got compression working really well, my laptops wifi card is able to talk to the nao fairly well through the door from the hallway

nao_video_server.py

@@ -3,8 +3,13 @@
 nao_video_server.py - RUNS ON THE ROBOT, under its own Python 2.7.
 
 Opens a qi session to NAOqi over localhost (zero network hop - it's
-the same box) and serves ONE persistent, length-framed TCP stream
-carrying both compressed video (JPEG) and compressed audio (mu-law).
+the same box) and serves TWO persistent, length-framed TCP streams -
+one for compressed video (JPEG), one for compressed audio (mu-law) -
+each on its own port/connection. They used to share one connection,
+but a single big video frame mid-send could block audio behind it for
+its whole transfer time, which is a bad tradeoff for something as
+latency-sensitive as audio. Splitting them means a slow video frame
+can never delay audio.
 
 Why not HTTP: the phone tunnel only forwards specific TCP ports, so
 this still rides over TCP - but a 5-byte binary header has none of
@@ -18,18 +23,20 @@ whole time. Subscribing to ALAudioDevice locally (same zero-hop trick
 as the camera) means only mu-law-compressed bytes ever cross the
 tunnel, half the size of raw PCM.
 
-Wire format (all integers big-endian, via `struct`):
+Wire format, identical on both connections (all integers big-endian,
+via `struct`):
     1 byte   type     0x00 HELLO   0x01 VIDEO   0x02 AUDIO
     4 bytes  length
     N bytes  payload
-        HELLO payload: JSON {"rate": 16000, "channels": 1}
-        VIDEO payload: JPEG bytes
-        AUDIO payload: mu-law encoded 8-bit samples (mono)
+        HELLO payload: JSON, informational only
+        VIDEO payload: JPEG bytes (video connection only)
+        AUDIO payload: mu-law encoded 8-bit samples, mono (audio connection only)
 
 Under load (slow/lossy tunnel), video quality and frame rate step
 down automatically (see QUALITY_TIERS) and recover once sends are
-fast again. Audio is prioritized over video each pass, since gaps in
-audio are far more noticeable than a slightly stale picture.
+fast again - and since encoding is throttled to match, this also cuts
+CPU use on a congested link instead of just wasting it on frames that
+would've been discarded anyway.
 
 Needs the SDK's own env vars, since the bindings aren't on a login
 shell's default path:
@@ -54,7 +61,9 @@ import qi
 
 # ---- tunables ----
 NAOQI_PORT = 9561          # local qi session port (confirmed working)
-TCP_PORT = 8000            # forward this one port through the phone tunnel
+VIDEO_PORT = 8000          # forward both of these through the phone tunnel
+AUDIO_PORT = 8001          # kept on its own connection so a slow video
+                            # frame can never block/delay audio delivery
 CAMERA_RESOLUTION = 1      # 0=kQQVGA(160x120) 1=kQVGA(320x240) 2=kVGA(640x480)
 CAMERA_COLORSPACE = 11     # kRGBColorSpace
 CAMERA_FPS = 20
@@ -209,66 +218,82 @@ class MicGrabber(object):
             pass
 
 
-class MediaServer(object):
-    """Accepts one client at a time and interleaves video + audio onto
-    its socket. Audio is drained first every pass since gaps in it are
-    much more noticeable than the picture being a beat stale."""
+def _send(conn, msg_type, payload):
+    conn.sendall(struct.pack(">BI", msg_type, len(payload)))
+    conn.sendall(payload)
 
-    def __init__(self, grabber, mic):
-        self.grabber = grabber
-        self.mic = mic
 
