware that is affected.

Upgrading Library from MAVLink 1

The MAVLink 1 pre-built library mavlink/c_library_v1 can be upgraded by simply dropping in the MAVLink 2 library from Github: mavlink/c_library_v2.

The MAVLink 2 C library offers the same range of APIs as was offered by MAVLink 1.

The major change from an API perspective is that you don't need to provide a message CRC table any more, or message length table. These have been folded into a single packed table, replacing the old table which was indexed by msgId. That was necessary to cope with the much larger 24 bit namespace of message IDs.

MAVLink 2 usage is covered in the following sections (this includes Working with MAVLink 1 which explains how you can communicate with both MAVLink 2 and MAVLink 1 (only) libraries.

Multiple Streams ("channels")

The C MAVLink library utilizes a "channel" metaphor to allow for simultaneous processing of multiple, independent MAVLink streams in the same program. All receiving and transmitting functions provided by this library require a channel, and it is important to use the correct channel for each operation.

By default up to 16 channels may be defined on Windows, Linux and macOS, and up to 4 channels may be define on other systems. Systems can specify a different maximum number of channels/comms buffers using MAVLINK_COMM_NUM_BUFFERS (for example, this might be reduced to 1 if running MAVLink on very memory constrained hardware).

If only one MAVLink stream exists, channel 0 should be used by specifying the MAVLINK_COMM_0 constant.

Receiving

MAVLink reception/decoding is done in a number of phases:

Parse the incoming stream into objects representing each packet (mavlink_message_t).
Decode the MAVLink message contained in the packet payload into a C struct (that has fields mapping the original XML definition).

These steps are demonstrated below.

Parsing Packets

The mavlink_parse_char(...) convenience function (mavlink_helpers.h) is used to parse incoming MAVLink data. The function parses the data one byte at a time, returning 0 (MAVLINK_FRAMING_INCOMPLETE) as it progresses, and 1 (MAVLINK_FRAMING_OK) on successful decoding of a packet. The r_mavlink_status parameter is updated with the channel status/errors as decoding progresses (you can check mavlink_status_t.msg_received to get the current byte's decode status/error and mavlink_status_t.parse_state for the current parse state). On successful decoding of a packet, the r_message argument is populated with an object representing the decoded packet.

The function prototype and parameters are shown below:

MAVLINK_HELPER uint8_t mavlink_parse_char(uint8_t chan, uint8_t c, mavlink_message_t* r_message, mavlink_status_t* r_mavlink_status)

Parameters:

uint8_t chan: ID of the current channel.
uint8_t c: The char to parse.
mavlink_message_t* r_message: On success, the decoded message. NULL if the message couldn't be decoded.
mavlink_status_t* r_mavlink_status: The channel statistics, including information about the current parse state.

Returns: 0 if the packet decoding is incomplete. 1 if the packet successfully decoded.

The code fragment below shows the typical use of this function when reading data from a serial port (serial):

#include <common/mavlink.h>

mavlink_status_t status;
mavlink_message_t msg;
int chan = MAVLINK_COMM_0;

while(serial.bytesAvailable > 0)
{
  uint8_t byte = serial.getNextByte();
  if (mavlink_parse_char(chan, byte, &msg, &status))
    {
    printf("Received message with ID %d, sequence: %d from component %d of system %d\n", msg.msgid, msg.seq, msg.compid, msg.sysid);
    // ... DECODE THE MESSAGE PAYLOAD HERE ...
    }
}

The mavlink_helpers.h include other parser functions: mavlink_frame_char() and mavlink_frame_char_buffer(). Generally you will want to use mavlink_parse_char() (which calls those functions internally), but reviewing the other methods can give you a better understanding of the parsing process.

Decoding the Payload

The message/packet object retrieved in the previous section(mavlink_message_t) contains fields in the MAVLink packet/serialization format - including the message id (msgid) and the payload (payload64).

