Workshop 4 Preview: Understanding yDx.M - What Yaanendriya Adds
Part 4 of 4: From DIY VIO to Production-Grade IMU Solutions
ros2
imu
perception
workshop
roscon-india
yaanendriya
ydxm
The Journey Complete
Over the past three posts, we experienced the full problem domain with real measurements:
| Part | Experience | Key Measurement |
|---|---|---|
| Part 1 | IMU Alone | 12cm drift in 5s, yaw drifts 2-5°/min |
| Part 2 | Vision Alone | 100% frames lost on fast motion & textureless |
| Part 3 | VIO Fusion | Only 15% lost, auto-recovery works! |
Now we understand WHY sophisticated IMU solutions matter - we measured the problems ourselves.
This final post introduces Yaanendriya and their yDx.M module - the professional solution to problems we experienced firsthand.
About Yaanendriya Pvt. Ltd.
Company Profile
┌─────────────────────────────────────────────────────────────────────────┐
│ YAANENDRIYA PVT. LTD. │
│ Indigenizing India's Sensor Ecosystem │
│ │
│ Founded: 2023 │
│ Location: Bengaluru, Karnataka, India │
│ Incubation: ARTPARK (AI & Robotics Technology Park) │
│ at Indian Institute of Science (IISc) Bangalore │
│ │
│ Support: Ministry of Heavy Industries │
│ Department of Science & Technology │
│ │
│ Mission: Building indigenous sensor systems for │
│ autonomous vehicles, drones, and robots │
│ │
└─────────────────────────────────────────────────────────────────────────┘ARTPARK Connection
ARTPARK (AI & Robotics Technology Park) is a unique initiative at IISc Bangalore that:
- Translates deep-tech research into commercial products
- Incubates startups focused on robotics and AI
- Connects academia with industry
- Receives government and industry funding
Yaanendriya emerged from this ecosystem, bringing academic research to production-grade products.
Their Product Line
| Product | Description | Use Case |
|---|---|---|
| yDx.M | Inertial Sensing Module (AHRS) | Primary motion estimation |
| Syncore Nano | Navigation stack controller | Drone/robot autopilot |
| YPS M9N | GNSS module (u-blox M9) | Outdoor positioning |
yDx.M: The Workshop’s Star
What is yDx.M?
yDx.M is an Attitude and Heading Reference System (AHRS) - a calibrated, integrated IMU solution that outputs orientation directly.
┌─────────────────────────────────────────────────────────────────────────┐
│ yDx.M ARCHITECTURE │
│ │
│ ┌─────────────────────────────────────────────────────────────────┐ │
│ │ SENSOR ARRAY │ │
│ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ │
│ │ │ Accelerometer│ │ Gyroscope │ │ Magnetometer │ │ │
│ │ │ ±2/4/8/16G │ │ 450-2000°/s │ │ (optional) │ │ │
│ │ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘ │ │
│ └──────────┼────────────────┼────────────────┼───────────────────┘ │
│ │ │ │ │
│ └────────────────┼────────────────┘ │
│ ▼ │
│ ┌─────────────────────────────────────────────────────────────────┐ │
│ │ ONBOARD PROCESSING │ │
│ │ ┌──────────────────────────────────────────────────────────┐ │ │
│ │ │ Factory Calibration + Sensor Fusion Algorithm │ │ │
│ │ │ (Runs on integrated MCU) │ │ │
│ │ └──────────────────────────────────────────────────────────┘ │ │
│ └─────────────────────────────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌─────────────────────────────────────────────────────────────────┐ │