-    def serve_forever(self, host, port):
+def serve_video(grabber, host, port):
+    """Video gets its own connection/thread - a slow sendall() here
+    (a big JPEG frame over a bad link) must never be able to delay
+    audio, which used to share this same socket and paid for it in
+    latency."""
     sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
     sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
     sock.bind((host, port))
     sock.listen(1)
-        print("media server listening on :%d" % port)
+    print("video server listening on :%d" % port)
     while True:
         conn, addr = sock.accept()
         conn.setsockopt(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)
         conn.settimeout(5.0)
-            print("client connected: %s" % (addr,))
+        print("video client connected: %s" % (addr,))
+        last_jpeg = None
+        last_send = 0.0
         try:
-                self._serve_client(conn)
+            _send(conn, TYPE_HELLO, json.dumps({"stream": "video"}).encode("utf-8"))
+            while True:
+                jpeg = grabber.latest()
+                _, min_interval = grabber.quality_and_interval()
+                now = time.time()
+                if jpeg is not None and jpeg is not last_jpeg and now - last_send >= min_interval:
+                    t0 = time.time()
+                    _send(conn, TYPE_VIDEO, jpeg)
+                    grabber.report_send_time(time.time() - t0)
+                    last_jpeg = jpeg
+                    last_send = now
+                else:
+                    time.sleep(0.01)
         except Exception as e:
-                print("client disconnected (%s)" % e)
+            print("video client disconnected (%s)" % e)
         finally:
             try:
                 conn.close()
             except Exception:
                 pass
 
-    def _serve_client(self, conn):
-        self._send(conn, TYPE_HELLO,
-                    json.dumps({"rate": AUDIO_RATE, "channels": AUDIO_CHANNELS}).encode("utf-8"))
-        last_jpeg = None
-        last_video_send = 0.0
+
+def serve_audio(mic, host, port):
+    """Audio's own connection/thread. Checked eagerly (short sleep,
+    not paced like video) since gaps are far more noticeable than a
+    video frame being a beat stale."""
+    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
+    sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+    sock.bind((host, port))
+    sock.listen(1)
+    print("audio server listening on :%d" % port)
     while True:
-            did_something = False
-
-            if self.mic is not None:
-                for chunk in self.mic.pop_all():
-                    self._send(conn, TYPE_AUDIO, chunk)
-                    did_something = True
-
-            jpeg = self.grabber.latest()
-            _, min_interval = self.grabber.quality_and_interval()
-            now = time.time()
-            if jpeg is not None and jpeg is not last_jpeg and now - last_video_send >= min_interval:
-                t0 = time.time()
-                self._send(conn, TYPE_VIDEO, jpeg)
-                self.grabber.report_send_time(time.time() - t0)
-                last_jpeg = jpeg
-                last_video_send = now
-                did_something = True
-
-            if not did_something:
-                time.sleep(0.01)
-
-    def _send(self, conn, msg_type, payload):
-        conn.sendall(struct.pack(">BI", msg_type, len(payload)))
-        conn.sendall(payload)
+        conn, addr = sock.accept()
+        conn.setsockopt(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)
+        conn.settimeout(5.0)
+        print("audio client connected: %s" % (addr,))
+        try:
+            _send(conn, TYPE_HELLO,
+                  json.dumps({"rate": AUDIO_RATE, "channels": AUDIO_CHANNELS}).encode("utf-8"))
+            while True:
+                chunks = mic.pop_all() if mic is not None else []
+                if chunks:
+                    for chunk in chunks:
+                        _send(conn, TYPE_AUDIO, chunk)
+                else:
+                    time.sleep(0.005)
+        except Exception as e:
+            print("audio client disconnected (%s)" % e)
+        finally:
+            try:
+                conn.close()
+            except Exception:
+                pass
 
 
 def main():
@@ -284,9 +309,12 @@ def main():
     except Exception as e:
         print("mic capture unavailable (%s) - video only" % e)
 
