Nokia: Controller multiprocessing, time chunking, cache safety, tracking

Adds parallel scene processing and performance improvements to Intel
SceneScape 2025.2 controller. No Triton/GPU dependencies.

For a detailed technical walkthrough of all changes with code references,
see docs/controller-enhancements-technical-reference.md

- scene_controller.py: ProcessPoolExecutor per scene, overwrite buffer,
  semaphore admission control, async MQTT publish, crash recovery watchdog
- time_chunking.py: SceneAwareCategoryBuffer with hybrid dispatch (event-
  driven when all cameras arrive, 200ms timer fallback for partial scenes)
- cache_manager.py: thread-safe _fast lookup methods, background HTTP
  refresh thread prevents MQTT callback thread from blocking on HTTP
- ilabs_tracking.py: O(1) object association via pre-built hash maps;
  UUID stability fix includes unreliable/suspended tracks in pruning
- robot_vision: expose getSuspendedTracks/getUnreliableTracks via C++
  bindings (required for UUID stability fix)
- metadata.schema.json: add reid, facemask, color, age, hat, gender fields
- tracker-config.json: retune for 10 FPS operation
- rest_client.py: fix token assignment, url=None guard, rootcert TLS verify
- controller/Dockerfile + Makefile: default to public Ubuntu image;
  Nokia mirror overridable via RUNTIME_OS_IMAGE build arg / env var

Signed-off-by: Mohammed Sufiyan Saqib <mohammed.sufiyan_saqib@nokia.com>

Author: Mohammed Sufiyan Saqib

controller/Dockerfile

@@ -1,10 +1,12 @@
 # SPDX-FileCopyrightText: (C) 2021 - 2025 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
+# Modifications:
+# Nokia VPOD (Emerging Products, BLR), 2026
 
-ARG RUNTIME_OS_IMAGE=ubuntu:24.04@sha256:a08e551cb33850e4740772b38217fc1796a66da2506d312abe51acda354ff061
+ARG RUNTIME_OS_IMAGE=ubuntu:noble-20260113@sha256:cd1dba651b3080c3686ecf4e3c4220f026b521fb76978881737d24f200828b2b
 
 # -------------- Common Base Stage (ported to Ubuntu 24.04) --------------
-FROM ubuntu:24.04@sha256:a08e551cb33850e4740772b38217fc1796a66da2506d312abe51acda354ff061 AS scenescape-common-base-24-04
+FROM ${RUNTIME_OS_IMAGE} AS scenescape-common-base-24-04
 
 # We use root for runtime init. The command in ENTRYPOINT will drop to an unprivileged user.
 # hadolint ignore=DL3002
@@ -113,6 +115,7 @@ SHELL ["/bin/bash", "-o", "pipefail", "-c"]
 
 USER root
 
+# Nokia addition: install ps, htop for easier resource usage / performance measurement inside the container
 RUN : \
     && apt-get update \
     && apt-get install -y --no-install-recommends \
@@ -124,6 +127,7 @@ RUN : \
         netbase \
         python3-pip \
         sudo \
+        procps htop \
     && rm -rf /usr/lib/x86_64-linux-gnu/libLLVM-15.so.1 \
     && rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
 

controller/Makefile

@@ -1,10 +1,17 @@
-# SPDX-FileCopyrightText: (C) 2025 Intel Corporation
+# SPDX-FileCopyrightText: (C) 2025 - 2026 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
+# Modifications:
+# Nokia VPOD (Emerging Products, BLR), 2026
 
 IMAGE := scenescape-controller
-RUNTIME_OS_IMAGE := ubuntu:24.04@sha256:a08e551cb33850e4740772b38217fc1796a66da2506d312abe51acda354ff061
+RUNTIME_OS_IMAGE ?= ubuntu:noble-20260113@sha256:cd1dba651b3080c3686ecf4e3c4220f026b521fb76978881737d24f200828b2b
 TARGET = scenescape-controller-runtime
 
+# PROJECT_RELATIVE_DIRS is needed for GIT_REVISION and GIT_REVISION_SHORT.
+# Some projects might depend on some other directory outside of their "top folder".
+# Set this before including common.mk, do not put these into quotes
+PROJECT_RELATIVE_DIRS:=. ../scene_common
+
 include ../common.mk
 
 .PHONY: test-build

controller/config/tracker-config.json

@@ -1,9 +1,9 @@
 {
-  "max_unreliable_frames": 10,
-  "non_measurement_frames_dynamic": 8,
-  "non_measurement_frames_static": 16,
-  "baseline_frame_rate": 30,
+  "baseline_frame_rate": 10,
+  "max_unreliable_frames": 5,
+  "non_measurement_frames_dynamic": 20,
+  "non_measurement_frames_static": 30,
   "time_chunking_enabled": false,
-  "time_chunking_interval_milliseconds": 50,
+  "time_chunking_interval_milliseconds": 200,
   "suspended_track_timeout_secs": 60.0
 }

controller/requirements-runtime.txt

@@ -1,13 +1,15 @@
-# SPDX-FileCopyrightText: (C) 2025 Intel Corporation
+# SPDX-FileCopyrightText: (C) 2025 - 2026 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 # This file is licensed under Apache 2.0 License.
+# Modifications:
+# Nokia VPOD (Emerging Products, BLR), 2026
 