│ │ OUTPUT │ │
│ │ • Calibrated acceleration (m/s²) │ │
│ │ • Calibrated angular velocity (rad/s) │ │
│ │ • Computed orientation (quaternion/Euler) ← Key difference! │ │
│ │ • Temperature compensated │ │
│ └─────────────────────────────────────────────────────────────────┘ │
│ │
│ Interface: I2C / UART / SPI │
│ DOF Options: 6-DOF, 9-DOF, 10-DOF │
│ │
└─────────────────────────────────────────────────────────────────────────┘Specifications
| Feature | yDx.M (6-DOF) | yDx.M (9-DOF) | yDx.M (10-DOF) |
|---|---|---|---|
| Accelerometer | ±2/4/8/16G | ±2/4/8/16G | ±2/4/8/16G |
| Gyroscope | 450-2000°/s | 450-2000°/s | 450-2000°/s |
| Magnetometer | ❌ | ✅ | ✅ |
| Barometer | ❌ | ❌ | ✅ |
| Yaw Reference | Relative | Absolute (N) | Absolute (N) |
| Altitude | ❌ | ❌ | Barometric |
| Calibration | Factory | Factory | Factory |
Unique Feature: Distributed Sensing
┌─────────────────────────────────────────────────────────────────────────┐
│ DISTRIBUTED SENSOR NETWORK │
│ │
│ Traditional Approach: yDx.M Approach: │
│ ────────────────────── ───────────────── │
│ ┌───────┐ ┌───────┐ ┌───────┐ ┌───────┐ │
│ │ Single│ │ yDx.M │ │ yDx.M │ │ yDx.M │ │
│ │ IMU │ │ #1 │ │ #2 │ │ #3 │ │
│ └───┬───┘ └───┬───┘ └───┬───┘ └───┬───┘ │
│ │ │ │ │ │
│ ▼ └─────────┼─────────┘ │
│ Single point ▼ │
│ of failure ┌─────────────────┐ │
│ │ Fusion & voting │ │
│ │ (redundancy) │ │
│ └─────────────────┘ │
│ │
│ Benefits: │
│ • Fault tolerance (sensor failure doesn't crash system) │
│ • Improved accuracy (multiple measurements averaged) │
│ • Vibration rejection (distributed sampling) │
│ • Flexible placement (sensors where needed) │
│ │
└─────────────────────────────────────────────────────────────────────────┘D435i DIY vs yDx.M Production
What We Built (D435i)
# Our DIY VIO stack
realsense2_camera # Raw IMU data
↓
imu_filter_madgwick # Compute orientation
↓
rtabmap_ros # VIO + SLAM
↓
robot_localization # EKF fusion (optional)What yDx.M Provides
# yDx.M integrated solution
yDx.M hardware # Calibrated sensors + onboard AHRS
↓
yDx.M ROS 2 driver # Direct orientation output
↓
Your VIO/SLAM stack # Integration pointFeature Comparison
| Feature | D435i DIY | yDx.M Production |
|---|---|---|
| Raw IMU | ✅ | ✅ |
| Orientation Output | Need Madgwick | ✅ Built-in |
| Magnetometer | ❌ | ✅ (9/10-DOF) |
| Barometer | ❌ | ✅ (10-DOF) |
| Factory Calibration | ❌ Manual | ✅ Factory |
| Distributed Sensing | ❌ | ✅ Network of modules |
| Temperature Compensation | ❌ | ✅ |
| Form Factor | Camera module | Standalone board |
| Interface | USB (via RealSense) | I2C/UART/SPI |
| Price Point | ~$250 (whole camera) | TBD |
When to Use Which
┌─────────────────────────────────────────────────────────────────────────┐
│ CHOOSING YOUR IMU SOLUTION │
│ │
│ D435i IMU (Our DIY Approach) │
│ ─────────────────────────────── │
│ Best for: │
│ • Learning and prototyping │
│ • Projects where camera is primary sensor │
│ • Cost-sensitive applications │
│ • Non-critical navigation │
│ │
│ Limitations: │
│ • Manual calibration required │
│ • No absolute heading (no magnetometer) │
│ • Single point of failure │
│ • USB interface only │
│ │
│ ─────────────────────────────────────────────────────────────────── │
│ │
│ yDx.M (Production Approach) │
│ ─────────────────────────────── │
│ Best for: │