-    server = MediaServer(grabber, mic)
+    video_thread = threading.Thread(target=serve_video, args=(grabber, "0.0.0.0", VIDEO_PORT))
+    video_thread.daemon = True
+    video_thread.start()
+
     try:
-        server.serve_forever("0.0.0.0", TCP_PORT)
+        serve_audio(mic, "0.0.0.0", AUDIO_PORT)
     except KeyboardInterrupt:
         pass
     finally:

naowalk.py

@@ -347,13 +347,140 @@ class LiveMicRelay:
             self._cleanup()
             print("🎙️ Talk relay: stopped")
 
+# ==========================================
+# ON-ROBOT VIDEO/AUDIO RELAY SERVER (auto-deploy)
+# ==========================================
+class VideoServerLauncher:
+    """
+    SCPs nao_video_server.py to the robot's home directory over SFTP and
+    launches it via a persistent, pty-backed SSH session - so the user
+    never has to manually copy the file over or SSH in to start/stop it.
+
+    Uses the same trick as LiveMicRelay: a pty ties the remote process
+    group to the SSH channel, so close() (killing the channel) reliably
+    kills the video server too instead of leaving it orphaned on the
+    robot, still holding VIDEO_PORT/AUDIO_PORT, for next time.
+    """
+    def __init__(self, nao_ip, ssh_user="nao", ssh_password="nao", ssh_port=22,
+                 local_path=None, remote_filename="nao_video_server.py"):
+        self.nao_ip = nao_ip
+        self.ssh_user = ssh_user
+        self.ssh_password = ssh_password
+        self.ssh_port = ssh_port
+        self.local_path = local_path or os.path.join(
+            os.path.dirname(os.path.abspath(__file__)), remote_filename)
+        self.remote_filename = remote_filename
+        self.ready = False
+        self._ssh_client = None
+        self._channel = None
+        self._drain_thread = None
+
+    def deploy_and_start(self):
+        if not HAS_SSH:
+            print("🎥 Video server auto-deploy needs paramiko: pip install paramiko")
+            return False
+        if not os.path.isfile(self.local_path):
+            print(f"🎥 Video server auto-deploy: can't find {self.local_path}")
+            return False
+        try:
+            self._ssh_client = paramiko.SSHClient()
+            self._ssh_client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
+            self._ssh_client.connect(self.nao_ip, port=self.ssh_port, username=self.ssh_user,
+                                      password=self.ssh_password, timeout=10)
+
+            # Best-effort: a previous session that didn't get to call
+            # close() cleanly (laptop lost network, crashed, etc.) can
+            # leave a stale instance squatting on the ports - clear it
+            # before starting a fresh one.
+            try:
+                _, out, _ = self._ssh_client.exec_command(
+                    f"pkill -f {self.remote_filename}", timeout=5)
+                out.channel.recv_exit_status()
+                time.sleep(0.3)
+            except Exception:
+                pass
+
+            sftp = self._ssh_client.open_sftp()
+            try:
+                remote_home = sftp.normalize(".")  # robot's home dir
+                remote_path = remote_home + "/" + self.remote_filename
+                print(f"🎥 Copying {self.remote_filename} -> {self.nao_ip}:{remote_path}")
+                sftp.put(self.local_path, remote_path)
+            finally:
+                sftp.close()
+
+            self._channel = self._ssh_client.get_transport().open_session()
+            self._channel.get_pty()
+            remote_cmd = (
+                "export PYTHONPATH=/opt/aldebaran/lib/python2.7/site-packages && "
+                "export LD_LIBRARY_PATH=/opt/aldebaran/lib && "
+                f"python2 {remote_path}"
+            )
+            self._channel.exec_command(remote_cmd)
+
+            self._drain_thread = threading.Thread(target=self._drain_output, daemon=True)
+            self._drain_thread.start()
+
+            time.sleep(1.0)  # give it a moment to bind before the client tries to connect
+            if self._channel.exit_status_ready():
+                print("🎥 Video server exited immediately - check the log lines above")
+                self.ready = False
+                return False
+
+            self.ready = True
+            print("🎥 Video server deployed and running on robot")
+            return True
+        except Exception as e:
+            print(f"🎥 Video server auto-deploy failed: {e}")
+            self.ready = False
+            return False
+
+    def _drain_output(self):
+        """Keeps reading the remote process's stdout/stderr so its pty
+        buffer never fills up and stalls the server, and so its own
+        status prints (quality tier changes, client connects) surface
+        locally too."""
+        chan = self._channel
+        while chan and not chan.closed:
+            try:
+                got = False
+                if chan.recv_ready():
+                    data = chan.recv(4096)
+                    if data:
+                        print(f"🎥 [nao_video_server] {data.decode('utf-8', 'replace').rstrip()}")
+                        got = True
+                if chan.recv_stderr_ready():
+                    data = chan.recv_stderr(4096)
+                    if data:
+                        print(f"🎥 [nao_video_server] {data.decode('utf-8', 'replace').rstrip()}")
+                        got = True
+                if not got:
+                    time.sleep(0.1)
+            except Exception:
+                break
+
+    def close(self):
+        """Kill the pty-backed channel (which kills the remote python
+        process with it) and tear down the SSH connection. Call once at
+        app shutdown."""
+        if self._channel:
+            try: self._channel.close()
+            except: pass
+            self._channel = None
+        if self._ssh_client:
+            try: self._ssh_client.close()
+            except: pass
+            self._ssh_client = None
+        self.ready = False
+
+
 # ==========================================
 # MAIN TELEOP
 # ==========================================
 class NaoTeleop:
     def __init__(self, session, nao_ip, audio_device_index=None, mic_channel=1, mic_gain=8.0, video_scale=2.0,
                  nao_ssh_user="nao", nao_ssh_password="nao", nao_ssh_port=22, relay_port=5555, relay_rate=16000,
-                 video_fps=25, media_relay=None):
+                 video_fps=25, media_relay=None, deploy_video_server=True, video_server_path=None):
         self.session = session
         self.nao_ip = nao_ip
         self.motion = session.service("ALMotion")
@@ -419,6 +546,17 @@ class NaoTeleop:
         self._movement_stop_event = threading.Event()
         self._movement_thread = None
 
