High Level Command — C++ Example

Source: rbq_sdk/cpp/example/src/rbq_high_level.cpp

Publishes HighLevelCommand messages at 50 Hz over DDS.
Gait selection is done from the terminal; velocity setpoints come from a joystick.

Overview

Three concurrent actors run together:

Actor	Rate	Role
Control loop thread	50 Hz	Publishes HighLevelCommand with the current gait + velocity
Joystick loop thread	100 Hz	Polls the joystick and updates atomic velocity globals
Main thread (terminal)	blocking	Reads gait names/IDs from stdin and updates the target gait

At startup the control loop also publishes switchControlMode = true once, which enables EXT_JOY mode on the robot stack — without this the HighLevelCommand topic is ignored.

Prerequisites

• RBQ SDK built and installed — see C++ SDK Overview.
• A Linux joystick device at /dev/input/js* (the joystick is required to start the program).
• Robot stack running (simulation or hardware).

Build

From rbq_sdk/cpp/example/:

bash scripts/setup.bash
bash scripts/build.bash

Binary: bin/rbq_high_level

Run

./bin/rbq_high_level <networkInterface> [joystickPath]
# e.g.
./bin/rbq_high_level eth0
./bin/rbq_high_level eth0 /dev/input/js1

The joystick auto-detection follows the same logic as the Sport Client (Joystick) example.

Terminal commands

At the > prompt type either a gait name or a numeric gait ID.

list              # print all available gaits
standing          # switch to STANDING by name
3                 # switch to TROTTING by ID
rl_trot           # switch to RL_TROT by name

Available gaits

Name	ID	Description
sitting	0	Robot sits on the ground
standing	1	Neutral standing posture (default at startup)
aiming	2	Aiming posture
trotting	3	MPC trot gait
trot_stairs	4	Stair-adaptive trot
waving	5	Slow walk gait
trot_running	6	High-speed trot
docking	10	Auto-docking sequence
rl_trot	30	RL-based trot
rl_front_walk	31	RL forward walk
rl_left_walk	33	RL lateral left walk
rl_right_walk	34	RL lateral right walk
rl_bound	35	RL bound gait
rl_pace	36	RL pace gait
rl_pronk	37	RL pronk gait
rl_3leg_hr	38	RL 3-leg (hind-right lifted)
rl_3leg_hl	39	RL 3-leg (hind-left lifted)
rl_3leg_fl	41	RL 3-leg (front-left lifted)
rl_trot_vision	42	RL trot with vision
rl_trot_run	45	RL trot run
rl_silent	46	RL silent gait
rl_stairs	47	RL stair climbing
rl_walk_vision	48	RL walk w/ terrain height feedback (height/pitch adjustable)
rl_walk	49	RL walk (height/pitch adjustable)
rl_end	80	RL end state

Joystick velocity controls

Axis	Maps to	Max value
Left stick Y (forward)	vel_x	±1.2 m/s
Left stick X (lateral)	vel_y	±0.5 m/s
Right stick X (yaw)	omega_z	±60 deg/s

A deadzone of 0.12 is applied before scaling.

Note on omega_z units: HighLevelCommand.omega_z is sent in degrees per second.
The on-robot QuadWalk multiplies by D2R internally, so the SDK side sends raw deg/s values (e.g. 60 deg/s ≈ 1.05 rad/s of yaw).

DDS topics used

Direction	Topic	Message type
Publish	rt/rbq/cmd/highLevel	rbq_msgs::msg::dds_::HighLevelCommand_
Publish	rt/rbq/cmd/switchControlMode	std_msgs::msg::dds_::Bool_

Key implementation notes

• gait_transition flag: Set to true only on the single tick immediately following a gait change, using std::atomic<bool>::exchange(). This ensures the QuadWalk gaitTimer is triggered exactly once rather than continuously.
• switchControlMode: Published once at startup (after a 200 ms delay to let DDS subscribers connect). If not published, the robot ignores HighLevelCommand and stays in joystick mode.
• Timing: The control loop uses clock_nanosleep(CLOCK_REALTIME, TIMER_ABSTIME, ...) for precise 20 ms scheduling. If the loop runs late, it resets timeNext to the current time rather than chasing a drifting target.

See also

• Sport Client
• Sport Client (Joystick)
• Low Level (RL Policy)
• C++ SDK Overview