 mapbox_earcut==1.0.3
 ntplib==0.4.0
 numpy==1.26.4
 open3d-cpu==0.19.0
 opencv-python-headless==4.11.0.86
-orjson==3.11.3
+orjson==3.11.5
 paho-mqtt==2.1.0
 scipy==1.16.1
 shapely==2.1.1
@@ -16,3 +18,4 @@ vdms==0.0.22
 opentelemetry-api==1.38.0
 opentelemetry-sdk==1.38.0
 opentelemetry-exporter-otlp-proto-grpc==1.38.0
+requests==2.32.3

controller/src/controller-cmd

@@ -2,12 +2,37 @@
 
 # SPDX-FileCopyrightText: (C) 2024 - 2025 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
+# Modifications:
+# Nokia VPOD (Emerging Products, BLR), 2026
 
 import argparse
 import os
+import threading
+from http.server import BaseHTTPRequestHandler, HTTPServer
 
 from controller.scene_controller import SceneController
 from controller.observability import metrics, tracing
+from controller.controller_mode import ControllerMode
+
+class HealthCheckHandler(BaseHTTPRequestHandler):
+  def do_GET(self):
+    if self.path == '/healthz':
+      self.send_response(200)
+      self.send_header('Content-type', 'text/plain')
+      self.end_headers()
+      self.wfile.write(b'OK')
+    else:
+      self.send_response(404)
+      self.end_headers()
+
+  def log_message(self, format, *args):
+    pass
+
+def start_health_server(port):
+  server = HTTPServer(('0.0.0.0', port), HealthCheckHandler)
+  thread = threading.Thread(target=server.serve_forever, daemon=True)
+  thread.start()
+  return server
 
 def build_argparser():
   parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
@@ -18,9 +43,6 @@ def build_argparser():
   parser.add_argument("--maxlag", help="Maximum amount of lag in seconds",
                       default=1.0, type=float)
 
-  # FIXME - configure mosquitto to authenticate against REST so that
-  # same user/pass can be used for both REST and MQTT
-  # https://pypi.org/project/django-mqtt/
   parser.add_argument("--broker", default="broker.scenescape.intel.com:1883",
                       help="hostname or IP of MQTT broker, optional :port")
   parser.add_argument("--brokerauth", default="/run/secrets/controller.auth",
@@ -43,19 +65,53 @@ def build_argparser():
   parser.add_argument("--visibility_topic", help="Which topic to publish visibility on."
                       "Valid options are 'unregulated', 'regulated', or 'none'",
                       default="regulated")
+  parser.add_argument("--profile", action="store_true",
+                      help="Enable cProfile profiling of controller (disabled by default)")
+  parser.add_argument("--profile-output", type=str, default="/dev/shm/controller_profile.stats",
+                      help="Output file for profile stats (default: /dev/shm/controller_profile.stats)")
+  parser.add_argument("--reid_config_file", help="JSON file with reid configuration",
+            default=None)
+  parser.add_argument("--analytics-only", dest="analytics_only", action="store_true",
+                      default=os.environ.get("CONTROLLER_ENABLE_ANALYTICS_ONLY", "false").lower() == "true",
+                      help="Run controller in analytics-only mode (tracker disabled)")
+  parser.add_argument("--healthcheck_port", type=int,
+                      default=int(os.environ.get("CONTROLLER_HEALTHCHECK_PORT", "0")),
+                      help="HTTP port for /healthz endpoint (0 disables)")
   return parser
 
 def main():
   args = build_argparser().parse_args()
+  
+  # Initialize profiler if requested
+  profiler = None
+  if args.profile:
+    import cProfile
+    profiler = cProfile.Profile()
+    profiler.enable()
+    print(f"[PROFILER] cProfile enabled, output: {args.profile_output}")
+  
   metrics.init()
   tracing.init()
-  controller = SceneController(args.rewriteBadTime, args.rewriteAllTime,
-                              args.maxlag, args.broker,
-                              args.brokerauth, args.resturl,
-                              args.restauth, args.cert,
-                              args.rootcert, args.ntp, args.tracker_config_file, args.schema_file,
-                              args.visibility_topic, args.data_source)
-  controller.loopForever()
+
+  ControllerMode.initialize(analytics_only=args.analytics_only)
+
+  if args.healthcheck_port > 0:
+    start_health_server(args.healthcheck_port)
+
+  try:
+    controller = SceneController(args.rewriteBadTime, args.rewriteAllTime,
+                                args.maxlag, args.broker,
+                                args.brokerauth, args.resturl,
+                                args.restauth, args.cert,
+                                args.rootcert, args.ntp, args.tracker_config_file, args.schema_file,
+                                args.visibility_topic, args.data_source)
+    controller.loopForever()
+  finally:
+    # Save profile on clean exit
+    if profiler is not None:
+      profiler.disable()
+      profiler.dump_stats(args.profile_output)
+      print(f"[PROFILER] Profile saved to: {args.profile_output}")
 
   return