+        # If we're using the on-robot media relay, deploy+launch
+        # nao_video_server.py over SSH now (before _init_video_maxfps
+        # tries to connect to it) instead of requiring it to already be
+        # running manually.
+        self.video_server_launcher = None
+        if self.media_relay and deploy_video_server:
+            self.video_server_launcher = VideoServerLauncher(
+                nao_ip, ssh_user=nao_ssh_user, ssh_password=nao_ssh_password,
+                ssh_port=nao_ssh_port, local_path=video_server_path)
+            self.video_server_launcher.deploy_and_start()
+
         self._init_audio_device()
         self._init_video_maxfps()
         if HAS_AUDIO:
@@ -451,8 +589,8 @@ class NaoTeleop:
         if self.media_relay:
             self.video = None
             self.video_client = "relay"  # truthy sentinel, unsubscribe() is a no-op for this path
-            print(f"✅ Camera + mic via on-robot media relay: {self.media_relay}")
-            threading.Thread(target=self._media_relay_loop, daemon=True).start()
+            print(f"✅ Camera via on-robot media relay: {self.media_relay}")
+            threading.Thread(target=self._video_relay_loop, daemon=True).start()
             return
         try:
             self.video = self.session.service("ALVideoDevice")
@@ -482,13 +620,13 @@ class NaoTeleop:
             except Exception:
                 time.sleep(0.05)
 
