CIKA - Autonomous Mobile Manipulation Robot (AMMR) for Waste Collection and Sorting

Project Structure

amr_ws/src/
├── cika_description/        # URDF, meshes, ros2_control config
├── cika_bringup/            # Top-level launch files
├── cika_navigation/         # Nav2, SLAM, RTAB-Map
└── cika_manipulation/       # MoveIt2, arm planning, gripper
└── cika_perception/         # Object detection, point cloud processing
└── cika_task_manager/       # High-level task execution, state machine, connections to perception and navigation 

Mapping and Navigation

# SLAM mode — builds and saves the map database
ros2 launch cika_bringup cika_full_sim.launch.py mode:=slam gui:=false

# Navigation mode — loads saved map and runs Nav2
ros2 launch cika_bringup cika_full_sim.launch.py mode:=navigation gui:=false

After a mapping run, the database is saved automatically to cika_navigation/maps/cika_map.db. To switch to navigation mode, simply relaunch with mode:=navigation — no manual copying needed.

Warning: launching mode:=slam deletes the existing database on start. Back up cika_warehouse.db before remapping if you want to keep a good map.

cp src/cika_navigation/maps/cika_warehouse.db \
   src/cika_navigation/maps/cika_warehouse_backup.db

Perception

# Simulation — YOLOv8 detector + EfficientNet-B0 classifier, simulated OAK-D Lite
ros2 launch cika_perception perception.launch.py backend:=sim use_sim_time:=true

# Hardware — same pipeline, tuned for Raspberry Pi 4B + physical OAK-D Lite
ros2 launch cika_perception perception.launch.py backend:=hardware use_sim_time:=false

The perception pipeline runs two nodes in sequence:

• detector_node — runs YOLOv8 on the RGB stream, back-projects detections to 3D using the depth image and camera intrinsics, and publishes on /cika/waste_detections
• classifier_node — crops each detection bounding box from the RGB frame, runs EfficientNet-B0 to verify the recyclable / non_recyclable label, and republishes enriched detections on /cika/waste_detections_classified

A debug image with bounding boxes and labels is published on /cika/perception/debug_image whenever a subscriber is present.

Note: backend:=sim and backend:=hardware load different parameter configs (perception_sim.yaml / perception_hardware.yaml) — inference rate is 5 Hz in sim and 2 Hz on hardware to stay within RPi 4B CPU limits.

Task Manager

# Launch task manager (simulation)
ros2 launch cika_task_manager cika_task_manager.launch.py use_sim_time:=true

# Launch task manager (hardware)
ros2 launch cika_task_manager cika_task_manager.launch.py use_sim_time:=false

The task manager bridges perception, navigation, and manipulation through a five-state machine:

• IDLE — listens to /cika/waste_detections_classified and scores candidates by combined detector + classifier confidence
• SELECTING — picks the highest-scoring detection with a valid 3D position and triggers navigation
• NAVIGATING — sends a NavigateToPose goal to Nav2, stopping approach_distance (0.4 m) short of the object; retries once on failure before returning to IDLE
• VERIFYING — confirms the target is still visible over 5 consecutive detection frames (~3 seconds) before committing to a pick
• PICKING — sends a PickAndDispose action goal to cika_manipulator

Visualization (Foxglove Studio)

Foxglove Studio is installed system-wide via apt and launched standalone — no ROS2 launch file is needed.

# Launch Foxglove Studio
foxglove-studio

Once open, connect to your running ROS2 stack via the Foxglove WebSocket bridge:

# Start the bridge (run this alongside your other nodes)
ros2 launch foxglove_bridge foxglove_bridge_launch.xml

Then in Foxglove Studio, select Open Connection → Foxglove WebSocket and enter: ws://<ip_address>:8765

Physical Robot

# Teleop only (default)
ros2 launch cika_bringup cika_real.launch.py nav:=false

# Compute launch
ros2 launch cika_bringup cika_compute.launch.py 

# SLAM — build a map
ros2 launch cika_navigation cika_nav.launch.py mode:=slam use_sim_time:=false

# Navigation — autonomous with saved map
ros2 launch cika_navigation cika_nav.launch.py mode:=navigation use_sim_time:=false

Cika Hardware Arm — How to Run

1. Upload Firmware to ESP32

• Open cika_arm_esp32.ino in Arduino IDE
• Board: ESP32 Dev Module | Port: /dev/ttyUSB1
• Click Upload, then press EN/RESET on ESP32

2. Give Port Permission

sudo chmod a+rw /dev/ttyUSB1

3. Build

cd ~/AMMR/cika_ws
colcon build --packages-select cika_hardware_arm
source install/setup.bash

4. Launch

ros2 launch cika_bringup cika_real_arm.launch.py

pip install sudo apt install xvfb
xvfb-run -a ros2 launch cika_manipulator moveit.launch.pi.py

```bash
ros2 run cika_manipulator paper_client 

5. Verify

ros2 control list_controllers

Expected:

joint_state_broadcaster   active
arm_controller            active

6. Monitor ESP32 (optional)

screen /dev/ttyUSB1 115200
# Close with: Ctrl+A then K → y

---

If Something Goes Wrong

Problem	Fix
Port busy	Close Arduino Serial Monitor or screen
Controllers fail	Unplug and replug ESP32, relaunch
Wrong port	Change ttyUSB1 to ttyUSB0 in cika_hardware_arm.cpp and rebuild