SkillAgentSearch skills...

matlab-connect-mavlink

Establish MAVLink connections between MATLAB and PX4/ArduPilot autopilots

Install / Use

npx skills add matlab/matlab-agentic-toolkit --skill matlab-connect-mavlink

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

90/100

Supported Platforms

Universal

Tags

Our assessment of matlab-connect-mavlink

matlab-connect-mavlink scores 90/100 on our quality scale, 1541st of 4,582 Development & Engineering skills we index (top 34%).

Its SKILL.md is 14 KB long, well organised into 19 sections with 12 code examples: a thorough specification that gives an agent plenty to work with.

With 1,098 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
30/30
Structure
20/20
Description
12/15
Adoption
13/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated 21 days ago, so matlab-connect-mavlink is actively maintained.
  • No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
  • Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

matlab-connect-mavlink compared with similar skills

All 4 of these similar skills score higher than matlab-connect-mavlink; compare them before choosing.

SkillScoreStarsUpdatedFormat
matlab-connect-mavlink (this skill)by matlab901.1k21d agoSKILL.md
ai-job-searchby MadsLorentzen10045.1k1d agoCLAUDE.md
claude-howtoby luongnv8910041.8k6d agoCLAUDE.md
algorithmic-artby anthropics100177.9k14d agoSKILL.md
pptxby anthropics100177.9k14d agoSKILL.md

Frequently asked questions

How do I install matlab-connect-mavlink?
Run npx skills add matlab/matlab-agentic-toolkit --skill matlab-connect-mavlink. The install tabs above show the steps for each supported agent.
Which AI agents does matlab-connect-mavlink work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is matlab-connect-mavlink safe to use?
It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is matlab-connect-mavlink still maintained?
The repository was last updated 21 days ago, so matlab-connect-mavlink is actively maintained.

name: matlab-connect-mavlink description: > Establish MAVLink connections between MATLAB and PX4/ArduPilot autopilots. Use when connecting to a drone, flight controller, or autopilot via MAVLink protocol over UDP. Covers dialect setup, UDP transport, timer-based heartbeat, and client discovery. Use when: "connect to PX4", "MAVLink connection", "heartbeat", "ground control station", "GCS", "connect to ArduPilot", "drone communication", "mavlinkio", "SITL". license: https://www.mathworks.com/content/dam/mathworks/license/pmrl/license.md metadata: author: MathWorks version: "1.0"

MAVLink Connection Setup

Establish a MAVLink UDP connection from MATLAB to a PX4 or ArduPilot flight controller, with MATLAB acting as a ground control station (GCS). This skill encodes the correct protocol sequence and heartbeat pattern that agents consistently get wrong.

When to Use

  • User wants to connect MATLAB to a PX4 or ArduPilot autopilot via UDP
  • User is building a ground control station in MATLAB
  • User needs to set up MAVLink communication with SITL or networked autopilot
  • User asks about heartbeat exchange or client discovery
  • User references mavlinkio, mavlinkdialect, or sendudpmsg

When NOT to Use

  • User wants to read/set parameters, upload missions, or download logs over an already-established connection — these work well without this skill
  • User is working with Simulink MAVLink blocks (different workflow)
  • User only wants to parse a .ulg log file offline — use ulogreader directly
  • User asks about MAVLink message serialization/deserialization only

Workflow

Follow this exact sequence. The order matters — skipping or reordering steps causes silent failures.

1. Create the dialect

dialect = mavlinkdialect("common.xml", 2);

Use "common.xml" for both PX4 and ArduPilot (covers all standard messages). Use "ardupilotmega.xml" only if you need ArduPilot-specific extension messages. The 2 specifies MAVLink protocol version 2.

Available dialects: common.xml, ardupilotmega.xml, standard.xml, minimal.xml

2. Create the MAVLink IO interface

mavlink = mavlinkio(dialect, 'SystemID', 255, 'ComponentID', 1);
  • SystemID 255, ComponentID 1 is the standard GCS identity
  • Do NOT create mavlinksub(mavlink, "HEARTBEAT") for connection verification. mavlinkio has a built-in heartbeat subscriber that feeds listClients(). Creating a manual HEARTBEAT subscriber is redundant and wasteful.

3. Connect UDP transport

connect(mavlink, "UDP", LocalPort=14550);

All transport options are name-value pairs. Never use positional arguments.