-    def _media_relay_loop(self):
+    def _video_relay_loop(self):
         """Same role as _video_loop, but speaks the lean length-framed
-        TCP protocol from nao_video_server.py instead of pulling an
-        HTTP MJPEG stream - and now demuxes BOTH video and mu-law-
-        compressed mic audio off the same connection, since the mic
-        subscription lives on the robot too now instead of shipping
-        raw PCM through the qi RPC session."""
+        TCP protocol from nao_video_server.py instead of pulling raw
+        frames over the qi session. Video-only connection - it used to
+        share one socket with audio, but a slow video send could then
+        block audio behind it, so they're split (see the audio port,
+        media_relay port + 1, handled by _audio_relay_loop)."""
         host, _, port_str = self.media_relay.partition(":")
         port = int(port_str) if port_str else 8000
         while not self._video_stop_event.is_set():
@@ -497,24 +635,50 @@ class NaoTeleop:
                 sock = socket.create_connection((host, port), timeout=5)
                 sock.settimeout(5.0)
                 sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)
-                print(f"✅ Media relay connected: {host}:{port}")
+                print(f"✅ Video relay connected: {host}:{port}")
                 while not self._video_stop_event.is_set():
                     msg_type, length = struct.unpack(">BI", _recv_exact(sock, 5))
                     payload = _recv_exact(sock, length) if length else b""
-
                     if msg_type == TYPE_VIDEO:
                         arr = np.frombuffer(payload, dtype=np.uint8)
                         frame = cv2.imdecode(arr, cv2.IMREAD_COLOR)
                         if frame is not None:
                             self._latest_frame = cv2.resize(frame, (320, 240), interpolation=cv2.INTER_NEAREST)
                             self._latest_frame_ts = time.time()
-                    elif msg_type == TYPE_AUDIO:
-                        if getattr(self, "sound_receiver", None):
+                    # TYPE_HELLO: nothing to do with it yet, just skip.
+            except Exception as e:
+                print(f"video relay error: {e}")
+            finally:
+                if sock:
+                    try: sock.close()
+                    except: pass
+            if not self._video_stop_event.is_set():
+                time.sleep(0.5)
+
+    def _audio_relay_loop(self):
+        """Mic audio's own connection - see _video_relay_loop for why
+        this is split off rather than sharing video's socket. Connects
+        to media_relay's host:port+1 by convention (matches AUDIO_PORT
+        = VIDEO_PORT + 1 in nao_video_server.py)."""
+        host, _, port_str = self.media_relay.partition(":")
+        video_port = int(port_str) if port_str else 8000
+        port = video_port + 1
+        while not self._video_stop_event.is_set():
+            sock = None
+            try:
+                sock = socket.create_connection((host, port), timeout=5)
+                sock.settimeout(5.0)
+                sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)
+                print(f"✅ Audio relay connected: {host}:{port}")
+                while not self._video_stop_event.is_set():
+                    msg_type, length = struct.unpack(">BI", _recv_exact(sock, 5))
+                    payload = _recv_exact(sock, length) if length else b""
+                    if msg_type == TYPE_AUDIO and getattr(self, "sound_receiver", None):
                         pcm = decode_ulaw(np.frombuffer(payload, dtype=np.uint8))
                         self.sound_receiver.feed(pcm.tobytes())
-                    # TYPE_HELLO carries {"rate", "channels"} for future use - nothing to do with it yet.
+                    # TYPE_HELLO: nothing to do with it yet, just skip.
             except Exception as e:
-                print(f"media relay error: {e}")
+                print(f"audio relay error: {e}")
             finally:
                 if sock:
                     try: sock.close()
@@ -533,11 +697,12 @@ class NaoTeleop:
         try:
             # Playback machinery is needed either way - only the source
             # of audio chunks differs (qi pub/sub below, vs mu-law
-            # chunks arriving over the media relay in _media_relay_loop).
+            # chunks arriving over the audio relay connection).
             self.sound_receiver = SoundReceiver(self.audio_device_index, self.mic_gain,
                                                  pyaudio_instance=self._pyaudio)
             if self.media_relay:
-                print(f"✅ Mic stream ready (gain={self.mic_gain}x, via media relay - compressed)")
+                print(f"✅ Mic stream ready (gain={self.mic_gain}x, via media relay - compressed, own connection)")
+                threading.Thread(target=self._audio_relay_loop, daemon=True).start()
                 return
             self.session.registerService(self.audio_service_name, self.sound_receiver)
             time.sleep(0.5)
@@ -970,6 +1135,8 @@ class NaoTeleop:
             if getattr(self, "mic_relay", None):
                 self._safe("mic relay stop", self.mic_relay.stop)
                 self._safe("mic relay close", self.mic_relay.close)
+            if getattr(self, "video_server_launcher", None):
+                self._safe("video server stop", self.video_server_launcher.close)
             if self.audio_device and self._qi_audio_subscribed:
                 self._safe("audio unsubscribe", lambda: self.audio_device.unsubscribe(self.audio_service_name))
             if getattr(self, "sound_receiver", None):
@@ -998,8 +1165,11 @@ if __name__ == "__main__":
                               "(e.g. 10-12) when tunneling over a constrained link. "
                               "Ignored when --media-relay is set (the server controls fps).")
     parser.add_argument("--media-relay", type=str, default=None,
-                         help="host:port of the on-robot media relay (nao_video_server.py), "
-                              "e.g. 127.0.0.1:8000 when tunneled through the phone. When set, "
+                         help="host:port to reach the on-robot media relay (nao_video_server.py) "
+                              "at, e.g. 127.0.0.1:8000 when tunneled through the phone. Setting "
+                              "this also makes naowalk auto-scp nao_video_server.py to the "
+                              "robot's home dir over SSH and launch it (see --no-deploy-video-"
+                              "server to skip that and connect to one you started yourself). "
                               "BOTH video and mic audio come compressed over this one TCP "
                               "connection (JPEG + mu-law) instead of raw frames/PCM over the "
                               "qi session - use this on a tunneled/shoddy link. --mic-channel "
@@ -1016,6 +1186,13 @@ if __name__ == "__main__":
                          help="TCP port used to stream mic audio to the robot's nc listener")
     parser.add_argument("--relay-rate", type=int, default=16000,
                          help="Sample rate for the live mic relay - try 48000 if audio sounds sped up/slow")
+    parser.add_argument("--no-deploy-video-server", action="store_true",
+                         help="Don't auto-scp/launch nao_video_server.py on the robot when "
+                              "--media-relay is set - use this if you're already running it "
+                              "yourself and just want naowalk to connect to it.")
+    parser.add_argument("--video-server-path", type=str, default=None,
+                         help="Local path to nao_video_server.py to deploy (default: the "
+                              "copy sitting next to naowalk.py)")
     args = parser.parse_args()
 