To get the fields of the specific message in the packet you need to further decode the payload. This is typically done by providing a case statement that maps the ids of the messages you wish to decode to appropriate decoder functions. The code fragment below shows this for two messages (the first decodes a whole message into a C struct, while the second gets just a single field):

if (mavlink_parse_char(chan, byte, &msg, &status)) {
...

    switch(msg.msgid) {
      case MAVLINK_MSG_ID_GLOBAL_POSITION_INT: // ID for GLOBAL_POSITION_INT
        {
          // Get all fields in payload (into global_position)
          mavlink_msg_global_position_int_decode(&msg, &global_position);

        }
        break;
      case MAVLINK_MSG_ID_GPS_STATUS:
        {
          // Get just one field from payload
          visible_sats = mavlink_msg_gps_status_get_satellites_visible(&msg);
        }
        break;
     default:
        break;
    }

...
}

The decoder/encoder functions and ids for each message in a dialect can be found in separate header files under the dialect folder. The headers are named with a format including the message name (mavlink*msgmessage_name.h*)

Individual message definitions for the dialect are pulled in when you include mavlink.h for your dialect, so you don't need to include these separately.

The most useful decoding function is named with the pattern mavlink*msgmessage_name_decode()*, and extracts the whole payload into a C struct (with fields mapping to the original XML message definition). There are also separate decoder functions to just get the values of individual fields.

For example, the common message GLOBAL_POSITION_INT is generated to common/mavlink_msg_global_position_int.h, and contains the following definitions:

// Message ID number definition
#define MAVLINK_MSG_ID_GLOBAL_POSITION_INT 33

// Function to decode whole message into C struct
static inline void mavlink_msg_global_position_int_decode(const mavlink_message_t* msg, mavlink_global_position_int_t* global_position_int)

// C-struct with fields mapping to original message
MAVPACKED(
typedef struct __mavlink_global_position_int_t {
 uint32_t time_boot_ms; /*< [ms] Timestamp (time since system boot).*/
 int32_t lat; /*< [degE7] Latitude, expressed*/
 int32_t lon; /*< [degE7] Longitude, expressed*/
 int32_t alt; /*< [mm] Altitude (MSL). Note that virtually all GPS modules provide both WGS84 and MSL.*/
 int32_t relative_alt; /*< [mm] Altitude above ground*/
 int16_t vx; /*< [cm/s] Ground X Speed (Latitude, positive north)*/
 int16_t vy; /*< [cm/s] Ground Y Speed (Longitude, positive east)*/
 int16_t vz; /*< [cm/s] Ground Z Speed (Altitude, positive down)*/
 uint16_t hdg; /*< [cdeg] Vehicle heading (yaw angle), 0.0..359.99 degrees. If unknown, set to: UINT16_MAX*/
}) mavlink_global_position_int_t;

// Function to get just a single field from the payload (in this case, the altitude).
static inline int32_t mavlink_msg_global_position_int_get_alt(const mavlink_message_t* msg)

Command Decoding

A MAVLink Command encodes a command defined in a MAV_CMD enum value into a COMMAND_INT or COMMAND_LONG message.

Command packets are parsed and decoded in the same way as any other payload - i.e. you switch on message id of MAVLINK_MSG_ID_COMMAND_INT/MAVLINK_MSG_ID_COMMAND_LONG and call the decoder functions mavlink_msg_command_int_decode()/mavlink_msg_command_long_decode() (respectively) to get a C struct mapping the original message.

The message types differ in that COMMAND_INT has int32 types for parameter fields 6 and 7 (instead of float) and also includes a field for the geometric frame of reference of any positional information in the command.

To decode the specific command you then switch on the value of the mavlink_command_int_t.command or mavlink_command_long_t.command field, which contains the particular MAV_CMD id.

Further interpretation and handling of the message then has to be done manually (there are no convenience functions).