Name-value options: LocalPort (default 0), ConnectionName (default "Connection#")

4. Build the GCS heartbeat message

createmsg signature: createmsg(dialect, msgName) — the dialect object is always the first argument, message name string is second. Do NOT pass the mavlinkio object to createmsg.

hbMsg = createmsg(dialect, "HEARTBEAT");
hbMsg.Payload.type(:) = dialect.enum2num("MAV_TYPE", "MAV_TYPE_GCS");
hbMsg.Payload.autopilot(:) = dialect.enum2num("MAV_AUTOPILOT", "MAV_AUTOPILOT_INVALID");
hbMsg.Payload.base_mode(:) = 0;
hbMsg.Payload.custom_mode(:) = 0;
hbMsg.Payload.system_status(:) = 0;  % GCS has no vehicle state

CRITICAL: Always access fields via msg.Payload.fieldname(:) — never msg.fieldname(:). The message struct has a .Payload sub-struct that contains all protocol fields. Writing hbMsg.type(:) = ... fails because type is not a top-level field — it lives at hbMsg.Payload.type.

CRITICAL: Always use (:) indexing on payload field assignments. Writing msg.Payload.type = 6 (without (:)) silently replaces the wire type (uint8) with double, producing corrupted MAVLink packets. The (:) preserves the original data type.

5. Start periodic heartbeat

There are two workflows depending on whether the autopilot is already broadcasting:

Workflow A: Auto-discovery (autopilot already broadcasting heartbeats)

If PX4 SITL configured to broadcast mavlink messages, the autopilot's heartbeats arrive automatically. Poll listClients first, then send heartbeats back to the discovered client:

% Wait for autopilot to appear
timeout = 10;
tic;
discovered = false;
while toc < timeout
    clients = listClients(mavlink);
    if height(clients) > 1
        discovered = true;
        break;
    end
    pause(0.5);
end

if discovered
    % Use SystemID/ComponentID from listClients output
    remoteClient = clients(clients.SystemID ~= 255, :);  % exclude local GCS
    autopilot = mavlinkclient(mavlink, remoteClient.SystemID, remoteClient.ComponentID);
    hbTimer = timer('ExecutionMode', 'fixedRate', 'Period', 1, ...
        'TimerFcn', @(~,~) sendmsg(mavlink, hbMsg, autopilot));
    start(hbTimer);
end

Workflow B: Manual initiation (autopilot not yet broadcasting)

If the autopilot requires GCS heartbeats before it will respond, use sendudpmsg with the autopilot's listening port. For PX4 SITL, find this in the build log: [mavlink] ... on udp port <SITL_PORT> remote port 14550. Ask the user for this port if not known.

sitlHost = "172.x.x.x";  % IP of SITL instance (use "ip a" in WSL to find it)
sitlPort = 18570;         % PX4 SITL listening port (from SITL build log "udp port" line)
hbTimer = timer('ExecutionMode', 'fixedRate', 'Period', 1, ...
    'TimerFcn', @(~,~) sendudpmsg(mavlink, hbMsg, sitlHost, sitlPort));
start(hbTimer);
  • sendudpmsg(io, msg, host, port) sends to a specific UDP endpoint — use when the client is not yet discovered
  • sendmsg(io, msg, client) sends to a discovered client — use after listClients shows the autopilot
  • MAVLink standard heartbeat rate is 1 Hz
  • Do NOT use port 14550 as the remote port — that is the GCS local port. The autopilot's listening port is different (e.g., PX4 SITL -u port).

6. Verify client discovery (Workflow B only)

After starting heartbeat via sendudpmsg, poll until the autopilot responds:

timeout = 10;
tic;
discovered = false;
while toc < timeout
    clients = listClients(mavlink);
    if height(clients) > 1  % local GCS client is always listed
        discovered = true;
        break;
    end
    pause(0.5);
end

if discovered
    disp(clients);
else
    error("Autopilot not discovered within %d seconds.", timeout);
end

listClients(mavlink) uses the built-in heartbeat subscriber — no manual mavlinksub needed. For Workflow A, discovery is already done in Step 5.

7. Clean up when done

stop(hbTimer);
delete(hbTimer);
disconnect(mavlink);

Always stop and delete the timer before disconnecting to prevent orphaned timers.

Key Functions

| Function | Signature | Purpose | |----------|-----------|---------| | mavlinkdialect | (xmlFile, version) | Parse dialect XML, create message definitions | | mavlinkio | (dialect, 'SystemID', N, 'ComponentID', N) | Create I/O interface | | connect | (io, "UDP", LocalPort=N) | Open UDP transport | | mavlinkclient | (io, systemID, componentID) | Create client handle for a remote system | | createmsg | (dialect, msgType) | Create message struct — dialect first, not io | | sendudpmsg | (io, msg, remoteHost, remotePort) | Send message to specific UDP endpoint | | sendmsg | (io, msg) or (io, msg, client) | Send to all or to a discovered client | | listClients | (io) | List all discovered clients (uses built-in subscriber) | | listTopics | (io) | List all received message topics | | listConnections | (io) | List active transport connections | | mavlinksub | (io, topic) or (io, client, topic) | Subscribe to messages | | latestmsgs | (subscriber, count) | Read most recent messages from subscriber | | disconnect | (io) | Close all connections |

Patterns

UDP Connection to PX4 SITL (Auto-Discovery)

PX4 SITL configured to broadcast to GCS port 14550 (check build log for remote port 14550). The autopilot appears in listClients automatically.