│ • Production robotics │
│ • Drones and UAVs │
│ • Applications requiring absolute heading │
│ • Safety-critical systems (distributed sensing) │
│ • Indian manufacturing ecosystem │
│ │
│ Advantages: │
│ • Factory calibrated (deploy immediately) │
│ • Magnetometer for absolute yaw │
│ • Multiple sensor support │
│ • Low-level interface flexibility │
│ │
└─────────────────────────────────────────────────────────────────────────┘Workshop Exercise Mapping
How Our Preparation Maps to Workshop
| Workshop Exercise | Our Preparation | What Workshop Adds |
|---|---|---|
| 1. IMU Data Acquisition | ✅ Part 1: Experiments 1-3 | yDx.M driver specifics |
| 2. IMU Filtering | ✅ Part 1: Experiments 4-7 | yDx.M built-in AHRS |
| 3. Sensor Fusion | ✅ Part 3: Experiments 11-12 | yDx.M fusion features |
| 4. Inertial-Aided SLAM | ✅ Part 3: Experiments 13-14 | yDx.M + SLAM integration |
| 5. End-to-End Pipeline | ✅ Complete VIO pipeline | yDx.M production deployment |
What We Know vs What We’ll Learn
┌─────────────────────────────────────────────────────────────────────────┐
│ KNOWLEDGE FRAMEWORK │
│ │
│ What We Now Understand (From Parts 1-3): │
│ ─────────────────────────────────────────── │
│ ✅ Why raw IMU has no orientation │
│ ✅ How Madgwick filter works │
│ ✅ Why yaw drifts without magnetometer │
│ ✅ How position integration fails │
│ ✅ Why vision fails on fast motion │
│ ✅ Why vision fails on textureless surfaces │
│ ✅ How VIO combines IMU + Vision │
│ ✅ EKF fusion configuration │
│ │
│ What Workshop Will Teach: │
│ ─────────────────────────────────────────── │
│ 🎯 yDx.M ROS 2 driver API │
│ 🎯 Their specific fusion algorithms │
│ 🎯 Distributed sensor configuration │
│ 🎯 Production calibration procedures │
│ 🎯 Best practices for deployment │
│ 🎯 Purchasing and support channels │
│ │
│ Our Advantage: We understand the "why" before learning the "how"! │
│ │
└─────────────────────────────────────────────────────────────────────────┘Questions to Ask at Workshop
Technical Questions
- Driver & Integration
- “What’s the ROS 2 message type for yDx.M output? Standard
sensor_msgs/Imuor custom?” - “Is the ROS 2 driver open-source? Where can we find documentation?”
- “What’s the recommended update rate for different applications?”
- “What’s the ROS 2 message type for yDx.M output? Standard
- Calibration & Accuracy
- “How does factory calibration compare to on-site calibration?”
- “What’s the typical yaw accuracy with the 9-DOF magnetometer?”
- “How do you handle magnetic interference in production robots?”
- Distributed Sensing
- “How do multiple yDx.M modules communicate in a sensor network?”
- “What’s the failure mode if one sensor in the network fails?”
- “Is there a ROS 2 node that handles multi-IMU fusion?”
- SLAM Integration
- “Which SLAM packages have you tested with yDx.M?”
- “Any specific recommendations for IMU preintegration parameters?”
- “How does yDx.M compare to Intel T265 (discontinued) for VIO?”
Business Questions
- Availability
- “How can workshop participants purchase yDx.M modules?”
- “What’s the typical lead time for orders?”
- “Are there educational/maker discounts?”
- Support
- “What documentation and tutorials are available?”
- “Is there a community forum or Discord?”
- “What’s the warranty and support policy?”