Additional Checks

The library have a number of #define values that you can set to enable various features:

MAVLINK_CHECK_MESSAGE_LENGTH: Enable this option to check the length of each message. This allows invalid messages to be caught much sooner. Use if the transmission medium is prone to missing (or extra) characters (e.g. a radio that fades in and out). Only use if the channel will only contain messages types listed in the headers.

Transmitting

Transmitting messages can be done by using the mavlink_msg_*_pack() function, where one is defined for every message. The packed message can then be serialized with mavlink_helpers.h:mavlink_msg_to_send_buffer() and then writing the resultant byte array out over the appropriate serial interface.

It is possible to simplify the above by writing wrappers around the transmitting/receiving code. A multi-byte writing macro can be defined, MAVLINK_SEND_UART_BYTES(), or a single-byte function can be defined, comm_send_ch(), that wrap the low-level driver for transmitting the data. If this is done, MAVLINK_USE_CONVENIENCE_FUNCTIONS must be defined.

Message Signing

Message Signing (authentication) allows a MAVLink 2 system to verify that messages originate from a trusted source.

The C libraries generated by mavgen provide almost everything needed to support message signing in your MAVLink system. The topic C Message Signing explains the remaining code that a system must implement to enable signing using the C library.

Connection/Heartbeat

The sections above explain how you can send and receive messages. What messages are sent/received depends on the systems that you're working with. The set of messages that most systems can send are documented in common.xml and there are various microservices microservices that you may want to use.

Minimally MAVLink components should implement the HEARTBEAT/Connection protocol as this is used by other systems as proof-of-life for the component, and also for routing.

Working with MAVLink 1

This section explains how to use the MAVLink 2 C library to work with MAVLink 1 systems.

Version Handshaking/Negotiation

MAVLink Versions explains the handshaking used to determine the supported MAVLink version of either end of the channel, and how to negotiate the version to use.

Sending and Receiving MAVLink 1 Packets

The MAVLink 2 library will send packets in MAVLink 2 framing by default. To force sending MAVLink 1 packets on a particular channel you change the flags field of the status object.

For example, the following code causes subsequent packets on the given channel to be sent as MAVLink 1:

mavlink_status_t* chan_state = mavlink_get_channel_status(MAVLINK_COMM_0);
chan_state->flags |= MAVLINK_STATUS_FLAG_OUT_MAVLINK1;

Incoming MAVLink 1 packets will be automatically handled as MAVLink 1. If you need to determine if a particular message was received as MAVLink 1 or MAVLink 2 then you can use the magic field of the message. In c programming, this is done like this:

if (msg->magic == MAVLINK_STX_MAVLINK1) {
   printf("This is a MAVLink 1 message\n");
}

In most cases this should not be necessary as the XML message definition files for MAVLink 1 and MAVLink 2 are the same, so you can treat incoming MAVLink 1 messages the same as MAVLink 2 messages.

MAVLink 1 is restricted to message IDs less than 256, so any messages with a higher message ID won't be received as MAVLink 1.

It is advisable to switch to MAVLink 2 when the communication partner sends MAVLink 2 (see Version Handshaking). The minimal solution is to watch incoming packets using code similar to this:

if (mavlink_parse_char(MAVLINK_COMM_0, buf[i], &msg, &status)) {

    // check if we received version 2 and request a switch.
    if (!(mavlink_get_channel_status(MAVLINK_COMM_0)->flags & MAVLINK_STATUS_FLAG_IN_MAVLINK1)) {
        // this will only switch to proto version 2
        chan_state->flags &= ~(MAVLINK_STATUS_FLAG_OUT_MAVLINK1);
    }
}

Examples

The following examples show the use of the API.

UART Interface: Simple C example of a MAVLink to UART interface for Unix-like systems.
UDP Example: Simple C example of a MAVLink UDP interface for Unix-like systems (Linux, MacOS, BSD, etc.).

In addition, the C library is used in numerous open source systems:

C (mavgen)