     session = qi.Session()
@@ -1037,4 +1214,6 @@ if __name__ == "__main__":
               nao_ssh_password=args.nao_ssh_password,
               nao_ssh_port=args.nao_ssh_port,
               relay_port=args.relay_port,
-              relay_rate=args.relay_rate).run()
+              relay_rate=args.relay_rate,
+              deploy_video_server=not args.no_deploy_video_server,
+              video_server_path=args.video_server_path).run()

start.sh

@@ -1,4 +1,5 @@
 #!/bin/bash
 pactl set-source-volume alsa_input.pci-0000_00_1b.0.analog-stereo 32%
 #python3.11 naowalk.py --ip spike.local --port 9561 --mic-gain 9 --video-scale 3.0 
-python3.11 naowalk.py --ip 127.0.0.1 --port 9561 --mic-gain 9 --video-scale 3.0 --nao-ssh-port 2222 --media-relay 127.0.0.1:8000
+#python3.11 naowalk.py --ip 127.0.0.1 --port 9561 --mic-gain 9 --video-scale 3.0 --nao-ssh-port 2222 --media-relay 127.0.0.1:8000
+python3.11 naowalk.py --ip spike.local --port 9561 --mic-gain 9 --video-scale 3.0 --nao-ssh-port 22 --media-relay spike.local:8000