% Setup
dialect = mavlinkdialect("common.xml", 2);
mavlink = mavlinkio(dialect, 'SystemID', 255, 'ComponentID', 1);
connect(mavlink, "UDP", LocalPort=14550);

% Build GCS heartbeat
hbMsg = createmsg(dialect, "HEARTBEAT");
hbMsg.Payload.type(:) = dialect.enum2num("MAV_TYPE", "MAV_TYPE_GCS");
hbMsg.Payload.autopilot(:) = dialect.enum2num("MAV_AUTOPILOT", "MAV_AUTOPILOT_INVALID");
hbMsg.Payload.base_mode(:) = 0;
hbMsg.Payload.custom_mode(:) = 0;
hbMsg.Payload.system_status(:) = 0;

% Wait for autopilot to be discovered
timeout = 10;
tic;
while toc < timeout
    clients = listClients(mavlink);
    if height(clients) > 1
        break;
    end
    pause(0.5);
end
disp(clients);

% Start GCS heartbeat back to the discovered autopilot
remoteClient = clients(clients.SystemID ~= 255, :);
autopilot = mavlinkclient(mavlink, remoteClient.SystemID, remoteClient.ComponentID);
hbTimer = timer('ExecutionMode', 'fixedRate', 'Period', 1, ...
    'TimerFcn', @(~,~) sendmsg(mavlink, hbMsg, autopilot));
start(hbTimer);

% ... perform operations ...

% Clean up
stop(hbTimer);
delete(hbTimer);
disconnect(mavlink);

Subscribing to Specific Messages

When you need to monitor a specific message type (beyond connection verification):

% Subscribe to all messages of a type
sub = mavlinksub(mavlink, "GLOBAL_POSITION_INT");

% Subscribe to messages from a specific discovered client
remoteClient = clients(clients.SystemID ~= 255, :);
autopilot = mavlinkclient(mavlink, remoteClient.SystemID, remoteClient.ComponentID);
sub = mavlinksub(mavlink, autopilot, "ATTITUDE");

% Read the latest message(s)
msgs = latestmsgs(sub, 1);
if ~isempty(msgs)
    disp(msgs.Payload);
end

Sending Commands After Discovery

Once a client is discovered via listClients, you can use sendmsg with the client:

% After discovery, send to specific client
clients = listClients(mavlink);
if height(clients) > 1
    remoteClient = clients(clients.SystemID ~= 255, :);
    autopilot = mavlinkclient(mavlink, remoteClient.SystemID, remoteClient.ComponentID);

    % Create command message
    cmdMsg = createmsg(dialect, "COMMAND_LONG");
    cmdMsg.Payload.target_system(:) = remoteClient.SystemID;
    cmdMsg.Payload.target_component(:) = remoteClient.ComponentID;
    cmdMsg.Payload.command(:) = 400;  % MAV_CMD_COMPONENT_ARM_DISARM
    cmdMsg.Payload.param1(:) = 1;     % arm

    sendmsg(mavlink, cmdMsg, autopilot);
end

Heartbeat Timer with Error Handling

For robust applications, wrap the timer callback to prevent silent failures:

% Using sendudpmsg (pre-discovery, to known SITL endpoint)
hbTimer = timer('ExecutionMode', 'fixedRate', 'Period', 1, ...
    'ErrorFcn', @(~,evt) warning("Heartbeat error: %s", evt.Data.message), ...
    'TimerFcn', @(~,~) sendudpmsg(mavlink, hbMsg, sitlHost, sitlPort));
start(hbTimer);

Gotchas

  • Port 14550 is the GCS local port, not the autopilot's port. When using sendudpmsg, the remote port must be the autopilot's listening port (PX4 SITL -u flag), not 14550. If the autopilot is already discovered via listClients, use sendmsg(io, msg, client) instead — it routes automatically.
  • Use sendudpmsg for pre-discovery messages, not sendmsg. Calling sendmsg(io, msg, client) throws an error if the client hasn't been discovered yet. Use sendudpmsg(io, msg, host, port) for heartbeats and any pre-discovery communication.
  • All message fields live under .Payload. Write msg.Payload.type(:) = ..., never msg.type(:) = .... The top-level message struct contains metadata; protocol fields are always at msg.Payload.fieldname.
  • Always use (:) on payload field assignments. msg.Payload.field(:) = value preserves the wire type (uint8, uint16, int32, etc.). Without (:), MATLAB replaces the field with a double, producing corr

Truncated for display — read the full file on GitHub.

Related Skills

View on GitHub
GitHub Stars1.1k
CategoryDevelopment
Updated21d ago
Forks134

Languages

MATLAB

Trust signals

88/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 medium