Workshop Day Preparation
Hardware Checklist
Software Verification
# Before workshop, verify everything works:
# 1. Container launches
./scripts/launch-container.sh 4
# Should see: workshop4-imu container running
# 2. RealSense with IMU
ros2 launch realsense2_camera rs_launch.py \
enable_gyro:=true enable_accel:=true unite_imu_method:=2
# Should see: Camera topics publishing
# 3. IMU filter
ros2 run imu_filter_madgwick imu_filter_madgwick_node --ros-args \
-r imu/data_raw:=/camera/camera/imu
# Should see: /imu/data with orientation
# 4. RTAB-Map VIO
ros2 launch rtabmap_launch rtabmap.launch.py \
visual_odometry:=true imu_topic:=/imu/data
# Should see: Odometry outputMental Model
Going into the workshop with this understanding:
┌─────────────────────────────────────────────────────────────────────────┐
│ YOUR PREPARATION ADVANTAGE │
│ │
│ Most Attendees: You (After Parts 1-4): │
│ ─────────────────── ───────────────────── │
│ "IMU gives orientation" "Raw IMU needs filtering" │
│ "Just plug and play" "Calibration matters" │
│ "Why is it drifting?" "Integration error, need fusion" │
│ "Vision should work" "Fast motion/textureless fails" │
│ "What's VIO?" "IMU + Vision = robust poses" │
│ │
│ You'll ask insightful questions and absorb advanced concepts! │
│ │
└─────────────────────────────────────────────────────────────────────────┘The Big Picture: Why This Matters
For Indian Robotics Ecosystem
Yaanendriya represents something important:
- Indigenous manufacturing - Not dependent on imports
- Local support - Same timezone, same language
- Cost optimization - Rupee pricing, no forex
- Government backing - Ministry support means stability
For Your Projects
Understanding IMU perception deeply means:
- Better debugging - Know where problems originate
- Architecture decisions - Choose right sensors for your needs
- Integration skills - Connect any IMU to ROS 2 pipelines
- Appreciation - Value what professional solutions provide
Series Conclusion
What We Actually Measured
┌─────────────────────────────────────────────────────────────────────────┐
│ 4-PART JOURNEY - REAL RESULTS │
│ │
│ Part 1: IMU Alone (7 Experiments) │
│ ────────────────────────────────── │
│ • Raw IMU: No orientation (pitch/roll = 0.0000) │
│ • After Madgwick: Tilt detection works! │
│ • Yaw drift: 2-5° per minute without magnetometer │
│ • Position: 12cm drift in 5 seconds → Unusable │
│ │
│ Part 2: Vision Alone (3 Experiments) │
│ ───────────────────────────────────── │
│ • Baseline: ~900 features, ~120 inliers → OK │
│ • Fast shake: 100% frames LOST │
│ • Textureless: 100% LOST, no recovery │
│ │
│ Part 3: VIO Fusion (3 Experiments) │
│ ────────────────────────────────── │
│ • Fast shake: Only 15% frames lost (6.7x better!) │
│ • Textureless: Lost, but AUTO-RECOVERED! │
│ • Combined stress: 40% lost, still usable │
│ │
│ KEY FINDING: VIO transforms UNUSABLE → USABLE │
│ │
│ ═══════════════════════════════════════════════════════════════ │
│ Ready for: "Perception in ROS 2 — IMU-Centric Hands-On Deep Dive" │
│ ═══════════════════════════════════════════════════════════════ │
│ │
└─────────────────────────────────────────────────────────────────────────┘The Learning Philosophy
We didn’t just read about IMU problems - we experienced them:
- Watched orientation stay at zero
- Felt yaw drift away from true heading
- Saw position explode to meters in seconds
- Broke visual odometry with rapid motion
- Stared at a white wall and lost tracking
- Then fixed everything with fusion
This experiential learning means the workshop concepts will click immediately.
Final Checklist
Before Workshop Day
During Workshop
After Workshop
Resources
Yaanendriya
- Company website (search: Yaanendriya Pvt Ltd)
- ARTPARK incubator information
ROS 2 IMU Ecosystem
Background Reading
See You at ROSCon India 2025!
Workshop 4: Perception in ROS 2 — IMU-Centric Hands-On Deep Dive Presenter: Yaanendriya Pvt. Ltd. Date: December 18-20, 2025 Venue: COEP Tech, Pune
May your IMU never drift, and your features always match!